The effects of state pharmacy drug product selection laws on statin patient generic-to-branded drug switch-backs.

Chressanthis, George A. ; Dahan, Nayla G. ; Fandl, Kevin J. 等

I. Introduction

Debate exists among medical and health policy researchers on the clinical equivalence of generic therapeutic interchange and substitution (Holmes et al. 2011). Generic therapeutic interchange and substitution are defined as follows, based on a report by the Clinical Quality Committee (CQC) of the American College of Cardiology Foundation (ACCF) (Holmes et al. 2011, 1289):

   Generic therapeutic interchange--The act of
   dispensing, with the authorization of the initial
   prescriber, an alternative drug that is believed
   to be therapeutically similar but may be chemically
   different, in a different category, with
   different pharmacokinetic properties. This
   interchange is based on the premise that the
   substituted drug will provide a similar clinical
   efficacy, desired outcome, and safety profile.
   Generic therapeutic substitution--The act of
   dispensing therapeutic interchange without
   the prior authorization of the initial prescriber.

Health policy recommendations by the CQC of the ACCF caution about the consequences to patients from generic therapeutic interchange and substitution (Holmes et al. 2011). The fundamental reason for this caution is that generic therapeutic substitution/interchange involves the switching between two different drugs. Therapeutically switched drugs could have differences in the indication profiles (FDA-approved uses to treat disease conditions supported by clinical evidence), effectiveness of drug therapy (how strong the drugs are to achieve clinical endpoints), and number of side-effects on patients (part of this is dependent on the individual response to specific drug therapies). An example using statin drugs analyzed in this study would be switching a patient from atorvastatin (brand name Lipitor[R]) to simvastatin (brand name Zocor[R]). While both are statin drugs, they have differences in the FDA-approved indication label, are different in their ability to reduce cholesterol, and have differences in their side-effect profile by dosing level (Medical Economics Company 2009; Ohsfeldt 2008). Thus, because generic therapeutic interchange involves the dispensing of a different drug than what a physician intended in consultation with their patient, this type of substitution carries with it greater risks to the patient that is best left to the attending physician and patient to decide. Pharmacists are outside this clinical discussion between the physician and patient. While pharmacists have specific knowledge and expertise about how drugs operate, proper dosing and administration, and potential for drug side effects and interactions with other drugs taken by patients, they do not have the deep clinical knowledge and information of patient history needed to make the best clinical decision for patients. This is why specific recommendations were delivered by the CQC of the ACCF on the effects of pharmacist actions through state drug product selection (DPS) laws that could interfere with clinical intentions decided by physicians that are in the best interests of the patients that are beyond pharmacist knowledge and training (Holmes et al. 2011). While the preceding discussion notes important legal and health policy implications surrounding DPS laws, the published research of an empirical nature is rather sparse on this subject and on average fairly dated. Thus a revisiting of this issue with better data and analysis is critical to see how DPS laws work in practice.

All 50 states and Washington, D.C. have DPS laws. DPS laws aim to control and reduce drug costs by allowing pharmacists to switch from prescribed branded drugs to generic bioequivalent cheaper generics, presumably without adversely affecting healthcare quality (Shrank et al. 2010; Carroll et al. 1987). Pharmacists cannot engage in generic substitution if a physician specifically notes otherwise, or if DPS laws exist that allow the patient to prevent the substitution through what is called "patient consent" actions as to the brand vs. generic drug choice.

Research has revealed that pharmacies accrue greater profit margins dispensing generic drugs (Danzon and Furukawa 2011). This current study does not explore the motivations of pharmacists and their lobbying groups to pass state DPS laws designed to make it easier for pharmacists to engage in generic switching. However, and what is critical to this current study, is whether those economic motivations for pharmacists to dispense generics is interfering with the intention of clinical decisions made by physicians in consultation with their patients as to what is in their best health interest. This issue can be viewed as a principal-agent problem where the patient (principal) and pharmacist (agent) are operating under different objective functions (Ross 1973). Previous health economic analyses have focused on the physician operating as an agent in healthcare and more recently insurance companies (Eggleston 2005; Fuchs 2000; Blomqvist 1991). However, the role of pharmacists as agents in healthcare decisions has largely been ignored. The analysis of generic-to-branded drug switch-backs may be a way to determine the intended and unintended effects of pharmacist agent actions in the area of brand-to-generic drug switching.

Not yet settled are the causes of drug switchbacks. Switch-backs are situations in which patients are prescribed a branded drug after being on a generic drug. Switch-backs can be seen as a potential indicator of clinical failure to achieve desired health outcomes from the original brand-to-generic drug substitution. Like generic substitution, switchbacks are broken down as either bioequivalent (within the same drug molecule, i.e., simvastatin to Zocor[R]) or therapeutic (across different drug molecules, i.e., simvastatin to atorvastatin). Using statin drugs as an example, changes to prescribing policies encouraging utilization of older generic statins such as simvastatin as opposed to patented atorvastatin predictively lowered lower statin drug costs but increased longer-term clinical and economic effects due to increased burden of the onset cardiovascular disease and death using a Markov model (Liew et al. 2012). The reason for this predictive effect is that older statins, like simvastatin, are not as effective in reducing cholesterol newer statins like atorvastatin (Liew et al. 2012). So while drug costs can be saved by moving patients to generic statin drugs, the question remains whether patients are getting the same clinical benefits through this switch (Liew et al. 2012). Less effective statin drugs means the greater likelihood of cardiovascular disease caused by a process called atherosclerosis, or plaque build-up in the walls of arteries, that can increase the risk of heart attack, heart failure, stroke, and other cardiovascular diseases (Liew et al. 2012). Thus, the legal design of DPS laws to encourage switching of more expensive patented branded drugs to cheaper off-patent generic drugs can have unintended health effects to patients but also consequences to the healthcare system in the form of higher treatment costs.

Another argument that could be given to generic-to-branded drug switch-backs is the effect from persuasive activities of pharmaceutical sales and marketing promotion to physicians (detailing). However, the results from this line of research are inconclusive. A commonly cited article on an extensive review of the research done on pharmaceutical promotion was unable to reach any definitive conclusions on the degree that information from pharmaceutical companies has on the frequency, cost, or quality of prescribing (Spurling et al. 2010). The authors noted that data limitations on the reviewed studies prevented a conclusive result because prior research has not analyzed pharmaceutical promotion on health outcomes. However, the authors also pointed out that there was no evidence they could find in support of beneficial effects of pharmaceutical promotion, and as they noted, there was some evidence pointing to negative consequences of promotion. To this last point, a recent article showed potential risks to patient safety in a cross-country comparison study of the negative effects of pharmaceutical sales representative activity to physicians by skewing product discussions more toward benefits while not fully explaining risks (Mintzes et al. 2013). Larkin et al. (2014) suggest that restrictions on detailing at academic medical centers reduced off-label prescribing of antidepressant and antipsychotic drug to children, thus benefitting patients, though no connection to health outcomes was empirically verified. Further, a recent report summarized in JAMA on conflicts of interest in medical education recommends prohibiting access of pharmaceutical sales representatives to academic medical centers (Korn and Carlat 2013). However, pharmaceutical promotion through publications can be an important vehicle for the dissemination of new medical information to physicians and the adoption of drugs that can be beneficial to patients (Majumdar et al. 2003). Further studies have argued that conflict of interest policies designed to ban pharmaceutical sales representatives to academic medical centers or from interacting with physicians are likely to produce greater costs than benefits and that any suggested benefits by advocates of restrictions on detailing to health outcomes has never been produced (Barton et al. 2014; Chressanthis et al. 2013; Lesko et al. 2012; Stell 2009; Huddle 2008; Stossel 2007; Stossel 2005; Stell 2005). Also, recent empirical research empirically demonstrates how increasing access restrictions of pharmaceutical sales representatives to physicians are related to slower prescription responses to new medical information events such as a black-box warning on a drug of potential danger to patients or a new first-in-class drug that can be of potential benefit to patients (Chressanthis et al. 2012). However, the previously cited study is limited in that variations in pharmaceutical sales representative access restrictions to physicians were not tied to changes in patient health outcomes. Finally, the LIS Supreme Court in Sorrell v. IMS Health ruled 6-3 in 2011 that restricting the flow of physician prescriber level data to pharmaceutical companies for the purpose of limiting detailing does not promote public health (Supreme Court of the United States 2011, 2):

   Vermont's law also rests on the illegitimate
   premise that physicians are not making what
   the State considers to be the optimal prescription
   decisions for their patients. The law is
   based on derogatory assumptions about doctors
   and the counterintuitive notion that doctors
   will make better decisions about patient
   health if they are deprived of FDA regulated
   speech about FDA-approved drugs. The First
   Amendment prevents a State from impeding
   the flow of truthful and non-misleading information
   based on such paternalistic assumptions.

Therefore, given the inconclusiveness of the preceding research, this current study will dismiss for now the argument that generic-to-branded drug switch-backs are connected to pharmaceutical sales and marketing practices but rather represent a potential clinical failure in the initial brand-to-generic drug substitution.

Do we see this unintended clinical and economic burden effects from brand-to-generic substitution in other therapy areas? Negative consequences in the treatment of epilepsy, such as the loss of seizure control resulting in the elimination of driving privileges and negative impacts on the quality of life, have been associated with brand-to-generic substitution for some specialized drugs (Berg et al. 2008; Richton-Hewett et al. 1988). Higher generic-to-branded switch-back rates were found for antiepileptic drugs, and higher utilization of medical services and longer lengths of hospital stay associated with brand-to-generic substitution (LeLorier et al. 2008; Andermann et al. 2007).

Despite the preceding discussion, other studies found little evidence of significant differences in clinical outcomes due to generic substitution. One study reported an extensive meta-analysis of seven randomized controlled trials and found no statistical difference in the odds of uncontrolled seizures for patients on generic anti-epileptic medicines as compared with patients on the branded drug medicine (Kesselheim et al. 2010). Another study statistically concluded that the odds ratio did not differ between patients who refilled their prescriptions with brand medicines and patients who refilled their prescriptions with generic medications (Gagne et al. 2010). A systematic review and meta-analysis investigated the clinical equivalence of generic and brand name drugs used in cardiovascular disease (Kesselheim et al. 2008). Generic and brand name cardiovascular drugs were similar in almost all clinical outcomes, and brand named drugs were not superior to generic drugs. Generic substitution can also have broader social benefits than using patented branded drugs because lower-cost generic drug therapies may produce overall better societal health given limited resource spending on some but not all available health technologies (Simoens 2011). That is, generic substitution can maximize population health by allowing more people access to technologies that provide improved health benefits.

Generic substitutions are typically enforced by managed care plans through various control mechanisms, but may also produce unintended effects (Mark et al. 2009). This study was cited because potentially like DPS laws that are designed to increase generic utilization and reduce drug cost spending, the use of managed care controls may produce similar short-term drug spending reductions, but create unintended longer-term health problems with patients requiring greater medical resource utilization and costs. These control mechanisms include prior authorization (permission must be granted first by the drug plan before a physician can prescribe a drug for a patient), step therapy condition (physicians are given guidelines as to the order of which drugs can be prescribed), and quantity limits (limits placed on the number of units and/or days of therapy physicians can prescribe a drug to a patient). Together, these control mechanisms are used by managed care drug plans to channel physicians into prescribing certain drug therapies for both clinical and cost-management reasons. Clinical and cost-management evidence are reviewed by a pharmacy and therapeutics (P&T) committee to determine which drugs are easier vs. more difficult to prescribe by physicians for given health conditions. Plans that are "high control" mean they aggressively enact and enforce these mechanisms to restrict physician prescribing practices, as opposed to low control plans, that allow for greater flexibility in physician prescribing of drugs. Plan control is non-therapy class specific, meaning that if a plan's view is to be restrictive in the enactment and enforcement of control mechanisms, it will do so across all therapy conditions and not just for specific drug classes.

The other side of control mechanisms is drug access, or the level of co-pay (out-of-pocket) expense a patient must pay to acquire a prescription. Contracting between pharmaceutical companies and managed care plans affect the level of co-pays for specific drugs. By affecting the level of co-pay to the patient, drug access will in turn affect the rate at which patients remain on drug therapy and whether they follow taking the drug in the appropriate manner prescribed (i.e., taking a medication as prescribed once a day or opposed every other day which may be done to reduce patient cost). Empirical evidence shows higher drug co-pays, all things being equal, reduce drug adherence and the appropriateness of drug use, both contributing to a lower likelihood that patients will receive the full benefit of taking their medication as indicated by the product label. When bringing managed care control and access together, how do DPS laws and decisions made by pharmacists operate in such a way to promote greater generic drug utilization in

a manner that may be contrary to the interest of protecting patient health as intended by physicians and the decisions they make with patients?

Further, since for-profit health plans are in the business to make money (even non-profit health plans must avoid making losses), generic substitution for economic reasons by plans will likely accelerate given numerous blockbuster drugs going off-patent through 2020 (DeRuiter and Holston 2012). Managed care drug plans will be under increasing economic pressure to use control mechanisms to lower drug costs as more generic drugs become available. Likewise, greater available of generics means that pharmacists will have greater economic incentives as well to dispense generics given their higher profit margin. However, research suggests care should be taken to ensure treatment decisions are made on clinical guidelines and not solely on cost considerations (Johnston 2010). Thus, analyzing the incremental effect of variations in DPS laws on generic-to-branded drug switch-backs is an important early indicator to see if potential patient health effects exist from initial brand-to-generic therapeutic substitutions over time.

II. Background

Drug product substitution (DPS) laws are the result of a balancing between the interests of the state in protecting consumers from potentially hazardous counterfeit drugs and the interests of consumers in spending as little as possible on their prescription medicines. States began passing anti-substitution laws in the 1950s in an effort to combat the high risk of dangerous counterfeit drugs produced by unscrupulous pharmacists (Federal Trade Commission 1979). As this fear was allayed over time, and as brand name medicines became both more common and more expensive for consumers, beginning in the 1970s states replaced their anti-substitution laws with DPS laws that encouraged the use of generic drugs (Cheng 2008). Today, every state allows such substitution (Cheng 2008).

Pennsylvania began allowing drug substitution by law in 1976 (Penn. Pub. L., [section] 1, 1976). In that Act, the Pennsylvania legislature explained that its purpose was to "permit consumers to secure necessary drugs at the most economical cost consistent with the professional discretion of the purchaser's physician and pharmacist." The Act expressly required pharmacists to substitute a less expensive generic equivalent drug product for a brand name product unless otherwise requested either by the purchaser or the physician. This Act established the requirement to add two lines to the bottom of drug prescription blanks. The first line must contain the words, "substitution permissible" and the second line must contain the words, "do not substitute."

Additional requirements for drug substitution in Pennsylvania include notification to the consumer that the drug provided will be a generic substitute rather than the name brand, and that if the generic substitute is unavailable that the name brand will be provided (Penn. Pub. L., [section] 3, 1976). Also, the law required pharmacists to post a prominent sign in all pharmacies with the following language: "Pennsylvania law permits pharmacists to substitute a less expensive generically equivalent drug for a brand name drug unless you or your physician indicates otherwise." (Penn. Pub. L., [section] 4(a), 1976).

This law was modified to adjust the requirements of the prescription blanks. Since this modification, prescription blanks shall contain only one signature line, under which it shall state, "substitution permissible." (Penn. Controlled Substance, Drug, Device and Cosmetic Act (1974)). Here, the signature of the physician automatically authorized the dispensing of the generic drug equivalent. In order for a physician to force the dispensing of the brand in lieu of the generic, the physician had to include in handwriting either, "brand necessary" or "brand medically necessary" on the form.

The prevalence of state DPS laws as a means to lower costs to consumers for prescription medications that are available in generic form initially had an unexpectedly low impact on pharmacists who had become accustomed to distributing brand name medications (Carroll et al. 1987). This study found that the reason for the unwillingness to substitute generic medicines for brand name equivalents is the lack of incentive for the pharmacist because cost-savings were generally passed on to the consumer. However, a more recent study suggests that rate of substitution is a product of state DPS law design (Shrank 2010). In states with permissive generic substitution laws (the vast majority of states), generic medicines are preferred by pharmacists; however, if those states also maintain a patient consent law requiring permission from the patient before making the generic substitution, substitution rates were far lower. That study concluded that while many pharmacists preferred generic substitutions, many patients did not.

Generic medications were more frequently dispensed as a result of these revised state DPS laws. However, it was the 1984 Hatch-Waxman Act that paved the way for rapid growth in generic pharmaceutical production. Prior to 1984, both brand name and generic drugs had to go through a comprehensive New Drug Application (NDA) process to prove their effectiveness and safety. Generic manufacturers had little incentive to undergo rigorous testing and evaluation for a much smaller portion of drug sale profits. The Hatch-Waxman Act significantly eased the approval process for generic drugs by establishing the Abbreviated NDA (ANDA) process (The Hatch-Waxman Act (1984)). Rather than requiring a generic drug manufacturer to prove safety and effectiveness through clinical trials, the ANDA process only requires the manufacturer to show that the generic drug is bioequivalent to the innovator drug in the following ways: (i) same dosage form, strength, active pharmaceutical ingredients, and route of administration as the brand product, (ii) bioequivalent to the brand product, and (iii) held to the same rigorous standard of review and good manufacturing practices required by the FDA for the brand product.

By creating and passing the ANDA process, the Hatch-Waxman Act removed one of the most significant obstacles to the production and approval of generic drugs (The Hatch-Waxman Act (1984)). The prevalence of state DPS laws that encouraged pharmacists to substitute brand name drugs with generic drugs provided the foundation for an expansive and lucrative generic drug market.

1. Review of DPS Laws

The few-existing studies on the effects of DPS laws generally show they encourage brand to generic substitution as intended (Shrank 2010; Carroll et al. 1987; Clouse et al. 1985; Masson and Steiner 1985; Goldberg et al. 1979; Kushner and Feierman 1978). However, most of the studies are fairly old, meaning they do not take into account more recent changes in the pharmaceutical landscape and the rise of numerous generic drugs available to physicians, the growth of managed care and influences on physician prescribing practices, and economic pressures placed on drug plans to cost drug costs. So an update on the effects of DPS is needed. Further, no studies have examined the effects of DPS laws on the incidence of switchbacks from generic-to-branded drugs using individual patient-level data. The main advantage of using patient-level data is greater precision on the effects of DPS laws where it matters most--on patients. Thus the primary aim of this study is to examine and discuss the effects of various provisions in state DPS laws on switch-backs for a large population of statin patients while statistically controlling for other relevant factors affecting generic substitution and switch-backs.

Several key aspects of the DPS laws impact brand-to-generic substitution (National Association of Boards of Pharmacy 2008). First, are state drug formulary laws, preferred lists of medications developed by healthcare professionals and drawn from the Food and Drug Administration's (FDA) Therapeutic Equivalence List known as the "Orange Book." Some state formularies are negative formularies, meaning pharmacists may substitute for any drug not listed in the formulary, whereas positive formularies permit pharmacists to substitute for any drug listed in the formulary. Second, a two-line prescription (Rx) form is also a factor. The format of the prescription form has shown to have major effects on substitution rates (Hellerstein 1998). Two-line prescription forms have a "DPS Permitted" line on the left. The physician generally signs prescriptions on the right side. One-line prescription forms have the "DPS one line" on the right. To prevent substitution in the two-line prescription format, prescribes sign the "Dispense as written" line on the left. In the case of the one-line prescription format, prescribers must write in their own handwriting other than the signature "Brand Necessary" or "No Substitution" or "Medically Necessary" to prevent substitution, otherwise the pharmacist is given the approval of drug substitution. Third, DPS laws are either permissive or mandatory. Permissive laws generally state pharmacists may substitute, while mandatory laws state pharmacists must substitute a lower-priced bioequivalent drug if the prescription was written for a branded product and the prescriber has not specifically prohibited substitution. Permissive generic substitution laws enacted in certain states give pharmacists more discretion by allowing, but not requiring, pharmacists to substitute generics. Fourth, pharmacists are not always required to pass on drug cost savings from substitution to patients, meaning that pharmacists can dispense a generic change for the branded drug and keep the difference. DPS laws that mandate a pass-on of drug cost savings mean lower costs to patients. The result of lower drug costs to patients means improved patient drug adherence (patients taking their medications as prescribed) and drug affordability (patients being able to pay for prescription drugs). In both cases, lower drug costs to patients mean improved health benefits from taking their medications (Roebuck et al. 2011; Goldman et al. 2007; Hsu et al. 2006; Sokol et al. 2005). Fifth, in some cases, substitution requires explicit patient consent. DPS law data from 2008 used in this study, 40 states and Washington, D.C. require patients to provide consent prior to generic substitution, meaning pharmacists have less influence in which to dispense generic medications (National Association of Boards of Pharmacy 2008).

The key legal and health policy questions posed here in this study resulting from DPS laws are as follows: (1) what is the relationship between variations in state pharmacy DPS laws and brand-to-generic drug substitution, given prior research analyzing pharmacist behavior and greater financial incentives for pharmacies to dispense generic drugs? (Carroll et al. 1987; Danzon and Furukawa 2011) and (2) if greater brand-to-generic substitution exists as a result of state pharmacy DPS laws, then do we see greater switch-backs of generic-to-branded drugs? A few qualifying and explanatory statements are in order as to the objectives addressed in this paper. First, this paper does not directly analyze variations in state pharmacy DPS law and patient health outcomes. There are no health outcome metrics analyzed in this study. Making that connection would require the use of a different database, such as leveraging patient claims data from health plans. However, the analysis of generic-to-branded drug switch-backs is a potential indicator that the original brand-to-generic substitution or generic drug use as a starting therapy did not achieve desired clinical outcomes or produced unwanted side-effects. Economic considerations cannot be the driving force behind generic-to-branded drug switch-backs because switch-backs involve greater costs borne by the patient through high co-pays. Second, caution needs to be exercised on extending results and implications from this paper on the effects of DPS laws from the statin drug class switch-backs and to those results for other drug classes. It is safe to say that drug classes that have a similar distribution of drug technology and patient response to differences in drug therapy to that seen in the statin drug class, then it is plausible that extensions of this study can be made. However, where significant within-class differences in drug technology, side-effect profiles, and individual patient response to each drug exist, DPS laws have the potential for even greater effects as seen through switch-back patterns. Regardless of the preceding qualifying statements, the understanding of DPS laws and their intended and unintended effects on patients will be an issue of growing legal and health policy importance.

III. Methodology

1. Patient-Level Data

Statin patient drug utilization data were gathered from LifeLink[TM] Anonymized Patient Level Data (APLD) from IMS Health (2006-2008). APLD contains information on drugs consumed by anonymized individual patients compliant with privacy requirements of the Health Insurance Portability and Accountability Act of 1996 (HIPAA) "Privacy Rule". The researchers did not deconstruct the patient-level data to determine the identity of any patient. Data from IMS Health is well-respected and trusted from the academic community for the conduct of peer-reviewed research for journal publication, pharmaceutical companies in making critical business decisions, and governmental agencies to make important health policy decisions.

This study on DPS laws on generic-to-branded drug switch-back patterns starts with a previous exploratory data analysis of 17.3 million unique statin patients over the period from March 2006 to June 2008 (Chressanthis et al. 2011). The principal reason for leveraging this prior study is that the concept of switch-backs already received validation as an appropriate measure from an expert pharmaceutical audience, thereby allowing this study to focus on the DPS effects on switch-backs. The second reason is that the study reported here employs statistical techniques not used in the initial study which only applied descriptive statistics to explain general patterns on observed switch-backs. Thus, this current study offers new evidence as to the relative importance of DPS laws relative to other factors that affect switch-backs. Third, the previous study was used a check on the statistical results found here, further validating the findings and legal/health policy implications.

These data captured about 50%-60% of all dispensed statin prescriptions to patients in the United States during the March 2006 to June 2008 time period (Vladutiu 2008). Statin drugs were selected because they treat a large number of patients, generate significant healthcare system drug spending, have a large number of available generics, and experienced a significant amount of brand-to-generic substitution (Chressanthis et al. 2011), especially given the introduction of generic simvastatin in June 2006 (Berenson 2006).

The drugs analyzed in the statin class were divided into the following categories according to patient utilization: (1) patients on branded drugs only (9.7 million), (2) patients started on branded drugs and who later switched to generic drugs (2.2 million), and (3) patients who started on generic drugs (5.4 million). A random sample of patients was drawn composed of 43 million prescriptions from a total of 210 million prescriptions from July 2006 to June 2008 from the original database of 17.3 million patients (Chressanthis et al. 2011). July 2006 was chosen for the start of the data analysis of brand-to-generic drug substitution and generic-to-branded drug switch-backs to allow time for patients to adjust during the transition period of the start of Medicare Part D which began in January 2006. March to June 2006 was used as the baseline to determine whether patients started on a generic drug or started on a branded drug then switched to a generic drug. The 9.7 million patients who were always on branded drugs were deleted from consideration. Of the 5.4 million patients who started on generic drugs, 5.2 million patients stayed on generics. These patients were also deleted from consideration. Therefore, adding up the 2.2 million patients who started on branded drugs but later switched to generics and the 200,000 patients who started on generic drugs but switched to branded drugs, comprised the total sample of patients to 2.4 million from which substitution and switchback prescription transactions were taken. Then the analysis on the effects of DPS laws was limited to transactions that either involved substitution from a branded-to-generic drug, or, a switch-back from a generic-to-branded drug. The analysis of branded-to-generic drug substitution and generic-to-branded drug switch-backs is not relevant if patients were either always on a generic or branded drug respectively.

The final random sample of prescriptions used for statistical analysis resulted in 1,031,172 prescription observations, broken down by 589,379 brand-to-generic drug substitutions and 441,793 generic-to-branded drug switch-backs. These prescriptions came from 339,427 and 186,041 patients, respectively. A total of 397,111 unique patients were engaged in combined generic substitutions and switch-backs because one patient can be involved in both substitutions and switch-backs. The final number of 1,031,172 prescriptions also accounts for deleting observations with missing values for variables in the model.

Finally, the sample encompassed all traditionally-defined statin drugs, including all patented branded drugs and all branded drugs with their generic equivalent available to patients in the 2006-2008 analysis time period. The list of statin drug treatments, from the earliest to the latest entry into the market, are as follows by their chemical names: lovastatin, pravastatin, fluvastatin, simvastatin, atorvastatin, rosuvastatin, and the combination drug ezetimibe/simvastatin. Research shows the newer statin drug therapies are more effective in lowering LDL-C per dosing strength than earlier statins (Medical Economics Company 2009; Ohsfeldt et al. 2008). The relevance of noting the timing of statin drug entry into the market is that earlier drugs are available at lower costs given that their patents have expired and now are available in generic form. This means there are ample generic statin drugs that pharmacists could dispense. However, the drug technology between early vs. later statins exhibit very different effectiveness profiles. Thus, while DPS laws and pharmacist economic incentives provide motivations for dispensing of generics, patients may not be receiving the clinically best drug option for their condition.

2. Modeling

The determinants of statin patient generic-to-branded drug switch-backs, modeled as a categorical dependent variable, are hypothesized to be affected by the following set of independent variables, utilizing prior research published by the author and other studies on DPS laws: (1) DPS state pharmacy laws expressed as a set of categorical measures, (2) patient demographic and health-status attributes, (3) patient method-of-payment, (4) physician attributes, (5) drug attributes, and (6) local managed care plan control (Shrank et al. 2010; Carroll et al. 1987; Clouse et al. 1985; Masson and Steiner 1985; Goldberg et al. 1979; Kushner and Feierman 1978). Table 1 shows the DPS laws analyzed in this study by state for 2008. Table 2 provides a complete description of the set of variables employed in the modeling. IMS Health is a trusted source for pharmaceutical data used by academics, pharmaceutical companies, and government (IMS Institute for Healthcare Informatics 2014).

A brief explanation follows on the reasons behind using each set of explanatory variables. Patient attributes of age, gender, and co-morbidity count represent individual characteristics and health status that could be associated with switch-backs. A higher patient co-morbidity count represents patients with greater health risk and the potential for being more sensitive to branded-to-generic drug substitutions that failed to reach clinical goals. Next, generic-to=branded drug switch-backs involve greater costs to the patient. Patient method-of-payment, using categorical measures that the prescription was paid for by either Medicaid or a third-party plan that includes Medicare Part D, measures the affect on drug plan payments that should make it easier for patients to engage in switch-back since they only have a co-pay charge relative to full-paying cash patients. The physician specialty associated with the patient prescription represents physician groups who are more likely to engage in statin switchbacks, such as, cardiologists and nephrologists (Card_Neph), endocrinologists and diabetologist (Endo_Dia), internal medicine (IM), and all other physicians (Other) relative to primary care physicians (PCPs) as the omitted category. This likewise suggests that PCPs are more likely to prescribe generic statin drugs relative to specialists. The prescription volume (Rx) by a physician represents the effects from a number of factors, all of which are likely to increase switch-backs: (i) the impact of pharmaceutical promotion since drug companies target physicians who have a greater level of Rx volume (promotion such as using sales representatives to call on physicians are generally reserved by companies promoting patented branded drugs), (ii) the degree of experience of the physician since greater Rx volume would translate into seeing more patients, and (iii) the level at which a physician may be familiar with using a greater number of statin drugs. Drug attributes represent product-level characteristics that would be associated with a generic-to-branded switch-back or substitution: (i) use of a lower dose versus higher dose as the omitted category, and (ii) change of molecule versus use by the same drug molecule as the omitted category). Later patented statin drugs are more effective in reducing cholesterol levels at a lower dose than older less effective generic statins. Likewise, generic-to-branded drug switch-backs are more likely associated with a change in molecule from older generic statin drugs to later patented branded statin drugs. Lastly, the local managed care control environment by plans in the metropolitan area associated per patient drug utilization captured potential barriers (plan control mechanisms) that make it more difficult for physicians to prescribe patented branded drugs. Local managed care plan control was defined as the ability of a drug plan to move product market share in a metropolitan area relative to the national average. Drug plans use various control mechanisms, as described earlier, to manage what drugs physicians can prescribe and encourage patients toward greater generic utilization (Mark et al. 2009). All control effects across drug plans for patients per physician were aggregated per metropolitan area to develop a local area managed care control measure for the area which the patient resided. The statin and proton pump inhibitor (PPI) markets were used to measure managed care control using data obtained through a pharmaceutical company. Local managed care plan control is not a therapy-class specific metric (i.e., if a drug plan is very controlling, this behavior should be seen across all drug classes). When defining local managed care plan control levels, we chose drug classes that affected many physicians and their patients, were consistent with the patients being studied in this research, and where we had data that were available to us to construct this metric.

3. Data Limitations

Measures were not available on the following attributes that represent areas for future research which will be discussed later: (i) pharmacist or pharmacy characteristics where patients received

their dispensed drugs, (ii) information of what physicians initially prescribed versus what prescription was finally dispensed, (iii) patient health outcomes, (iv) total patient drug costs, (v) total patient treatment costs, (vi) patient-level managed care access or co-pay measures, and (vii) level of pharmaceutical promotions at the physician level. The lack of pharmacist or pharmacy characteristics means no inferences can be made on the effect of varying directions and incentives given by individual pharmacies to instruct their pharmacists to reach quotas on the number of generic prescriptions dispensed, which in turn, could affect the number of generic-to-branded drug switch-backs. Pharmacies with greater quota targets to reach and higher financial incentives to pharmacists to dispense generics would suggest lower switch-backs to branded drugs. The lack of data on the original physician prescription means no analysis of comparing what prescription was prescribed versus dispensed. The lack of data on patient health outcomes means no direct inference can be made how branded-to-generic drug substitution or generic-to-branded drug switch-backs affect health outcomes. This analysis can only suggest that if a generic-to-branded drug switch-back occurs, it is because of a failure of the branded-to-generic drug substitution to achieve desired clinical outcomes and/or generated unwanted side effects and not due to economic considerations or managed care incentives. Total patient drug and treatment costs are also not measured. Capturing these measures requires leveraging claims data from health plans. Existence of this data would raise interesting questions whether the enactment of DPS laws, for example, lowers drug costs since dispensed generic drug prescriptions would increase, but what happens to treatment costs and health outcomes in response to greater generic drug utilization would be open for analysis. This is a matter for more detailed discussion later in this paper. Managed care co-pay data were not available at the patient level, but were partially captured by using patient-level method-of-payment data. Patients using Medicaid or third party (including Medicare Part D) payment plans have lower co-pays than patients paying cash. While Medicaid patients have lower absolute co-pays than patients on third party plans, the cost of these co-pays relative to patient disposable income is unclear since Medicaid patients are by design poorer than third party patients. The influence of pharmaceutical promotion, mainly through visits by sales representatives delivering product-specific discussions ("detailing") by encouraging physicians to prescribe branded drugs was likely captured by physician prescription volume, with more promotions by companies targeting physicians with larger prescription volume. Lastly, the patient-level database does not provide the race of patients. However, the effect of race is likely present in other measures already captured, though more crudely, in data such as Medicaid method-of-payment, because black and Hispanic patients are proportionally more impoverished than white patients according to census data. These are also at-risk population groups that have less access to quality healthcare and thus more likely to have multiple health conditions which is captured in the model through patient co-morbidity count (Betancourt and Green 2010; Betancourt and Blumenthal 2007).

While the preceding data limitations we surmise do not alter our ability to conduct empirical tests on the research objectives, they do limit the precision of the statistical analysis to infer certain specific reasons that may be behind the relationship between DPS laws and generic-to-branded drug switch-backs. The lack of the preceding data elements also means care has to be done to ensure tests are conducted to see if there exists any remaining unexplained systematic variation that may be a result of missing data.

4. Empirical Estimation

Logistic regression estimation was used because the dependent variable "switch-category" was a dichotomous variable as it included patients who engaged in switch-backs (defined as "2") and patients who engaged in brand-to-generic substitutions (defined as "1") (Greene 1993). Censored regressors were used by dropping all observations with any missing values, since recent research has derived theoretical reasons why an alternative dummy variable approach is questionable (Rigobon and Stoker 2009, 2007). Standard errors were estimated to account for repeated patient observations, while other tests were performed to check for model specification issues. Wald test was run to test the null hypothesis that one or all the coefficients = 0. The null hypothesis was rejected at a significance level of p < 0.0001 (Greene 1993). Tests did not find the existence of multicollinearity using a variance inflation factor (vif) method (Robinson and Schumaker 2009). All the variables in the model generated a vif < 2, far below a vif threshold value of 10, thus rejecting existence of multicollinearity. Model estimates and significance levels were also invariant to alternative specifications.

A few comments are in order to explain the reasons behind the statistical method and tests employed on this data: (i) logistic regression estimation is a very common and accepted statistical method for analyzing data that is dichotomous in measurement, (ii) accounting for repeated measures is necessary since a single patient can be represented more than once in having multiple brand-to-generic drug substitutions and/or generic-to-branded drug switch-backs, (iii) specification tests were performed to ensure that the model is providing robust insights that are insensitive to alternative ways to define what other variables are employed in the model, especially in light of some data limitations noted earlier, and (iv) tests to determine the existence of multicollinearity are important to ensure proper attribution of effects are assigned to each variable employed in the model. All the statistical tests performed are commonly employed when conducting this type of modeling analysis.

IV. Results

7. Descriptive Statistics

Table 3 provides descriptive statistics per variable, total number of prescriptions and patients by therapeutic and bioequivalent substitution generic drug substitution and switch-backs. The data sample revealed that 93.5% of patients in the sample engaged in therapeutic substitution and 96.4% of patients engaged in switch-backs relative to the total number of patients engaged in brand-to-generic drug substitution and generic-to-branded drug switchbacks. A similar percentage breakdown occurred with the number of therapeutic brand-to-generic substitution and generic-to-branded drug prescriptions. This means we have many patients to analyze that have engaged in either brand-to-generic substitution or generic-to-branded drug switch-backs rather than a sample that is skewed to just having substitution or switch-back transactions. Also, over 93% of prescriptions or patients were involved in therapeutic as opposed to bioequivalent substitution or switch-backs. Patient demographic information reveals a mean age of 64 years, a mean cardiovascular risk co-morbidity count per patient of 4.9, and that 51% are male. The physician specialty breakdown is PCP (41%), IM (41%), Card_ Neph (13%), Endo_Dia (2%), with 3% being Other physicians. Among the method-of- payment measures, third party payment represents 95% of the transactions with 2% each being Medicaid and cash. Patient consent DPS laws are the most represented in the sample (40 out of 51-50 states plus D.C.), followed by permissive (39 out of 51), formulary substitution (17 out of 51), two-line Rx format (10 out of 51), and no cost saving pass on (10 out of 51). Switch to a lower dose occurred in 45% of the transactions while a switch to a different molecule, either in a brand-to-generic substitution or generic-to-branded drug switch-back, occurred in 96% of the transactions. Finally, the positive mean managed care control measure of 0.21 with a wide dispersion from 0.00 to 0.87 implies there exists a significant variation in drug plans that can move market share beyond the national average, from plans that have no ability to move market share (0.00) to those than can substantially move market share.

2. Overall Model Empirical Results

Table 4 presents results of DPS state laws effects on the probability of patient switch-backs from generic to branded statin drugs, while controlling for other factors affecting generic substitution and switch-backs. The table also lists log odds estimates ranked from highest to lowest in magnitude by absolute value. Model results showed strong statistical significance between generic-to-branded drug switch-backs and DPS laws. All DPS model variables specified were highly significant to at least the 99.99% confidence level with the direction (positive or negative) on the log odds estimates generally as predicted according to prior research studies. Switch-backs to branded drugs occurred despite pressure from factors limiting such activity, such as, effects from DPS laws, local managed care plan control, recessionary effects of 2008 on patient affordability of drugs (Martin et al. 2011), lower co-pays to encourage patients using generic drugs, and the effects from other factors not specified directly in the model but reveal themselves by the highly significant and large negative intercept sign. The economic effects from the last recession, though not specifically modeled, are likely revealed through the strong negative intercept, suggesting factors outside the model design strongly contributed to preventing switch-backs. The deep and lengthy recession affected people's income/wealth and loss of drug insurance, making paying for more expensive branded drugs through switch-backs more difficult, thus reducing the likelihood of switch-backs occurring. The fact that switch-backs occurred despite these strong factors working against it from happening, provides suggestive evidence that important clinical reasons were forcing switch-backs to occur. The model also showed a high association of predicted probabilities to observed responses, a reasonable proxy [R.sup.2] measure, at 81.9% (Greene 1993).

Finally, regarding the interpretation of the log odds estimates, for purposes of the discussions here, the sign and magnitude of the estimates will be the focus. The sign (positive or negative) signifies the relationship direction between changes in an explanatory variable and switch-backs: (i) a positive-signed estimate means an increase in an explanatory variable will increase switch-backs, while (ii) a negative-signed estimate means an increase in an explanatory variable will decrease switch-backs. Logistic regression is a statistical procedure that normalizes the estimates, meaning direct comparisons in the magnitude of the estimate can be made as to which explanatory variable has a greater likelihood in affecting the dependent variable (switchbacks vs. substitution). Table 4 lists the log odds estimates in magnitude order from the largest to smallest number in absolute value so one can clearly see what explanatory variables have a greater influence on the dependent variable but also by what quantitative degree relative to other variables.

3. DPS Laws Empirical Results

Results indicated switch-backs to branded drugs were more prevalent in states not requiring patient consent prior to substitution, having mandatory DPS laws, utilizing two-line prescriptions forms, not requiring the passing of savings to consumers, and having formulary restrictions. States having patient consent requirements decreased the log odds of switch-backs by -0.063, compared to states without. States having permissive requirements decreased the log odds of switch-backs by -0.040 compared to states without. Thus, the influence of patient consent DPS laws was more than 50% greater the effect from permissive requirements in decreasing the likelihood of switch-backs. Increasing influences on switch-backs were states having DPS laws utilizing a two-line prescription form (0.079), not requiring passing of savings to patients (0.083), and the existence of a drug formulary requirement (0.103), compared to states without. These DPS laws have similar effects to each other on switch-backs given log odds estimates close to each other, but are significantly higher than the effects from patient consent and permissive requirement DPS laws. Physicians in two-line prescription form states had greater influence affecting prescribing decisions and reducing the ability for pharmacists to affect drug dispensing decisions. The existence of a formulary limits the number of drugs that pharmacists may substitute. Hence, one would expect switch-backs to be more frequent in states requiring formularies. The log odds of switch-backs for states requiring a formulary increased by 0.103 as compared with states having no such formulary requirement. Overall, switchback probabilities were higher in states with DPS laws designed to lessen patient and pharmacist autonomy. Overall, the results were consistent with prior empirical studies on DPS laws (Shrank 2010; Carroll et al. 1987; Masson and Steiner 1985).

4. Non-DPS Laws Empirical Results

All non-DPS specified variables were statistically significant to at least the 99% confidence level explaining generic-to-branded drug switchbacks. A higher probability of switch-backs occurred with an increase in or existence of these factors: (i) greater patient health risk measured as higher co-morbidity count (0.022), (ii) male patients (0.028), (iii) existence of a drug benefit payment making out-of-pocket costs for branded drugs less expensive through third party (0.068) and Medicaid (0.239), (iv) physician specialists relative to primary care physicians (PCP), such as, Card_ Neph (0.448), Endo_Dia (0.319), Other physicians (0.189), and IM (0.017), (v) physician Rx volume (0.011), (vi) switch to a lower dose (2.695), and (vii) switch to a different molecule (0.932). The most common switch-backs were from higher doses of simvastatin, an older and less effective statin, to lower doses of later and more effective statins in atorvastatin, ezetimibe/simvastatin, and rosuvastatin. This switch-back pattern was consistent with differences in drug efficacy between branded statins where an atorvastatin 20 mg, rosuvastatin 10 mg, or ezetimibe/simvastatin 10 mg have approximately the same LDL-C reduction effect as a simvastatin 80 mg (Medical Economics Company 2009; Ohsfeldt et al. 2008).

A lower probability of switch-backs occurred with an increase in or existence of these factors: (i) managed care plan control (-1.531), (ii) patient age (-0.003), and (iii) the intercept (-2.308). The interpretation of the intercept is the predicted log odds estimates when all explanatory variables are zero. This large negative sign suggests that the remaining influences on switch-backs not specifically specified in the model point to a strong resistance to switch-backs. The intercept log odds estimate is the 2nd largest in absolute value only to switch to a lower dose.

5. Ranking of ALL Log Odds Empirical Estimates

The strongest factors increasing switch-backs according to the log odds estimates were switch to lower dose (2.695) and different molecule (0.932), Card_Neph (0.448), Endo_Dia (0.319), and Medicaid payment (0.239). Local managed care plan control was the largest and most significant factor preventing switch-backs (-1.531). The managed care control effect was larger in absolute value than all model effects except for switch to lower dose. While all DPS law variables are statistically significant to at least the p < 0.01 level, the magnitude of the log odds estimates are neither the largest nor smallest values among all model variables. The DPS variables have higher log odds estimates in absolute value than patient gender and age, patient co-morbidity count, IM physicians, and physician Rx volume. However, DPS variables have significantly lower log odds estimates in absolute value than switch to lower dose, managed care plan control, switch to different molecule, Card_Neph, Endo_Dia, Medicaid payment, and Other physicians. For example, the effect of "switch to a lower dose" is over 26 times greater than the DPS variable "drug formulary" (2.695 vs. 0.103), while "switch to a different molecule" is about 9 times greater than "drug formulary (0.932 vs. 0.103). Thus, the empirical results show that the existence of some DPS variables are strongly statistically associated with increasing generic-to-branded drug switch-backs while others are strongly statistically associated with increasing brand-to-generic drug substitution. However, as a group, given the lower magnitude of the log odds estimates relative of other model variables, state pharmacy DPS law variables are not the principal drivers of such switch-back or brand-to-generic substitution patterns.

V. Discussion

7. Discussion and Implications of Present Study Results

The issue of generic therapeutic interchange and substitution is under debate by medical and health policy researchers as to whether it has consequences to patient health on the grounds of differences in clinical equivalence. DPS laws allow pharmacies to more easily switch prescriptions for branded-to-generic drugs. The present research measured the effects of state pharmacy DPS laws on statin patient brand-to-generic substitution and generic-to-branded drug switch-backs, using switch-backs as a potential indicator of clinical failure to achieve desired health outcomes from the brand-to-generic substitution. LifeLink[R] Anonymized Patient Level Data from IMS Health from March 2006 through June 2008 were analyzed on the drug utilization patterns for 397,111 U.S. statin patients from office-based physicians. Logistic regression estimated the effects of DPS laws and non-DPS variables on switch-backs. Existence of DPS mandatory laws, drug formulary laws, use of a two-line prescription form, and absence of patient consent laws increased switch-backs. The log odds estimates for the DPS variables ranged from 0.040 to 0.103 in absolute value.

The strongest factors associated with increasing switch-backs (using log odds) were switch to lower dose (2.695), switch to different molecule (0.932), specialist physicians involved in prescribing statin drugs [cardiologists and nephrologists (0.448), endocrinologists and diabetologists (0.319)], and Medicaid method-of-payment (0.239). Local managed care plan control was the largest factor preventing switch-backs (-1.531), with weaker effects from the existence of patient consent (-0.063) and permissive (-0.040) DPS laws. So while DPS variables contribute to brand-to-generic substitution or generic-to-branded drug switch-backs, the present study empirical results show they are not the principal drivers of such drug utilization patterns relative to other stronger non-DPS law factors.

Most of the present study empirical results on the effects of DPS laws were consistent with prior empirical studies, with some exceptions. A 1980 study of DPS laws using IMS Health data found the opposite effect from permissive versus mandatory laws on state generic substitution rates (Masson and Steiner 1985). One possible explanation for the difference is greater availability of generics in 2006-2008 versus 1980. Another possible reason is that state pharmacy DPS laws in 1980 were still in their infancy. In another study on Medicaid-only patients, researchers found the existence of patient consent laws lowered generic substitution rates (Shrank et al. 2010). The difference in results may be attributed to the following reasons: (i) this current study analyzes patient-level drug-utilization data patterns between generic and branded drugs as opposed to data analysis at an aggregated state-level, thus revealing a different relationship that is only seen when conducting a more granular analysis of the data, (ii) the prior study used Medicaid-only data whereas our study looks more broadly across patients with various method-of-payment instruments, (iii) greater availability of generic drug options for patients, and (iv) greater confidence by patients in the quality and safety of generic drugs.

These last-stated potential reasons for differences between present and previous study results deserve further elaboration and discussion. First, patient-level data on drug utilization patterns has been qualitatively improving and is more readily available to researchers. The existence of granular patient-level data means that greater insights into reasons behind brand-to-generic substitution and generic-to-branded drug switch-backs are more possible by today's researchers relative to past researchers. The availability of health plan claims data to researchers allows for the analysis of changes in drug utilization patterns through plan pharmacy data along with data on changes in patient health outcome, drug cost, and treatment cost. Health plan claims data can also reveal how changes in the utilization between branded and generic drugs result in differences of the following effects: (i) unwanted drug side-effects (adverse events) that can be harmful to patients, (ii) patient drug adherence (patients maintaining drug utilization as intended and prescribed by the physician), and (iii) appropriate use of medicines (patients using their medications according to the drug's indicated label as approved by the FDA). Thus, using this very detailed data would allow for developing a more complete picture how state pharmacy DPS laws affect patients and the healthcare system than previously possible.

Second, the effects of patient consent DPS laws warrant further scrutiny. Patient consent DPS laws were designed to provide a check on pharmacist actions when dispensing a generic drug in place of a branded drug. Patient consent laws also indirectly bring physician information and intent on how to treat patients into the decision on brand versus generic drug choice. Where no DPS patient consent laws exist, pharmacists are free to dispense a bioequivalent generic drug in place of the prescribed branded drug without informing or receiving consent from patients. However, with patient consent laws, pharmacists are now confronted with patient decisions based on their interactions with physicians on how to move forward in the drug treatment of their condition. The present study empirical results reveal a negative sign on the patient consent variable, meaning that the existence of patient consent DPS laws decrease generic-to-branded drug switch-backs and increase brand-to-generic drug substitution. When faced with no patient consent laws, the empirical results suggest greater generic-to-branded drug switch-backs. As noted earlier, pharmacies and pharmacists have financial incentives to dispense generic drugs over branded drugs. Thus, greater switch-backs occur when pharmacists are less constrained to make decisions in favor of their own financial interests relative to the interests stated by patients and indirectly from physicians through patients.

What is not known here, because of data limitations by not explicitly looking at health outcome data, is whether patients made decisions that are in the best interest of their own health and interests conveyed by their attending physicians, or were such consent decisions made more on economic grounds in order to obtain drugs at a lower cost? The question left unanswered by this current analysis is the health outcome comparison of two different approaches in treating high cholesterol through statin drugs: (i) patients receive lower co-pay cost generic drugs obtained through DPS patient consent laws which would improve drug adherence, thereby allowing patients to more likely achieve the benefits of drug therapy, albeit from older less effective statin drug technology, or (ii) patients receive higher co-pay cost patented branded drugs which would lessen drug adherence, though receive newer more effective statin drug technology.

Regarding the effects from other DPS laws, the empirical results reveal greater generic-to-branded drug switch-backs when drug formulary, no cost saving pass on, and two-line Rx form DPS laws exist. The existence of DPS drug formularies, especially if they are designed similar to managed care plan drug formularies, give greater emphasis to generic drug utilization. This results in pharmacists in pursuing their own financial interests to dispense generic drugs, with again, the empirical results finding greater generic-to-branded drug switch-backs. The existence of no cost saving pass on DPS laws allow pharmacists to reap greater returns from brand-to-generic substitution, thereby greater generic-to-branded drug switch-backs. Finally, the existence of two-line Rx forms have DPS permitted, thereby allowing pharmacists to pursue their own financial interests by engaging in brand-to-generic substitution, resulting in increased generic-to-branded drug switch-backs. The end result in each of the preceding DPS laws is to give pharmacists greater latitude to engage in brand-to-generic substitutions, with the effect being greater generic-to-branded drug switch-backs.

Finally, the effects of permissive DPS laws are discussed. Here, the present study empirical result generates potentially conflicting insights from what would be expected given previous study research findings. Permissive DPS laws generally state that pharmacists may substitute branded drug prescriptions with generic drugs. Thus, pharmacists have greater latitude to pursue their own financial interests to dispense generic drugs. However, rather than the current study empirical model generating greater generic-to-branded drug switch-backs, the opposite is found. The answer may lie in the type of permissive DPS laws that are in effect. Permissive DPS laws that are mandatory state that the pharmacist must substitute a lower priced product when the Rx was written for a branded product and the prescriber has not prohibited substitution. Thus, if most permissive state DPS laws are in the form of mandatory laws, then pharmacists would be acting more in line with physician wishes, thereby reducing expected generic-to-branded drug switch-backs.

The empirical results presented here therefore raise broader questions as to whether DPS laws operate in practice as originally intended. The intent of DPS laws is to encourage the dispensing of generic drugs that are therapeutically equivalent to more expensive patented branded drugs. The drugs being switched by pharmacists under strict applications of DPS laws are bioequivalent, with the only real difference being cost. The effects of DPS laws, both intended and unintended, will have growing importance as an increasing number of drags lose patent protection and thus become available as generic drags. This means there will be a growing opportunity for patients on patented branded drugs to be able to replace their more expensive branded prescription with a less expensive bioequivalent prescription. The dispensing of a bioequivalent brand-to-generic interchange/ substitution is less controversial since the switch is between two drugs that are essentially the same though are available at different costs. However, the data analyzed here showed much a greater proportion of therapeutic brand-to-generic interchange. The questions raised then are the effects of DPS laws in practice on brand-to-generic therapeutic interchange since this switch is between two different drugs with potentially different indication profiles and different effects on patients. With more branded drugs losing patent protection and greater numbers of generic drugs available, there will be growing opportunities for brand-to-generic drug therapeutic interchange. Pharmacists will encourage patients who are seeking to reduce patented branded drug co-pays, to receive generic drugs in the same therapy class at a lower co-pay in place of drugs currently on patent, subject to physician approval. The problem created when pharmacists engage in encouraging therapeutic interchange is they are stepping out of their role as persons who understand the pharmacology of drugs into the realm of diagnosing what is best for patients which is what physicians discuss with their patients. This goes beyond the legal intent of state pharmacy DPS laws as merely replacing a more expensive patented branded drug with less expensive bioequivalent generic drug.

What is the suspected mechanism how pharmacists engage in brand-to-generic therapeutic drug interchange since DPS laws only allow bioequivalent interchange or substitution? We say "suspected" since our database does not contain records by individual pharmacists and/or pharmacies. We suspect pharmacists encourage patients to seek out therapeutic generic substitutes, given their financial incentives to dispense generics, when a patient attempts to fill a physician prescription for a patented branded drug with a higher drug co-pay than what is charged for generic drugs. The patient, upon hearing the cost difference between the co-pay for the prescribed patented branded drug versus a generic drug, is more willing to have the generic drug, though not fully understanding the complete ramifications of engaging in brand-to-generic therapeutic substitution. The pharmacist must call into the physician and request the change in prescription for dispensing, which likely finds the physician very busy with other patients and approves the request. The question here is whether this outlined process is what health policymakers had in mind when creating DPS laws, to reduce drug costs by allowing pharmacists to switch between drug that are therapeutically equivalent, and not have pharmacists interfere in physician decisions that are made in the best interest of the patient.

Therefore, the previous discussion suggests important questions to be asked regarding decisions made by actors in the healthcare system that interact with each other and operate under different motivations and incentives which DPS laws act to contribute, that ultimately affect patient care and costs to the healthcare system: (i) what is the role of physicians to prescribe what drug he/she thinks is in the best interest of the patient based on their diagnosis, (ii) what is the role of the patient in making decisions affecting their own health as a non-physician, yet have to bear a co-pay in obtaining any prescribed drug, (iii) what is the role of the pharmacist in making brand-to-generic substitutions who are outside the diagnosis and treatment discussions between physicians and patients, (iv) what is the role of pharmacies that offer incentives to pharmacists to dispense generic drugs as a way to accrue greater profit, and (v) what is the role of managed care plans in determining what drug the physician can prescribe and the patient can receive through control mechanisms to restrict drug choice.

2. Discussion and Implications of Study Limitations and Alternative Model Designs

Not studied here were patient perceptions of risk and resistance to generic substitution due to fear and lack of knowledge affecting patient drug choice. One study estimated up to 30% of patients were skeptical about the efficacy and safety of generic equivalent drugs in the absence of strong demonstrable evidence (Ganther and Kreling 2000). Research results in this current study indicate patients in states where patient consent DPS laws exist have lower switch-backs and higher brand-to-generic substitution. Thus the current study research produces results that deviate from previous research expectations about patient perceptions on the efficacy and safety of generic drugs. The improvement in the quality, efficacy, and safety of generic drugs would diminish patient concerns.

Also not studied because of data limitations are the motivations of patients through patient consent

DPS laws to allow greater brand-to-generic substitution for cost-only reasons and against clinical evidence that would suggest branded drug therapy would be the optimal choice. Generic-to-branded drug switch-backs represent a potential indicator that the original brand-to-generic substitution, mainly through therapeutic generic substitution as revealed in the current study database, may not have been appropriate. As revealed through lower dose and different molecule effects, therapeutic switch-backs were the strongest reasons for moving patients back to branded drugs. The results here suggest more study is needed on effects of patient consent DPS laws on drug choice relative to what is considered clinically optimal. Moving the analysis from an analysis of switchback prescription patterns to patient health outcomes, drug cost, and total cost of treatment would provide more conclusive evidence on the clinical effects of DPS laws.

Finally, alternative analyses, each with different interpretations, were recognized and left out for later research. For example, in the original database of 17.3 million statin patients that preceded this current study analysis, 538,000 patients engaged in generic-to-branded drug switch-backs, and of these patients, 469,000 were therapeutic switchbacks and 69,000 were bioequivalent switch-backs (Chressanthis et al. 2011). The total switch-back number of patients translates into 3.1% of the total 17.3 million patient population, or 7.1% of the 7.6 million patients who started on generic drugs or started on a branded drug and switched to a generic. These relatively small percentages show that generic drugs are working for well over 90% of the patients, or patients are being prevented to engage in switch-backs either because of institutional controls, such as those imposed by managed care, or by the higher co-pay cost of branded drugs. Also, these small percentages do translate into large numbers of patients when analyzing switch-backs in the context of significant chronic diseases that affect millions of Americans. However, why did 56% (9.7 million) of all patients choose to stay on branded drugs and 30% (5.2 million) of all patients choose to stay on generic drugs? Breaking down the original database of statin patients even further, of the 9.7 million patients on branded drugs, 9.2 million patients were on only one branded drug, for a 95% retention rate on one drug. If pharmaceutical promotion is designed to encourage physicians to switch patients onto a different drug, the data in the original database of statin patients presents no evidence that significant brand switching occurred. This relationship also likely shows the clinical treatment bias of physicians keeping patients on a therapy that is working rather than expose patients to unknown effects from switching to a new drug treatment. Also not studied is whether statin drug promotion increased the overall size of patients by encouraging people to seek treatment for high cholesterol. Looking at generic drug utilization, 31% (5.4 million patients) of all 17.3 million patients in the original database started on generic drugs, with 30% (5.2 million) staying on generics, for a 96% retention rate on generic therapy. This shows that once patients start on a generic drug, either the drug works in the vast majority of patients and/or there are controls in place that make it difficult for patients (and prescribing physicians) to switch to a branded drug. Another question is whether switch-back rates were different for patients who started on generic drugs versus patients who started on a branded drug and then switched to a generic? We chose to simplify the analysis by focusing this study only on patients who were involved in brand-to-generic drug substitutions or generic-to-branded drug switch-backs, regardless of the pathway how patients came to utilize generic drugs, leaving these other questions and relationships for other research studies.

VI. Conclusions

Few studies have studied the effects of DPS laws on prescriptions dispensed to patients, none have tracked patient-utilization of drugs, and none have measured their effects on brand-to-generic drug substitution and generic-to-branded drug switch-backs. DPS laws primarily exist to reduce drug costs, encouraging substitution to generics without affecting healthcare quality. The empirical results showed statin patient generic-to-branded drug switch-backs were more prevalent in states with drug formulary, no cost saving pass on, and two-line Rx form DPS laws. The empirical results showed statin patient brand-to-generic drug substitutions were greater in states with patient consent and permissive DPS laws. Local managed care plan control played the largest role in generating generic substitution, thus preventing switch-backs. DPS laws, as practiced by pharmacists, played a smaller, yet significant role in affect drug substitutions and switch-backs. Switching to a lower dose and different molecule, and prescriptions by physician specialists increased switch-backs. Some classes of patients who were engaged in generic substitution, required switch-backs to branded drugs. The statistical effects revealed non-DPS variables affirming findings from a previous research study looking at a descriptive analysis of statin patient switch-back patterns (Chressanthis et al. 2011).

Principal-agent theory would suggest that physicians and their patients are more aligned in their objective functions given their private discussions about the diagnosis and best drug treatment options for their condition that patients can afford. Pharmacists operate under a different economic incentive arrangement and are removed from physician-patient private healthcare discussions. Therefore pharmacists would be expected to be less aligned with healthcare objectives determined by physicians in consultation with patients. The results here provide empirical support for implications of pharmacists operating as agents in healthcare, though the findings were weaker than other effects. However, states with DPS patient consent laws did have lower switch-backs than states without such a provision as predicted by principal-agent theory, ceteris paribus, suggesting that physician influence through the patient contributed to ensuring that the best drug option would be administered which would decrease the probability of switch-backs. Lastly, the results strongly support implications from principal-agent theory from the perspective of health insurance companies desiring to constrain drug costs through managed care control mechanisms to force generic drug utilization at the potential expense of patient care. The managed care controls demonstrate constraints on physician practice to prescribe more effective newer patented branded drug technology at a higher cost in favor of less effective older generic drug technology at a lower cost. Switch-backs primarily go from older statin generic-drug technology at a higher dose (which carries some potential side effect risks, such as simvastatin 80 mg) to newer patented- branded drug technology at a lower dose. Further, specialists most knowledgeable about the cardiovascular condition and the optimal drug option needed for patients are the primary drivers of switch-backs.

States can affect the extent of brand-to-generic substitution prescription drugs by amending their pharmacy DPS laws. The present research suggests therapeutic brand-to-generic substitution is not recommended, without understanding limitations across all classes of patients, such as those with higher cardiovascular risk and greater sensitivity to higher doses, and particularly when done for cost-only considerations, is not recommended. This caution is consistent with a recent major health policy statement on the dangers of therapeutic interchange and substitution, and institutional frameworks through state pharmacy laws and managed care formularies, when done for non-clinical reasons (Holmes et al. 2011). Switches to a lower dose and switches to a different molecule were the two biggest drivers related to increasing generic-to-branded drug switchbacks, affirming the concern of potential inappropriate initial brand-to-generic therapeutic substitution. Newer patented branded statin drugs are more effective in achieving reductions cholesterol levels at a lower dose than older generic drugs that require higher dosing to achieve the same efficacy in cholesterol reduction. Lower drug dosing also means patients are less susceptible to drug side-effects that increase in likelihood at higher drug doses. Physician specialists who treat higher-risk cholesterol patients such as cardiologists and nephrologists, and endocrinologists and diabetologists, were the principal drivers of generic-to-branded drug switch-backs relative to primary care physicians. There are a few reasons for these specialists being the principal drivers of switch-backs relative to primary care physicians: (i) primary care physicians, the initial gatekeepers in the healthcare system to be seen by patients, are more familiar with and accustomed to prescribing older generic drugs, (ii) more severe high-cholesterol patients who are uncontrolled under older generic drug technology, are likely to be transferred to specialists who are more familiar with the latest and more powerful statin drugs, and (iii) specialists represent physicians who generate high-prescriptions per prescriber and thus are more lucrative targets by pharmaceutical companies for promotion by industry sales representatives.

Unknown was the role that pharmacists played in potentially influencing patients to seek lower-cost generic drug treatments through therapeutic substitution since DPS laws only allow bioequivalent generic substitutions, when permitted by physicians. Further research is needed to answer whether DPS laws encouraged decisions by pharmacists to choose generics or influence generic substitution over clinically superior brand treatments in some patients for monetary gain at the expense of patient health. Another question for further and later discussion is what position should be taken by physicians to resist allowing generic substitution by pharmacists on grounds other than for clinical reasons? In addition, how much should physicians be allowed to challenge the controls imposed by managed care drug plans to force generic drug prescribing first when branded drug therapy is recommended as the best option for patients? How much independence should be given to physicians to make decisions that they feel are in the best health interests of patients? Otherwise, DPS laws can produce unintended effects of driving some classes of at-risk patients into less effective older generic drug options that may potentially come at the expense of patient health (Liew et al. 2012). The long-term consequences can be greater overall treatment costs as a result of particular patients being on less desirable therapy (Liew et al. 2012). States should consider revisions in DPS laws to make it harder for pharmacists and patients to counter physician decisions who selected the best therapies for specific patients, primarily based on clinical factors. Further, revisions in DPS laws should include the prohibition of pharmacists to discuss with patients brand-to-generic therapeutic substitution since this type of substitution involves the prescription of a different drug than originally intended by the physician. The current process to change a prescription through a therapeutic substitution involves the pharmacist calling the physician for approval. We believe a better process which places the health interest of patients first is for the physician and patient to sit down and discuss the complete ramifications of the drug choice, understanding the full range of benefits and risks to that particular patient. The pharmacist is ill-suited to recommend a different drug choice since they do not have the full medical history and other relevant clinical matters of the patient to make the best drug decision. Future research will determine whether these recommendations not only produce better health outcomes but also be more long-term cost-effective to the healthcare system. Lastly, the current analysis is on statin patient switch-back patterns. More needs to be understood if the same effects occur across other disease conditions. The empirical findings in this study suggest the potential for even greater switch-back effects in other chronic disease conditions for the following reasons: (i) where significant therapeutic substitution by the forces modeled in this research also occur, (ii) greater effectiveness differences by dose between generic and patented branded drug technology, and (iii) greater patient-level idiosyncratic response to drug therapy. Examples of drug therapy classes that exemplify the preceding characteristics include but not limited to oncologies, antidiabetics, antipsychotics, autoimmune diseases, and central nervous diseases such as Alzheimer's disease, Parkinson's disease, and Multiple sclerosis. Spending by therapeutic drug class over $10 billion as of the end of 2011 included the following areas out of total United States drug spending of $320 billion: oncologies ($23.2B), respiratory agents ($21.0B), lipid regulators ($20.IB), antidiabetics ($19.6B), antipsychotics ($18.2B), autoimmune diseases ($12.0B), antidepressants ($11.0B), HIV antivirals ($ 10.3B), and anti-ulcerants ($10.IB) (IMS Institute for Healthcare Informatics 2012b). Healthcare spending on patients with chronic diseases account for a larger proportion of total healthcare spending, thus better understanding how DPS laws affect drug choice and healthcare system costs are of significant health policy importance (IMS Institute for Healthcare Informatics 2012a).

Disclosures

Kevin Fandl has no financial conflicts to report. Nayla Dahan worked on this paper while on a postdoctoral research fellowship in pharmaceutical economics and health policy research from September 2010 to September 2011 in the Department of Risk, Insurance and Healthcare Management in the Fox School of Business and Management at Temple University funded by AstraZeneca Pharmaceuticals LP (AZ). She has no direct relationships with industry except for receiving post-doctoral support through a grant funded by AZ. George Chressanthis served as the principal investigator. George Chressanthis was an employee of AZ, having left there at the end of December 2009. He holds no direct financial interests in AZ, having sold off all remaining stock grants in the company in July 2012. He completed another prior research project that was funded by AZ and also received a private grant for academic research from TGaS Advisors, a pharmaceutical benchmarking firm that works with drug companies on their commercial operations processes. He received travel funds from the AZ and TGaS Advisors grants to disseminate research findings at pharmaceutical industry and academic conferences. Temple University retained intellectual property rights and independent control over the design and implementation of this study, analysis, writing of the manuscript, and information dissemination. The exception to this control was the exclusion from publication of any specific information that was deemed commercially sensitive and proprietary that AZ wished to remain confidential. Such an exception is common with academic-corporate research relationships. This exception did not alter in any way the research approach pursued by the researchers. Oversight by Temple University officials of any work completed by George Chressanthis in connection with AZ was done to ensure all research meets rigorous academic standards as an added precaution against claims of conflict of interest. In addition, George Chressanthis is current on all reporting of potential financial conflicts of interest to Temple University compliance authorities. Further, the AZ product Crestor[R] is among the statin drugs studied in this research. The statements, findings, conclusions, views, and opinions contained and expressed in this article are based in part on data obtained under license from IMS Health Incorporated Information service^): LifeLink* Anonymized Patient Level Data (2006-2008), Xponent[R] (2008), and Xponent PlanTrak[R] (2008), IMS Health Incorporated. All Rights Reserved. The statements, findings, conclusions, views, and opinions contained and expressed herein are not necessarily those of IMS Health Incorporated or any of its affiliated or subsidiary entities.

References

American Medical Association. 2006-2008. "Physician Specialty Designations." Chicago, IL: 2006-2008.

Andermann, F., Duh, M., Gosselin, A. and Paradis, P. 2007. "Compulsory Generic Switching of Antiepileptic Drugs: High Switchback Rates to Brand Compounds Compared to Other Drug Classes." Epilepsia, 48: 464-469.

Barton, D., Stossel, T. and Stell, L. 2014. "After 20 Years, Industry Critics Bury Skeptics, Despite Empirical Vacuum." International Journal of Clinical Practice, 68: 666-673.

Berenson, A. 2006. "Merck Loses Protection for Patent on Zocor." New York Times, June 23. Available from URL: http://www.nytimes.com/ 2006/06/23/business/23statin.html. Accessed October 6, 2014.

Berg, M., Gross. R., Tomaszewski, K., Zingaro, W. and Haskins, L. 2008. "Generic Substitution in the Treatment of Epilepsy: Case Evidence of Breakthrough Seizures." Neurology, 12: 525-530.

Betancourt, J. and Blumenthal, D. 2007. "Racial and Ethnic Disparities in Health Care." In Harrison's Principles of Internal Medicine, Anthony Fauci, et al. (editors), Chapter 11.

Betancourt, J. and Green, A. 2010. "The Cost of Inaction: Developing a Business Case for Addressing Racial and Ethnic Disparities in Health and Health Care." Howard Health Policy Review, 11: 13-16.

Blomqvist, A. 1991. "The Doctor as Double Agent: Information Asymmetry, Health insurance, and Medical Care." Journal of Health Economics, 10: 411-432.

Carroll, N., Fincham, J. and Cox, F. 1987. "The Effect of Differences in State Drug Selection Laws on Pharmacists' Substitution Behavior." Medical Care, 25: 1069-1077.

Cheng, J. 2008. "An Antitrust Analysis of Product Hopping in the Pharmaceutical Industry." Columbia Law Review, 108: 1471-1515.

Chressanthis, G., Khedkar, P., Jain, N., Poddar, P. and Seiders, M. 2013. "Suppressing Commercial Speech is Not in the Best Interest of Physicians and Patients." Journal of Clinical Hypertension, 15: 219-220.

Chressanthis, G., Khedkar, P., Jain, N., Poddar, P. and Seiders, M. 2012. "Can Access Limits on Sales Representatives to Physicians Affect Clinical Prescription Decisions? A Study of Recent Events with Diabetes and Lipid Drugs." Journal of Clinical Hypertension, 14: 435-446.

Chressanthis, G., Khedkar, P., Jain, N., Poddar, P. and Seiders, M. 2011. "Generic-to-Branded Drug Switch-Back Patterns: Exploratory Analysis and Insights from the Statin Class." Journal of Pharmaceutical Health Services Research, 2: 217-226.

Clouse, E., McCormick, W., Angom, R., Kimberlin, C. and Bradham, D. 1985. "Drug Product Selection: The Florida Experience Revisited." American Journal of Public Health, 75: 283-284.

Danzon, P. and Furukawa, M. 2011. "Cross-National Evidence of Generic Pharmaceuticals: Pharmacy vs. Physician-Driven Markets. NBER working paper, No. 17226, July.

DeRuiter, J. and Holston, P. 2012. "Drug Patent Expirations and the "Patent Cliff."" U.S. Pharmacist, 37: 12-20.

Eggleston, K. 2005. "Multitasking and mixed systems for provider payment." Journal of Health Economics, 24: 211-223.

Federal Trade Commission. 1979. "Drug Product Selection." Staff report to the U.S. FTC, Bureau of Consumer Protection, Washington, D.C.

Fuchs V. 2000. "The Future of Health Economics." Journal of Health Economics, 19: 141-157.

Gagne, J., Avorn, J., Shrank, W. and Schneeweiss, S. 2010. "Refilling and Switching of Antiepileptic Drugs and Seizure Related Events." Clinical Pharmacology & Therapeutics, 88: 347-353.

Ganther, J. and Kreling, D. 2000. "Consumer Perceptions of Risk and Required Cost Savings for Generic Prescription Drugs." Journal of the American Pharmaceutical Association, 40: 378-383.

Goldberg, T., DeVito, A., Smith, D., Stano, M., Vidis, J., Moore, W. and Dickson, M. 1979. "Evaluation of Economic Effects of Drug Product Selection Legislation." Medical Care, 17: 411-419.

Goldman, D., Joyce, G. and Zheng, Y. 2007. "Prescription Drug Cost Sharing: Associations with Medication and Medical Utilization and Spending and Health." JAMA, 298: 61-69.

Greene, W. 1993. Econometric Methods. Englewood Cliffs, NJ: Prentice Hall.

Hatch-Waxman Act, Pub. L. No. 98-417 (1984).

Hellerstein, J. 1998. "The Importance of the Physician in the Generic versus Trade-Name Prescription Decision." Rand Journal of Economics, 29: 108-136.

Holmes, D. Jr., Becker, A., Granger, C., Limacher, M., Page, R. II. and Sila, C. 2011. "ACCF/AHA 2011 Health Policy Statement on Therapeutic Interchange and Substitution." Journal of the American College of Cardiology, 58: 1287-1307.

Hsu, J., Price, M., Huang, J., Brand, R., Fung, V., Hui, R., Fireman, B., Newhouse, J. and Selby, J. 2006. "Unintended Consequences of caps on Medicare Drug Benefits." New England Journal of Medicine, 354: 2349-2359.

Huddle, T. 2008. "Drug Reps and the Academic Medical Center: A Case for Management Rather than Prohibition." Perspectives in Biology and Medicine, 51: 251-260.

IMS Health. 2006-2008. "LifeLink[TM] Anonymized Patient Level Data (APLD)." Norwalk, CT.

IMS Health. 2006-2008. "Xponent[TM]." Norwalk, CT.

IMS Health. 2008. "Xponent PlanTrak[TM]." Norwalk, CT.

IMS Institute for Healthcare Informatics. 2014. Available from URL: http://www.imshealth.com/ portal/site/ims/menuitem.5ad1c081663fdf9b41d84 b903208c22a/?vgnextoid=e39f79d7f269e210Vgn VCM10000071812ca2RCRD&vgnextfmt=default. Accessed October 5, 2014.

IMS Institute for Healthcare Informatics. 2012. "Healthcare Spending Among Privately Insured Individuals Under Age 65." February. Available from URL: http://www.imshealth.com/deployed files/ims/Global/Content/Insights/IMS%20Institute %20for%20Healthcare%20Informatics/Healthcare %20Spending/IHII_Spending_Report.pdf. Accessed October 6, 2014.

IMS Institute for Healthcare Informatics. 2012. "The Use of Medicines in the United States: Review of 2011." April. Available from URL: http://www.imshealth.com/ims/Global/Content/ Insights/IMS%20Institute%20for%20Healthcare% 20Informatics/IHII_Medicines_in_U.S_Report_ 201 l.pdf. Accessed October 6, 2014.

Johnston, A. 2010. "Challenges of Therapeutic Substitution of Drugs for Economic Reasons: Focus on CVD Prevention." Current Medical Research and Opinion, 26: 871-878.

Kesselheim, A., Misono, A., Lee, J., Stedman, M., Brookhart, A., Choudhry, N. and Shrank, W. 2008. "Clinical Equivalence of Generic and Brand-Name Drugs Used in Cardiovascular Disease: A Systematic Review and Meta-Analysis." JAMA, 300: 2514-2526.

Kesselheim, A., Stedman, M., Bubrick, E., Gange, J., Misono, A., Lee, J., Brookhart, A., Avorn, J. and Shrank, W. 2010. "Seizure Outcomes Following the Use of Generic versus Brand-Name Antiepileptic Drugs: A Systematic Review and Meta-Analysis." Drugs, 70: 605-621.

Korn, D. and Carlat, D. 2013. "Conflicts of Interest in Medical Education: Recommendations From the Pew Task Force on Medical Conflicts of Interest." JAMA, 310: 2397-2398.

Kushner, D. and Feierman. R. 1978. "RPh's Pass Up Most Opportunities to Substitute, A.D. Survey Discloses. American Druggist, 178: 7-17.

Larkin, I., Ang, D., Avorn, J. and Kesselheim, A. 2014. "Restrictions on Pharmaceutical Detailing Reduced Off-Label Prescribing of Antidepressants and Antipsychotics in Children." Health Affairs (Millwood), 33: 1014-1023.

LeLorier, J., Duh, M., Paradis, P., Lefebvre, P., Weiner, J., Manjunath, R. and Sheehy, O. 2008. "Clinical Consequences of Generic Substitution of lamotrigine for Patients with Epilepsy." Neurology, 70: 2179-2186.

Lesko, R., Scott, S. and Stossel, T. 2012. "Bias in High-Tier Medical Journals Concerning Physician-Academic Relationships with Industry." Nature Biotechnology, 30: 320-322.

Liew, D., Webb, K., Marbaix, S. and Annemans, L. 2012. "Changes to the Statin Prescribing Policy in Belgium: Potential Impact in Clinical and Economic Terms." American Journal of Cardiovascular Drugs, 12: 225-232.

Majumdar, S., McAlister, F. and Soumerai, S. 2003. "Synergy between Publication and Promotion: Comparing Adoption of New Evidence in Canada and the United States." American Journal of Medicine, 115: 467-472.

Mark, T., Gibson, T. and McGuigan, K. 2009. "The Effects of Antihypertensive Step-Therapy Protocols on Pharmaceutical and Medical Utilization and Expenditures." American Journal of Managed Care, 15: 123-131.

Martin, A., Lassman, D., Whittle, L., Catlin, A. and the National Health Expenditure Accounts Team. 2011. "Recession Contributes to Slowest Annual Rate of Increase in Health Spending in Five Decades." Health Affairs (Millwood), 30: 11-22.

Masson, A. and Steiner, R. 1985. "Generic Substitution and Prescription Drug Prices: Economic Effects of State Drug Product Selection Laws." Washington, DC: Federal Trade Commission.

Medical Economics Company. 2009. Physicians Desk Reference 64th edition. Montvale, NJ: PDR Network, LLC.

Mintzes, B., Lexchin, J., Sutherland, J., Beaulieu, M., Wilkes, M., Durrieu, G. and Reynolds, E. 2013. "Pharmaceutical Sales Representatives and Patient Safety: A Comparative Prospective

Study of Information Quality in Canada, France and the United States." Journal of General Internal Medicine, 28: 1368-1375.

National Association of Boards of Pharmacy. 2008. "Annual Survey of Pharmacy Law.

Ohsfeldt, R., Gandhi, S., Fox, K. and McKenney, J. 2008. "Statin Cost-Effectiveness Comparisions Using Real-World Effectiveness Data: Formulary Implications." Value in Health, 11: 1061-1069.

Penn. Pub. L., [section][section] 1, 3, 4 (1976).

Richton-Hewett, S., Foster, E. and Apstein, C. 1988. "Medical and Economic Consequences of a Blinded Oral Anticoagulant Brand Change at a Municipal Hospital." Archives of Internal Medicine, 148: 806-808.

Rigobon, R. and Stoker, T. 2009. "Bias from Censored Regressors." Journal of Business & Economic Statistics, 27: 340-353.

Rigobon, R. and Stoker, T. 2007. "Estimation with Censored Regressors: Basic Issue." International Economic Review, 48: 1441-1467.

Robinson, S. and Schumacker, R. 2009. "Interaction Effects: Centering, Variance Inflation Factor, and Interpretation Issues." Multiple Linear Regression Viewpoints, 35: 6-11.

Roebuck, M., Liberman, J., Gemmill-Toyama, M. and Brennan, T. 2011. "Medication Adherence Leads to Lower Health Care Use and Costs Despite Increased Drug Spending." Health Affairs (Millwood), 30: 91-99.

Ross, S. 1973. "The Economic Theory of Agency: The Principal's Problem." American Economic Review, 63: 134-139.

Shrank, W., Choudhry, N., Agnew-Blais, J., Federman, A., Liberman, J., Liu, J., Kesselheim, A., Brookhart, M. and Fischer, M. 2010. "State Generic Substitution Laws Can Lower Drug Outlays Under Medicaid." Health Affairs (Millwood), 29: 1383-1390.

Simoens, S. 2011. "Generic and Therapeutic Substitution: Ethics Meets Health Economics." International Journal of Clinical Pharmacy, 33: 469-470.

Sokol, M., McGuigan, K., Verbrugge, R. and Epstein, R. 2005. "Impact of Medication Adherence on Hospitalization Risk and Healthcare Cost." Medical Care, 43: 521-530.

Spurling, G., Mansfield, P., Montgomery, B., Lexchin, J., Doust, J., Othman, N. and Vitry, A. 2010. "Information from Pharmaceutical Companies and the Quality, Quantity, and Cost of Physicians' Prescribing: A Systematic Review." PLOS Medicine, 7: 1-22.

Stell, L. 2009. "Drug Reps Off Campus! Promoting Professional Purity by Suppressing Commercial Speech." Journal of Law, Medicine & Ethics, 37: 431-443.

Stell, L. 2005. "Against the Flow: Why Physicians Should Listen to Drug Reps." American Society for Bioethics and Humanities Exchange, 8: 1, 4-5.

Stossel, T. 2007. "Regulation of Financial Conflicts of Interest in Medical Practice and Medical Research." Perspectives in Biology and Medicine, 50: 54-71.

Stossel, T. 2005. "Regulating Academic-Industrial Research Relationship--Solving Problems or Stifling Progress?" New England Journal of Medicine, 353: 1060-1065.

Supreme Court of the United States. 2011. William H. Sorrell, Attorney General of Vermont, et al., Petitioners v. IMS Health Inc. et al. No. 10-779. Opinion of the Court, 23 June.

Vladutiu, G. 2008. "Genetic Predisposition to Statin Myopathy." Current Opinion in Rheumatology, 20: 648-655.

George A. Chressanthis, Professor of Healthcare Management and Marketing, Department of Risk, Insurance and Healthcare Management, Fox School of Business, Temple University. Professor, Department of Clinical Sciences, Temple School of Medicine, 618 Alter Hall, 1801 Liacouras Walk, Philadelphia, PA 19122. Phone: 215-204-6030 Email: george.chressanthis@temple.edu

Nayla G. Dahan, Lecturer, School of Business, Gwynedd Mercy University, 893 Upper State Road, Chalfont, PA 18914. Phone: 215-908-0585 Email: dahan.n@gmercyu.edu

Kevin J. Fandl, Assistant Professor, Department of Legal Studies, Fox School of Business, Temple University, 410A Alter Hall, 1801 Liacouras Walk, Philadelphia, PA 19122. Phone: 215-204-6674 Email: kevin.fandl@temple.edu

Research was funded by AstraZeneca Pharmaceuticals LP.

TABLE 1.
Drug Product Selection (DPS) Laws by State Governing
Generic Substitution: l=Yes and 0=No per Category

                   Patient                 State Drug
State              Consent    Permissive   Formulary

Alabama               0           1            0
Alaska                1           1            0
Arizona               1           1            0
Arkansas              1           1            0
California            1           1            0
Colorado              1           1            0
Connecticut           1           1            0
Delaware              1           1            1
Florida               1           0            1
Georgia               1           1            0
Hawaii                1           0            1
Idaho                 1           1            0
Iowa                  1           1            0
Illinois              1           1            1
Indiana               1           1            0
Kansas                1           1            0
Kentucky              1           0            1
Louisiana             1           1            0
Maine                 1           1            0
Maryland              1           1            0
Massachusetts         0           0            1
Michigan              1           1            0
Minnesota             1           0            1
Mississippi           1           0            0
Missouri              1           1            1
Montana               1           1            0
Nebraska              1           1            1
Nevada                0           0            1
New Hampshire         0           1            1
New Jersey            0           0            1
New Mexico            0           1            0
New York              1           1            1
North Carolina        1           1            0
North Dakota          1           1            0
Ohio                  1           1            0
Oklahoma              1           1            0
Oregon                            0            I
Pennsylvania          1           0            0
Rhode Island          0           0            0
South Carolina        1           1            0
South Dakota          1           1            0
Tennessee             0           1            0
Texas                 1           1            0
Utah                  1           1            1
Vermont               1           1            0
Virginia              1           1            0
Washington            0           0            0
Washington, D.C.      1           1            1
West Virginia         1           0            1
Wisconsin             1           1            1
Wyoming               0           1            0

                                         No Cost
                      Two Line           Saving
State                 Rx Format          Pass On

Alabama                   1                 0
Alaska                    0                 1
Arizona                   0                 0
Arkansas                  0                 1
California                0                 0
Colorado                  0                 0
Connecticut               0                 0
Delaware                  0                 0
Florida                   0                 0
Georgia                   0                 0
Hawaii                    0                 0
Idaho                     0                 0
Iowa                      0                 0
Illinois                  0                 0
Indiana                   1                 1
Kansas                    1                 0
Kentucky                  1                 0
Louisiana                 0                 0
Maine                     0                 1
Maryland                  0                 0
Massachusetts             0                 0
Michigan                  0                 0
Minnesota                 0                 0
Mississippi               1                 0
Missouri                  1                 0
Montana                   0                 0
Nebraska                  0                 1
Nevada                    0                 0
New Hampshire             0                 0
New Jersey                1                 0
New Mexico                0                 0
New York                  0                 0
North Carolina            1                 0
North Dakota              0                 0
Ohio                      0                 0
Oklahoma                  0                 1
Oregon                    0                 0
Pennsylvania              0                 0
Rhode Island              0                 0
South Carolina            1                 1
South Dakota              0                 1
Tennessee                 0                 0
Texas                     0                 0
Utah                      0                 1
Vermont                   0                 1
Virginia                  0                 0
Washington                1                 0
Washington, D.C.          0                 0
West Virginia             0                 0
Wisconsin                 0                 0
Wyoming                   0                 0

Source: National Association of Boards of Pharmacy (2008).

TABLE 2.
Explanatory Variable Names, Descriptions, and Data Sources

DPS State Pharmacy Law Variables

Patient Consent   1 =states where generic substitution
                  required explicit patient consent DPS
                  law exist. 0=states having "No Patient
                  Consent" serve as the omitted variable.

Permissive        l=states where permissive DPS law exist.
                  Permissive laws generally state that the
                  pharmacist may substitute; mandatory
                  laws state that the pharmacist must
                  substitute a lower priced product when
                  the prescription (Rx) was written for a
                  branded product and prescriber has not
                  prohibited substitution. 0=states having
                  a mandatory DPS law serve as the omitted
                  variable.

Drug Formulary    1 =states where required formulary DPS
                  law exist. A formulary is a list of drug
                  products. If it is "negative" formulary,
                  the pharmacist may substitute for any
                  drug not listed on the formulary. If it
                  is a "positive" formulary, the
                  pharmacist, may substitute for any drug
                  listed on the formulary. 0=states not
                  requiring a drug formulary serve as the
                  omitted variable.

Two-Line          l=states where a two-line Rx form is
Rx Format         mandated DPS law exist. Two-line Rx
                  forms have a "Drug Product Selection
                  Permitted" line on the left. 0=states
                  not requiring two-line Rx forms serve as
                  the omitted variable.

No Cost           1=states where pharmacists were not
Savings           required to pass on cost savings
Pass On           resulting from substitution to patients
                  DPS law exist. 0=requirement to pass on
                  cost savings serve as the omitted
                  variable.

Patient Attribute Variables

Age               Age of the patient as of the Rx date.

Male              1=male patient. 0=female patient serves
                  as the omitted variable.

Co-Morbidity      Number of co-morbidities are based on 13
Count             broad cardiovascular disease categories
                  created by grouping together related
                  ICD9 diagnoses. High cardiovascular risk
                  patients having one of the following
                  diagnoses: Diabetes; 250.X;
                  Hypertension: 401.X, 402.X, 403.X,
                  404.X, 405 .X; Anti-Platelet: 286.X.

Method of Payment Variables

Medicaid          l=payment of Rx through Medicaid. 0=cash
                  payment serves as the omitted variable.

Third Party       l=payment of Rx through third party drug
                  plans (includes Medicare Part D plans).
                  0=cash serves as the omitted variable.

Physician
Attribute
Variables

Physician         Five physician specialty grouping based
Specialty         on American Medical Association (AMA)
                  specialty codes using one-zero
                  categorical measures: 1 for physicians
                  who are cardiologists and nephrologists
                  (Card_Neph), diabetologists and
                  endocrinologists (Dia_Endo), internal
                  medicine (IM), and all others (Other). 0
                  for primary care physicians (PCP) serves
                  as the omitted variable. PCPs are
                  defined as the sum of family practice
                  (FP). general practice (GP), and doctor
                  of osteopathy (DO). Physicians may have
                  multiple specialties, though the
                  analysis used only one specialty
                  provided in the AMA database.

Physician         Number of statin prescriptions by
Rx Volume         physician.

Drug Attribute
Variables

Switch to a       l=switch to a lower dose. 0=switch to a
Lower Dose        higher dose serves as the omitted
                  variable.

Switch to a       1 =therapeutic switch. O=bioequivalent
Different         switch serves as the omitted variable.
Molecule

Institutional/External Variable

Managed           Managed care plan control metric
Care Plan         measures the degree that a healthcare
Control           plan is able to move product share
                  within a local area away from the
                  national share. Managed care control was
                  defined as the average control seen by a
                  physician across all plans in the statin
                  and proton pump inhibitor market using
                  July 2008 data.

Sources: American Medical Association (2006-2008). National
Association of Board of Pharmacy (2008). All IMS Health data
sources: IMS Health, Lifelink[R] Anonymized Patient Level Data
(2006-2008); IMS Health, Xponent[R] (2008); IMS Health,
Xponent PlanTrak[R] (2008), IMS Health Incorporated. All
Rights Reserved.

TABLE 3.
Descriptive Statistics and Total Number of Prescriptions and Patients

Variables                     Mean        SD       Min       Max

Dependent Variable
Switch category               1.43      0.49         1         2
Patient Attributes
Age                          64.24     12.18         1       108
Male                          0.51      0.50         0         1
Co-morbidity count            4.91      7.80         0       103
Physician Attributes
Physician Rx volume           5.64      2.84         1        12
PCP                           0.41      0.49         0         1
IM                            0.41      0.49         0         1
Card_Neph                     0.13      0.33         0         1
Endo Dia                      0.02      0.14         0         1
Other                         0.03      0.18         0         1
Method of Payment
Cash                          0.02      0.15         0         1
Medicaid                      0.02      0.14         0         1
Third party                   0.95      0.21         0         1
Drug Attributes
Switch to lower dose          0.45      0.50         0         1
Switch to                     0.96      0.18         0         1
  different molecule
DPS Pharmacy State Laws
Patient consent               0.82      0.39         0         1
Permissive                    0.47      0.50         0         1
Formulary substitution        0.40      0.49         0         1
Two-line Rx format            0.20      0.40         0         1
No cost saving pass on        0.06      0.24         0         1
Institutional/External
Managed care control          0.21      0.07         0      0.87

                            Generic Drug        Generic to Branded
                            Substitution        Drug Switch-Backs
Number of Prescriptions
  Therapeutic               563,108 (95.5%)     431,495 (97.7%)
  Bioequivalent             26,271 (4.5%)       10,298 (2.3%)
  Total                     589,379             441,793
Number of Patients
  Therapeutic               317,511 (93.5%)     179,284 (96.4%)
  Bioequivalent             21,916 (6.5%)       6,757 (3.6%)
  Total                     339,427             186,041

                            Combined Numbers for
                            Substitutions and Switch-Backs
Number of Prescriptions
  Therapeutic               994,603 (96.5%)
  Bioequivalent             36,569 (3.5%)
  Total                     1,031,172
Number of Patients
  Therapeutic               372,187 (93.7%)
  Bioequivalent             24,924 (6.3%)
  Total                     397,111

Sources: American Medical Association (2006-2008). National
Association of Board of Pharmacy (2008). All IMS Health data
sources: IMS Health, Lifelink[R] Anonymized Patient Level
Data (2006-2008); IMS Health, Xponent[R] (2008); IMS Health,
Xponent PlanTrak[R] (2008), IMS Health Incorporated. All
Rights Reserved.

TABLE 4.
Logistic Regression Model Estimates Ranked
by Absolute Value

Dependent Variable = Switch Category
(l=substitution to generic drug,
2 = switch-back to branded drug

                                            Log Odds
Explanatory Variables                      Estimates

Switch to lower dose                          2.695 ***
Intercept                                    -2.308 ***
Managed care plan control                    -1.531 ***
Switch to different molecule                  0.932 ***
Card_Neph                                     0.448 ***
Endo_Dia                                      0.319 ***
Medicaid payment                              0.239 ***
Other physicians                              0.189 ***
Drug formulary                                0.103 ***
No cost saving pass on                        0.083 ***
Two-line Rx form                              0.079 ***
Third party payment                           0.068 ***
Patient consent                              -0.063 ***
Permissive                                   -0.040 ***
Male                                          0.028 ***
Co-morbidity count                            0.022 ***
IM                                            0.017 **
Physician Rx volume                           0.011 ***
Age                                          -0.003 ***
N                                     1,031,172
Proxy [R.sup.2]                              81.9%

IMS Health data sources: IMS Health, Lifelink[R]
Anonymized Patient Level Data (2006-2008): IMS Health.
Xponcnf (2008); IMS Health, Xponent PlanTrak[R] (2008),
IMS Health Incorporated. All Rights Reserved.

SAS 9.2 used in all statistical and diagnostic tests.

Ranking of the log odds estimates are from highest to
lowest in absolute values.

Standard errors are available upon request.
Statistical significance levels: *** p < 0.0001;
** p < 0.01.