摘要：In their letter, Kogevinas and Pearce (2012) suggested that meta-analyses should be more routinely prepared for the evaluations of the International Agency for Research on Cancer (IARC) Monographs program. We concur that meta-analyses are useful in many cases, but there are also counter examples where they have not been useful. For example, when Kogevinas et al. (1998) reviewed the carcino-genicity of cancer hazards in the rubber-manufacturing industry, they argued against using meta-analytic techniques because of the hetero geneity of exposure circumstances within and between manufacturing plants and differences of exposure classifications used in the studies. They concluded that a single summary risk estimate would be uninformative. Based on their systematic narrative review, the authors concluded that there is an increased risk of neoplasms of the urinary bladder, lung, and larynx and an increased risk of leukemia (Kogevinas et al. (1998). In contrast, Alder et al. (2006) performed a meta-analysis of cancer occurrence among workers in the rubber-manufacturing industry. Based on summary estimates for the entire rubber industry and two major sectors of this industry, these authors concluded that excesses other than for leukemia were not substantiated by their synthetic meta-analysis (Alder et al. 2006). After reviewing all the pertinent studies, a later IARC Working Group concluded that there is sufficient evidence for an increased risk of several types of cancer in rubber manufacturing (Baan et al. 2009). In contrast, when t he IARC Working Group for Volume 98 reviewed the evidence on shift work and cancer, a published meta-analysis had reported a statistically signifi-cantly increased risk for breast cancer among women who regularly worked the night shift (Megdal et al. 2005). Nevertheless, the IARC Working Group concluded that there was only limited evidence for carcino genicity in humans (IARC 2010).
