Comparison of Apples with their Parts : Rethinking a Meta-analysis on the Association between Cooking and Uveal Melanoma

We read with interest Ge et al. (2012) article entitled “Occupational Cooking and Risk of Uveal Melanoma: a Metaanalysis”. We believe that this meta-analysis is helpful because a synthesis of the growing evidence of the association between cooking and uveal melanoma risk is overdue. Uveal melanoma is a rare disease and only a few consistent risk factors have been identified (Behrens et al., 2011; Epidemiologisches Krebsregister NRW, 2013). However, we are concerned about three aspects. First, the authors included published study results that were not derived from independent study samples. Guenel et al. (2001), Stang et al. (2003), Lutz et al. (2005) and Behrens et al. (2011) used data of the multicentric case– control study entitled “Study of Occupational Risk Factors for Rare Cancers of Unknown Aetiology” (in short “Rare Cancer Study”) from nine European countries (Ge et al., 2012). Guenel et al. (2001). similar to Stang et al. only included their national data, that is, French and German data, respectively. Lutz et al. used data from Italy, Denmark, Latvia, Sweden, Spain, Portugal. That means, these three papers used disjunct data sets from the international Rare Cancer Study. Finally, Behrens et al. (2011) used the largest data set of the Rare Cancer Study including France, Germany, Italy, Denmark, Latvia, Sweden, Spain, Portugal, UK. Therefore, the effect estimate from Behrens et al. is not independent of the effect estimates presented in the previous papers of the Rare Cancer Study (Table 1). We therefore excluded the data from Behrens et al. and re-ran the metaanalysis of Ge et al. (2012) to avoid statistical dependence between the study-specific effect estimates. We provide both results from the fixed-effect and from the random-effect model. The newly estimated Comparison of Apples with their Parts: Rethinking a Metaanalysis on the Association between Cooking and Uveal Melanoma Risk


Dear Editor
We read with interest Ge et al. (2012) article entitled "Occupational Cooking and Risk of Uveal Melanoma: a Metaanalysis".We believe that this meta-analysis is helpful because a synthesis of the growing evidence of the association between cooking and uveal melanoma risk is overdue.Uveal melanoma is a rare disease and only a few consistent risk factors have been identified (Behrens et al., 2011;Epidemiologisches Krebsregister NRW, 2013).However, we are concerned about three aspects.
First, the authors included published study results that were not derived from independent study samples.Guenel et al. (2001), Stang et al. (2003), Lutz et al. (2005) and Behrens et al. (2011) used data of the multicentric casecontrol study entitled "Study of Occupational Risk Factors for Rare Cancers of Unknown Aetiology" (in short "Rare Cancer Study") from nine European countries (Ge et al., 2012).Guenel et al. (2001).similar to Stang et al.only included their national data, that is, French and German data, respectively.Lutz et al. used data from Italy, Denmark, Latvia, Sweden, Spain, Portugal.That means, these three papers used disjunct data sets from the international Rare Cancer Study.Finally, Behrens et al. (2011) used the largest data set of the Rare Cancer Study including France, Germany, Italy, Denmark, Latvia, Sweden, Spain, Portugal, UK.Therefore, the effect estimate from Behrens et al. is not independent of the effect estimates presented in the previous papers of the Rare Cancer Study (Table 1).
We therefore excluded the data from Behrens et al. and re-ran the metaanalysis of Ge et al. (2012) to avoid statistical dependence between the study-specific effect estimates.We provide both results from the fixed-effect and from the random-effect model.The newly estimated

Comparison of Apples with their Parts: Rethinking a Metaanalysis on the Association between Cooking and Uveal Melanoma Risk
pooled OR for the association between cooking and uveal melanoma risk were higher than the original once reported by Ge et al. ( 2012    This is not correct.However, data from Monarrez-Espino et al. (2002) and Stang et al. (2003) used in part the same population.Therefore it was correct to exclude the study from Monarrez-Espino et al. (2002).Third, Ge et al. (2012) concluded that "more welldesigned observational studies with large sample size or prospective cohort studies are needed".To the best of our knowledge, the case-control studies included in the meta-analysis by Ge et al. (2012) were well-designed.Furthermore, a prospective cohort study appears to be statistically inefficient.For example, a prospective population-based cohort study of people aged 40-79 years in Germany that included 200.000 people would provide an expected number of 28 newly diagnosed uveal melanoma events during a 10 year follow-up.For this calculation, we used current incidence estimates (2006)(2007)(2008)(2009)(2010) of the cancer registry in Northrhine-Westphalia (Epidemiologisches Krebsregister NRW, 2013), Germany and assumed that no cohort members died due to competing outcomes.
Our concerns emphasize that obviously all three stages of peer-review are important for quality control: pre-peer review (authors), peer-review (journal), and post publication peer-review (readers) (Stang et al., 2008).

Figure
Figure 1.Meta-analysis of the Association between Occupational Cooking and the Risk Of Uveal Melanoma (fixed-effect model)

Table 1 . Overview of the Five Publications on the Etiology of Uveal Melanoma that Contributed to the Pooling by Ge et al.
RIFAStudy of Occupational Causes of Rare Cancers of Unknown Aetiology (Rare Cancer Study)