Author's reply to: Meta‐analysis of cancer risks of professional firefighters

We would like to thank Dr Casjens and colleagues for their critical and thorough comments on our article: cancer incidence and mortality among firefighters.1 We are happy to discuss all the points raised below. The authors argue for stratification of the results by study design and disagree with the pooling of different risk estimates. We agree that pooling of different risk estimates can lead to biased results. However, if the outcome is rare (such as all cancers studied in our review) and the risk estimates are close to one (almost all cancers among firefighters) estimates from case–control and cohort studies can be pooled without bias.2, 3 Because of the few numbers of studies in some of the cancer types, we therefore preferred to not stratify by study design. Moreover, the proportion of case–control studies among all included studies was maximally 30% (for incidence/mortality of lung cancer). It is true that using multiple estimates of specific cancer sites from one study would give more weight to this one study. However, since it was only two studies that reported two estimates based on the same number of cases, we do not think that this distorts our results. Because the estimates reported differed, we thought it was important to include both of them. For example, Ahn et al.4 reported a significantly elevated standardized mortality ratio of 1.56 (95% confidence interval [95% CI] 1.01–2.41) for kidney cancer, with no significant elevation of standardized rate ratio (0.69; 95% CI 0.16–2.99) for this organ. It is correct that the method by Hamling et al.5 was originally developed to aggregate categories of exposure within one variable but also to aggregate estimates of different categories of disease. We therefore think that it can be used without bias to aggregate estimates of cancers of different ICD codes. The alternative would have been to combine them with fixed effects meta-analysis, but this approach treats each estimate as an independent measure, which we think is not correct. The method of Hamling et al. takes the correlation between the estimates into account and this is why we preferred this method. In any case, out of approximately 800 extracted risk estimates, only one standardized mortality ratio, five standardized incidence ratios and three odds ratios were calculated with the method of Hamling et al. We therefore think that using another method would not have changed the overall results of this meta-analysis. We fully agree with Casjens and colleagues’ comment on our criteria to assess the strength of association between firefighting occupation and cancer. The method uses arbitrary thresholds (similar to the widely used significance level of p < 0.05) to interpret findings. This is why we show all the individual estimates with confidence intervals and just add the interpretation by applying the method of LeMasters et al. This gives the reader an easily understandable idea on the strength of associations and makes the results comparable to the last meta-analysis of LeMasters et al.6 In addition, we tried to enhance the method by separating incidence and mortality risk estimates. We used the DerSimonian–Laird estimator to calculate the between-study variance. We agree that studies have shown that the Paule and Mandel estimator might be more accurate under some circumstances. We decided to use the DerSimonian–Laird estimator because it is most widely used and makes our results comparable. The DerSimonian–Laird estimator performs well with low mean squared errors when τ2 is small.7 Although the Newcastle–Ottawa Scale (NOS) is not the best measure to assess the quality of studies in systematic reviews, it is widely used and accepted for this aim and has been suggested by several investigators.8, 9 However, evidence suggests that quality scales can be problematic and the conclusions of a meta-analysis can be affected by the choice of the quality scale.9 We agree that the NOS was not sensitive to differentiate well between the studies included in our review. We have therefore also put little emphasis on the NOS rating when interpreting our results. In our meta-analysis, we primarily extracted effect sizes of male firefighters but included studies were the estimates included both sexes combined. Casjenns and colleagues were concerned about this approach and suggested to add an analysis in men only. We agree that this approach could be problematic if there were many studies including men and women. However, we extracted more than 800 risk estimates through 48 studies and only six estimates included both sexes combined. Furthermore, the populations in the underlying studies included less than 5% women. We therefore think that a bias because of some estimates including also women is unlikely. We confirm that there is a problem with the graphical representation of the confidence intervals in two studied cancers (intestine and colon) of Figure 2. The underlying numbers for the calculations and reporting of these numbers in the text are, however, correct. Additionally, we spotted that in Figure 2, the number of studies for colon cancer must be rectified to 10. Finally, we tried to find and include all relevant studies by using a broad search strategy including a wide range of words for the population (firefighters) and outcome (cancer) of interest in three important databases (see Supporting Information Table S2 of the article), identifying 2,630 records. In addition, we screened the reference lists of previous reviews on the same topic. However, we can of course not guarantee that we have not missed some publications. As described in Figure 1 of our article, we excluded seven publications that overlapped with other studies (based on the same data material). Moreover, we exclude additional nine studies, including Demers et al.,10 because they reported risk estimates that include more than one occupation. For example, Demers et al.10 reported an odds ratio of 1.90 (95% CI 0.50–9.40) for multiple myeloma among “firefighting and prevention occupations.” Yours sincerely, Hamed Jalilian Mansour Ziaei Elisabete Weiderpass Yahya Khosravi Kristina Kjaerheim Corina S. Rueegg Where authors are identified as personnel of the International Agency for Research on Cancer / World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer / World Health Organization.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.