TO THE EDITOR
I recently read the systematic review and meta-analysis conducted by Ilic et al comprising 10 case-control studies and 11 cohort studies, which concluded that the use of oral contraceptives (OCU) was associated with a decreased risk of pancreatic cancer in women (PCW) [summary relative risk (sRR) = 0.85; 95% confidence intervals (CI) = 0.73-0.98; P = 0.03]. Interestingly, the subgroup analysis according to the study design showed no statistical significance in case-control studies but showed borderline statistical significance in cohort studies (sRR = 0.84; 95%CI = 0.70-1.00; P = 0.05).
However, while reviewing the results of the 11 selected cohort studies, I found the following potential errors. First, among the 11 selected studies, the study by Teras et al was a cohort study that analyzed the mortality of PCW; therefore, excluding this study would be valid based on the research hypothesis; second, it would be necessary to include the two cohort studies[3,4] that were considered in other studies on the risk of various cancers associated with OCU[5,6]; finally, in the two studies that did not provide an RR for the ever group[7,8], the RR's direction was opposite to that of the forest plot shown in the study by Ilic et al.
Considering these issues, I recalculated the sRR of the longest duration (LD) group as well as the ever group. The statistical significance disappeared in both groups, and the sRRs were 1 or higher (Figure 1). Egger’s test was performed to evaluate publication bias, and no statistical significance was noted in either group (P = 0.439 and 0.817 in the ever group and LD group, respectively).
Figure 1 Forest plots in the ever and the longest duration group.
Eight of the 12 selected cohorts[3,7-13] provided the information necessary for performing a dose-response meta-analysis. A two-stage random-effects dose-response model was used with a dosing unit of 1 year (P of goodness-of-fit = 0.041). The results showed borderline statistical significance with a linear dose-response relationship between OCU duration and PCW risk (sRR = 1.015; 95%CI = 0.999-1.030; P = 0.057) (Figure 2).
Figure 2 The linear dose-response relationship between duration (year) of oral contraceptive usage and risk of pancreatic cancer in women.
RR: Relative risk.
Based on the results of the recalculated sRRs and DRMA, the OCU may not be associated with the risk of PCW. Because my conclusion contradicts that reported by Ilic et al, it is necessary to re-evaluate the direction and statistical significance of risk through an updated meta-analysis in the future.