The numbers given are the number of cases:
Ever versus never oral contraceptive use: Population based: 3 cohort studies (571 women) and 3 case-control studies (353 women).
Hospital-based: 4 case-control studies (589 women).
Current versus never oral contraceptive use: Population based: 3 cohort studies (354 women) and 2 case-control studies (205 women).
Hospital-based: 3 case-control studies (341 women).
Longest duration of exposure in years versus never oral contraceptive use: Population based: 2 cohort studies (208 women) and 2 case-control studies (more than 47 women).
Hospital-based: 3 case-control studies (more than 55 women).
Four meta-analyses of studies on oral contraceptive use and the development of rheumatoid arthritis were used to compare ever use versus never used.
The intraclass correlation coefficient bewteen the two raters was 0.68 (95%CI: 0.13, 0.91). For part A initial scoring produced agreement on roughly 70% of the individual items and part B produced agreement on approximately 40% on the individual items.
The regression coefficient for the overall quality was not significantly different from zero (P = 0.308). The overall quality index was therefore not used as a weight in the final analysis but was used in the sensitivity analysis.
A funnel plot using the individual studies provided no evidence of publication bias.
The 'odd man ou't procedure using ever versus never used produced a disjointed area suggesting heterogeneity.
Summary statistics: results from direct method used where heterogeneity chi-squared P > 0.10, otherwise DerSimonian and Laird method results given.
Ever versus never use of oral contraceptives: overall RR 0.84 (95%CI: 0.66, 1.08), heterogeneity chi-squared P = 0.0006;population based: RR = 1.01 (95%CI: 0.87, 1.17) heterogeneity chi-squared P = 0.5924; hospital based: RR = 0.54 (95%CI: 0.36, 0.80) heterogeneity chi-squared P = 0.0748.
Current versus never use of oral contraceptives: overall RR 0.93 (95%CI: 0.66, 1.32), heterogeneity chi-squared P = 0.0100;population based: RR = 1.05 (95%CI: 0.83, 1.34) heterogeneity chi-squared P = 0.1557; hospital based: RR = 0.67 (95%CI: 0.37, 1.19) heterogeneity chi-squared P = 0.0645.
Longest duration versus never use of oral contraceptives: overall RR 0.91 (95%CI: 0.58, 1.42), heterogeneity chi-squared P = 0.0135; population based:RR = 1.11 (95%CI: 0.81, 1.53) heterogeneity chi-squared P = 0.3795; hospital based: RR = 0.71 (95%CI: 0.37, 1.37) heterogeneity chi-squared P = 0.0162.
In the meta-regression, only the source of cases and the source of controls were statistically significant in accounting for the heterogeneity. A final model including only the source of controls as an explanatory variable showed no evidence of residual heterogeneity.
In the sensitivity analysis, done by omitting certain studies, none of the analysis resulted in a significant change and heterogeneity was always detected. Adding a weighting of the overall quality resulted in only minor changes and strong heterogeneity was still detected. No differences were found after repeating the analysis, once the three studies with the lowest quality rating were omitted.