The increased risk of venous thromboembolism and the use of third generation progestagens: Role of bias in observational research

A matched case-control study was undertaken in 10 centers in Germany and the United Kingdom to explore the association of current use of major combination oral contraceptives with the occurrence of venous thromboembolism. The cases recruited were 505 women aged 16-44 years with venous thromboembolism, controls were 1877 women (at least 3 controls per case) matched for 5-year age group and region without VTE. The main outcome measures were odds ratios derived by comparing current use of a specific oral contraceptive or group of OC against current use of other groups or against no current use of OC. The odds ratios (95% confidence intervals) for venous thromboembolism were: for third generation products (low dose ethinyloestradiol, gestodene and desogestrel) versus second generation products (low dose ethinyloestradiol, no gestodene and desogestrel, 1.5 (1.1 to 2.0), for third versus second generation products with norgestimate included in third generation, 1.6 (1.2 to 2.2). The odds ratios for current use for women aged 16-44 of specific progastagens versus levonorgestrel-containing compounds were 1.7 (1.1 to 2.6) for gestodene, 1.8 (1.2 to 2.6) for desogestrel, 1.9 (1.0 to 3.6) for norgestimate and 1.3 (0.7 to 2.5) for progestagen-only pills. For women aged 25 to 44 likely to be exposed to any of these progestagens, odds ratios for the comparison of progestagens versus levonorgestrel showed a successive increase by market introduction ranging from 1.5 (0.9 to 2.5) for desogestrel with 30 pg oestrogen content (introduced 1981) to 2.8 (1.3 to 6.5) for desogestrel with 20 mu g oestrogen content (introduced 1992) significant in linear trend analysis (p=0.00012). The influence of norgestimate classification as third or second generation product does not significantly alter the results regarding the association of third generation products and venous thromboembolism. A direct comparison of current use of norgestimate (which is primarily metabolized to levonorgestrel) versus levonorgestrel shows an increased odds ratio. The trend of increasing risk of progestagens by recency of market introduction when compared with levonorgestrel is strongly indicative of the existence of external bias due to attrition of susceptibles.