Sixteen trials (15,236 patients) were included. Most trials had at least one potential bias associated with methodology, which resulted in moderate to high risk of bias in 15 of the 16 included trials.
There were significantly fewer pregnancies with TCu380A compared to MLCu375 at all time points including five-year follow-up (RR 0.49, 95% CI 0.33 to 0.72; 16 trials). There were significantly fewer expulsions with TCu380A compared to MLCu375 at five years (RR 0.63, 95% CI 0.45 to 0.89; 16 trials) and at three years, but not at earlier follow-up.
There were more removals with TCu380A than MLCu375 for bleeding and/or pain at three years (RR 1.83, 95% CI 1.28 to 2.62; 13 trials) and at three months and two years, but not at six months, one year or five years.
There was higher risk of pelvic inflammatory disease when using TCu380A compared to MLCu375 when all studies that measured pelvic inflammatory disease were pooled irrespective of length of follow-up (RR 1.47, 95% CI 1.03 to 2.11; eight trials). There were no significant differences in continuation rates, uterine perforation or removal for bleeding at any measured follow-up.
Visual examination of the funnel plots for pregnancy and continuation outcomes showed no discrimination in effects between large and small studies.