Article type
Year
Abstract
Introduction: The literature on the use of estrogens in the prevention of osteoporosis lacks RCTs reporting fracture as the main clinical outcome.
Objective: To examine the quality of the observational studies that represent the main source of evidence on the effect of estrogen treatment on hip fracture risk.
Methods: Searching of Medline and Excerpta Medica databases, bibliographies of published reviews, and review of the reference list by experts and pharmaceutical manufacturers. Data extraction and quality assessment was performed in duplicate. Fixed- and random-effects models were used to pool relative risks (RR - cohort studies) and odds ratios (OR - case-control studies).
Results: Five cohort and 10 case-control studies of estrogens and hip fracture were identified (1977-1993). For the case-control studies, total quality scores tended to increase over time and were inversely related to estrogen's estimated protective benefit. However, none of the individual quality items appear to have a direct relationship with the estimated odds ratios. The pooled estimates of effect differed for the case-control (OR=0.57, 95% CI 0.48-0.67) and cohort (RR=0.85, 0.68-1.07) studies. Plotting the results for the individual studies against total quality scores suggests that the better quality case-control studies yielded smaller estimates of protective benefit that were more consistent with those from the cohort studies.
Discussion: In this literature there was a relationship between study quality and estimated effect of estrogens, with the better quality studies suggesting a smaller protective benefit. The poorer quality case-control studies appeared to over-estimate the true effect of estrogens in the prevention of hip fracture.
Objective: To examine the quality of the observational studies that represent the main source of evidence on the effect of estrogen treatment on hip fracture risk.
Methods: Searching of Medline and Excerpta Medica databases, bibliographies of published reviews, and review of the reference list by experts and pharmaceutical manufacturers. Data extraction and quality assessment was performed in duplicate. Fixed- and random-effects models were used to pool relative risks (RR - cohort studies) and odds ratios (OR - case-control studies).
Results: Five cohort and 10 case-control studies of estrogens and hip fracture were identified (1977-1993). For the case-control studies, total quality scores tended to increase over time and were inversely related to estrogen's estimated protective benefit. However, none of the individual quality items appear to have a direct relationship with the estimated odds ratios. The pooled estimates of effect differed for the case-control (OR=0.57, 95% CI 0.48-0.67) and cohort (RR=0.85, 0.68-1.07) studies. Plotting the results for the individual studies against total quality scores suggests that the better quality case-control studies yielded smaller estimates of protective benefit that were more consistent with those from the cohort studies.
Discussion: In this literature there was a relationship between study quality and estimated effect of estrogens, with the better quality studies suggesting a smaller protective benefit. The poorer quality case-control studies appeared to over-estimate the true effect of estrogens in the prevention of hip fracture.