Screening Mammography & Breast Cancer Mortality: Meta-Analysis of Quasi-Experimental Studies

BACKGROUND: The magnitude of the benefit associated with screening has been debated. We present a meta-analysis of quasi-experimental studies on the effects of mammography screening. METHODS: We searched MEDLINE/PubMed and Embase for articles published through January 31, 2013. Studies were included if they reported: 1) a population-wide breast cancer screening program using mammography with 5+ years of data post-implementation; 2) a comparison group with equal access to therapies; and 3) breast cancer mortality. Studies excluded were: RCTs, case-control, or simulation studies. We defined quasi-experimental as studies that compared either geographical, historical or birth cohorts with a screening program to an equivalent cohort without a screening program. Meta-analyses were conducted in Stata using the metan command, random effects. Meta-analyses were conducted separately for ages screened: under 50, 50 to 69 and over 70 and weighted by population and person-years. RESULTS: Among 4,903 published papers that were retrieved, 19 studies matched eligibility criteria. Birth cohort studies reported a significant benefit for women screened <age 50, but not for women screened ages 50–69. Significant reductions in breast cancer mortality were observed in historical comparisons. For geographical comparisons, there was a significant 20% reduction in mortality for women <age 50 and a significant 21–22% reduction for women ages 50–69. Studies that tested the interaction of geographical and historical comparisons produced a pooled, significant 13–17% reduction in incident breast cancer mortality for women ages 50–69, but the effects in most individual studies were non-significant. All studies of women ages 70+ were non-significant. CONCLUSIONS: Mammography screening may have modest effects on cancer mortality between the ages of 50 and 69 and non-significant effects for women older than age 70. Results are consistent with meta-analyses of RCTs. Effects on total mortality could not be assessed because of the limited number of studies.