Table 4.
Evidence summary of benefits associated with screening mammography: The content of the evidence profile table is provided in 13 standardized columns. The first column provides information about the number of studies and the study design used to determine the effectiveness of screening mammography for women in this age range (N = 8 RCTs). Columns 2 to 7 provide an assessment of the quality of these studies. Footnotes provide further explanations as required. For instance, in column 3 (risk of bias) we indicate a serious concern about the potential risk of bias in the studies. This is based on the fact that only 3 of the 8 trials were considered truly randomized; in 5 of the trials it was not clear if investigators were blinded to the groups to which the patients were assigned or whether those enrolling patients were aware of which group patients were being assigned to. There were no other concerns about quality: results of all trials were consistent, the patients and the interventions were similar to the patients that we were studying, the samples sizes were large, the CIs were narrow, and there was no evidence of publication bias. Columns 8 to 11 in the table present the summary of our meta-analysis to determine the overall effectiveness of mammography screening in women aged 40–49 y. The number of deaths seen in the control and experimental groups are provided in columns 8 and 9. In columns 10 and 11, the estimates of the relative and absolute risk reductions that can be attributed to screening mammography are provided. Relative risk is used to compare risks between 2 different groups of people, often those who were exposed to an intervention and those who were not. Meta-analysis of mammography screening studies with women aged 40–49 y found a reduction of breast cancer risk of 15% (equivalent to an RR of 0.85) for women who were screened compared with women who were not screened. Absolute risk focuses on an individual’s risk of getting a disease in a specific period of time and can be expressed as a percentage or a rate (eg, 10% or 1 in 10). In this example, this means that 474 fewer women per million (or 1 in approximately 2100) will die as a result of screening. Column 12 provides an overall rating of the quality. Column 13 highlights the importance of the results.
|
SUMMARY OF FINDINGS
|
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
QUALITY ASSESSMENT FOR BREAST CANCER MORTALITY AT AGE 40–49 Y*W
|
NO. OF BREAST CANCER DEATHS/ NO. OF PATIENTS (%)
|
ESTIMATE OF MORTALITY EFFECT
|
QUALITY | IMPORTANCE | ||||||||
| NO. OF STUDIES | DESIGN | RISK OF BIAS | INCONSISTENCY | INDIRECTNESS | IMPRECISION | OTHER CONSIDERATIONS | SCREENING WITH MAMMOGRAPHY | CONTROL | RELATIVE (95% CI) | ABSOLUTE | ||
| 8 | Randomized trials† | Serious‡ | No serious inconsistency§ | No serious indirectness|| | No serious imprecision¶ | None# | 448/152 300 (0.29) | 625/195 919 (0.32) | RR=0.85** (0.75 to 0.96) | 474 fewer per 1 000 000 (from 115 fewer to 792 fewer) NNS=2108†† | Moderate | Critical |
NNS—number needed to screen, RCT—randomized controlled trial, RR—relative risk.
The available data were based on women aged 39–49 y; however, the focus of the review was for those aged 40–49 y.
Of the 8 studies, 5 were quasi-randomized and 3 were truly randomized.
Blinding and concealment were not clear for 5 studies, so only 3 trials were considered truly randomized.
No heterogeneity exists; P value for testing heterogeneity is 0.48 and I2 = 0%.4
The question addressed is the same for the evidence regarding the population, intervention, comparator, and outcome.
Total sample size is large and the total number of events is > 300 (a threshold rule-of-thumb value).
Insufficient number of studies to assess publication bias.
Estimates are based on a random-effects meta-analysis.
NNS is the number of patients who would need to be invited to be screened about once every 2 years over a median of about 11 years to prevent 1 death from breast cancer.