To systematically review the evidence on screening asymptomatic adults for cardiovascular disease (CVD) risk using resting or exercise electrocardiography (ECG) for populations and settings relevant to primary care in the United States.
PubMed/MEDLINE, the Cochrane Library, and trial registries through May 30, 2017; reference lists of retrieved articles; outside experts; and reviewers, with surveillance of the literature through April 4, 2018.
Two investigators selected English-language studies using a priori criteria. Eligible studies focused on the use of resting or exercise ECG for adults without symptoms or a diagnosis of CVD. Eligib le de signs included controlled trials comparing ECG screening with no ECG screening and prospective cohort studies reporting reclassification, calibration, or discrimination that compared risk assessment using ECG plus traditional risk factors versus traditional risk factors alone. For harms of ECG, prospective cohort studies, large retrospective cohort studies, and case-control studies were also eligible. For harms from exercise ECG or subsequent procedures/interventions, large registries or multicenter studies without a control group were also eligible.
One investigator extracted data and a second checked accuracy. Two reviewers independently rated quality for all included studies using predefined criteria.
Sixteen studies (77,140 participants) were included. Two randomized, controlled trials (RCTs) (1,151 participants) found no significant improvement in all-cause mortality, cardiovascular-related mortality, myocardial infarction (MI), heart failure, or stroke for screening with exercise ECG in asymptomatic adults ages 50 to 75 years with diabetes compared with no screening. In addition, there was no significant improvement for their primary composite outcomes (hazard ratio [HR] 1.00 [95% confidence interval [CI], 0.59 to 1.71] for all-cause mortality, nonfatal MI, nonfatal stroke, or heart failure requiring hospitalization or emergency service intervention, and HR 0.85 [95% CI, 0.39 to 1.84] for nonfatal MI or cardiac death). No controlled trials evaluated screening with resting ECG. Although potential harms of exercise or resting ECG include arrhythmias, acute MI, sudden cardiac death, and harms of subsequent angiography or revascularization procedures after an abnormal test, evidence on their frequency in asymptomatic persons was scant. Evidence from five cohort studies (9,582 participants; mean baseline Framingham Risk Score [FRS] 10.8 to 12.3 in studies reporting it) shows that the addition of exercise ECG abnormalities to traditional CVD risk factors results in small improvements in discrimination (absolute improvement in area under the curve [AUC] or C-statistics 0.02 to 0.03; 95% CIs rarely reported), but it is uncertain whether calibration or appropriate risk classification improves. Evidence from nine cohort studies (66,407 participants; mean baseline risk ranging from low to high across studies) shows that the addition of resting ECG findings to traditional CVD risk factors results in very small or small improvements in discrimination (absolute improvement in AUC or C-statistics 0.001 to 0.05) and improvements for calibration and appropriate risk classification for prediction of multiple outcomes (e.g., all-cause mortality, CVD mortality, CHD events). Total net reclassification improvements (event;nonevent) ranged from 3.6 percent (2.7%; 0.6%) to 30 percent (17%; 19%) for studies using Framingham Risk Score (FRS) or pooled cohort equations (PCE) base models (95% CIs were rarely reported).
The RCTs that evaluated exercise ECG in asymptomatic diabetic patients did not reach sample size targets and were stopped early because of trouble recruiting; both followed participants for about 3.5 years. For risk prediction with the addition of ECG, evidence was limited by imprecision, quality, and considerable heterogeneity. Consistency of findings for specific risk thresholds is unknown because all studies used different risk categories. About half of the included risk prediction studies did not use the published coefficients of externally validated base models such as FRS or PCE; only one used the PCE as a base model. For risk prediction with resting ECG, it is unclear what proportion of participants was truly asymptomatic because most studies did not report any assessment of symptoms.
The overall strength of evidence was low or insufficient for each of the questions and outcomes evaluated. RCTs of screening with exercise ECG in asymptomatic participants found no improvement in health outcomes despite focusing on higher risk populations with diabetes. For asymptomatic persons without a history of CVD, the harms of exercise or resting ECG can include arrhythmias, acute MI, sudden cardiac death, and harms of subsequent angiography or revascularization procedures after an abnormal test, but the frequency of these harms is uncertain. Evidence on whether the addition of exercise ECG to traditional CVD risk factors results in accurate reclassification is lacking. Cohort studies found that the addition of multiple resting ECG abnormalities to traditional CVD risk factors accurately reclassifies persons, and improves discriminat ion and calibration, but evidence was limited by imprecision, quality, considerable heterogeneity, and inconsistent use of risk thresholds that align with clinical decisions and recommendations.
