Systematic Reviews to Answer Health Care Questions

193

Chapter 12 • Assessing and Rating the Strength of the Body of Evidence

navigation refers to services that provide personal guidance through the healthcare sys tem to meet an individual patient’s needs during the course of care. In the cancer screen ing studies, patient navigation included any of an array of services, such as education, scheduling, transportation, information, financial, referral, and reminders. A strength of evidence summary is outlined in Table 12.4. Using the AHRQ EPC approach, the strength of evidence for each type of cancer screening was assessed for: study limitations (low, medium, or high level); consistency (consistent, inconsistent, or unknown/none); directness (direct or indirect); precision (precise or imprecise); and reporting bias (suspected or undetected). Study limitations: The study limitations domain was rated medium for colorectal, breast, and cervical cancer screening because most included studies were fair-quality RCTs with some methodological deficiencies. The single lung cancer screening trial was given a poor-quality rating because of lack of reporting on randomization and allocation con cealment, unclear masking of assessors or patients, and large loss to follow-up. This led to a domain rating of high study limitations. Consistency: Most studies for all types of cancer screening showed increased screening rates for intervention versus usual care groups. Consequently, colorectal, breast, and cervical cancer screening studies achieved a rating of consistent for the consistency domain. This rating was none for the single lung cancer screening study. Directness: This domain was rated direct for all types of cancer screening because studies closely matched the PICOTS elements defined by the key questions. Most studies were RCTs that directly compared patient navigation with usual care and reported screening rates as the primary outcome measure. Precision: The precision domain was rated precise for colorectal, breast, and cervical cancer studies. Meta-analyses of RCTs of colorectal cancer screening (RR 1.64; 95% CI 1.42-1.92; 22 trials) and breast cancer screening (RR 1.50; 95% CI, 1.22-1.91; 10 trials) indicated statistically significant effects with narrow confidence intervals. Although studies of cervical cancer screening could not be combined because of statistical heterogeneity, results were statistically significant and effect sizes were clinically relevant. The domain for lung cancer was rated imprecise because of the uncertainty of the results of the single screening study in the context of its high loss to follow-up. Reporting bias: As reporting bias is difficult to assess, the investigators did not detect small study effects for the meta-analyses and did not suspect selective outcome report ing because the main outcome, the screening rate, was prespecified for each study. This domain was rated undetected for all types of cancer screening.

Using the definitions in Table 12.3, the strength of evidence for each type of can cer screening was assigned an overall grade of high, moderate, low, or insufficient by evaluating and weighing the combined results of the above domains, and by applying the systematic reviewers’ comprehensive understanding of the studies. The strength of evidence was graded high for colorectal cancer screening; medium for breast and cer vical cancer screening; and low for lung cancer screening. The large numbers of trials and participants and the extensive sensitivity analysis of the meta-analysis placed the colorectal cancer screening evidence at a higher level than the other types. Results of the single poor-quality RCT for lung cancer screening showed higher screening rates for the intervention versus usual care group, demonstrating an effect consistent with the other types of screening. Although a low grade was assigned, an insufficient grade could also be justified. Copyright © 2024 Wolters Kluwer, Inc. Unauthorized reproduction of the content is prohibited.