Matching of controls may lead to biased estimates of specificity in the evaluation of cancer screening tests
ABSTRACT In the evaluation of cancer screening tests, cancer-free controls are often matched to cancer cases on factors such as sex and age. We assessed the potential merits and pitfalls of such matching using an example from colorectal cancer (CRC) screening.
We compared sex and age distribution of CRC cases and cancer-free people undergoing screening colonoscopy in Germany in 2006 and 2007. We assessed specificity by sex and age of two immunochemical fecal occult blood tests (iFOBTs) in a study among screening colonoscopy participants conducted in the same years, and we assessed the expected impact of matching by sex and age on the validity of specificity estimates at various cut points.
In the screening colonoscopy program, the proportion of men and mean age were 59.6% and 68.6 years among 10,324 CRC patients compared with 45.6% and 64.7 years, respectively, among 997,490 cancer-free participants. The specificity of the iFOBTs was higher among women than among men and decreased with age. Matching of cancer-free controls by age and sex would have led to the underestimation of specificity at all cut points assessed.
In the evaluation of cancer screening tests, matching of controls may lead to biased estimates of specificity.
- SourceAvailable from: informahealthcare.com[Show abstract] [Hide abstract]
ABSTRACT: Background: Faecal immunochemical tests (FITs) for haemoglobin are increasingly used for non-invasive screening for colorectal cancer (CRC) but large scale comparative studies of different FITs for detection of CRC, overall and by stage, are sparse. We aimed to determine and compare performance of different FITs for the detection of CRC, and to assess their stage-specific sensitivities. Material and methods: We assessed sensitivity, specificity and their corresponding 95% confidence intervals for six qualitative FITs among 74 CRC cases (59% stage I or II cancers) and 1480 controls free of colorectal neoplasm. Overall and stage-specific receiver operating characteristic curves were derived for three quantitative FITs. The areas under the curves (AUCs) were calculated and compared. Results: Pairs of overall sensitivity and specificity of the qualitative FITs ranged from 66% and 96% to 92% and 62%, respectively. For the three quantitative tests, AUCs ranged from 0.90 to 0.92, with sensitivities ranging from 80% to 87% at cut-offs yielding 90% specificity. AUCs ranged from 0.85 to 0.92, 0.94 to 0.96, and 0.86 to 0.93 for stage I, stage II and advanced stages (stage III and IV) cancers, respectively. At a specificity of 90%, the tests detected 65%-85% of stage I cancers. Conclusion: The diagnostic performance of FITs regarding detection of CRC is promising, even though the pre-defined cut-offs of some of the qualitative FITs need to be adjusted to limit false-positive rates in screening setting. At cut-off levels yielding 90% specificity, the quantitative tests detected the vast majority of CRCs, even at early stages.Acta oncologica (Stockholm, Sweden) 04/2013; 52(8). DOI:10.3109/0284186X.2013.789141 · 3.00 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Novel noninvasive blood-based screening tests are strongly desirable for early detection of colorectal cancer. We aimed to conduct a head-to-head comparison of the diagnostic performance of 92 plasma-based tumor-associated protein biomarkers for early detection of colorectal cancer in a true screening setting. Among all available 35 carriers of colorectal cancer and a representative sample of 54 men and women free of colorectal neoplasms recruited in a cohort of screening colonoscopy participants in 2005-2012 (N = 5,516), the plasma levels of 92 protein biomarkers were measured. ROC analyses were conducted to evaluate the diagnostic performance. A multimarker algorithm was developed through the Lasso logistic regression model and validated in an independent validation set. The .632+ bootstrap method was used to adjust for the potential overestimation of diagnostic performance. Seventeen protein markers were identified to show statistically significant differences in plasma levels between colorectal cancer cases and controls. The adjusted area under the ROC curves (AUC) of these 17 individual markers ranged from 0.55 to 0.70. An eight-marker classifier was constructed that increased the adjusted AUC to 0.77 [95% confidence interval (CI), 0.59-0.91]. When validating this algorithm in an independent validation set, the AUC was 0.76 (95% CI, 0.65-0.85), and sensitivities at cutoff levels yielding 80% and 90% specificities were 65% (95% CI, 41-80%) and 44% (95% CI, 24-72%), respectively. The identified profile of protein biomarkers could contribute to the development of a powerful multimarker blood-based test for early detection of colorectal cancer. Clin Cancer Res; 1-9. ©2015 AACR. ©2015 American Association for Cancer Research.Clinical Cancer Research 05/2015; DOI:10.1158/1078-0432.CCR-14-3051 · 8.72 Impact Factor