Clinical utility estimation for assay cutoffs in early phase oncology enrichment trials

A Bayesian model to estimate the cutoff value of TSH for management of preterm birth

BACKGROUND

Determining a thyroid hormone cutoff value in pregnancy is challenging issue and several approaches have been introduced to optimize a utility function. We aimed to estimate the cutoff value of TSH using Bayesian method for prediction of preterm-birth.

METHODS

This study was a secondary-analysis of the population-based data collected prospectively within the framework of the Tehran Thyroid and Pregnancy Study. A total of 1,538 pregnant women attending prenatal clinics.

RESULTS

Using Bayesian method resulted a TSH-cutoff of (3.97mIU/L,95%CI:3.95-4.00) for distinguishing pregnant women at risk of preterm-birth. The cutoff was associated with acceptable positive predictive and negative predictive values (0.84,95% CI:0.80-0.88) and 0.92 (95%CI: 0.91-0.94), respectively). In women who were negative for thyroid peroxides antibody (TPOAb) with sufficient urinary iodine concentration (UIC), the TSH cutoff of 3.92 mIU/L(95%CI:3.70-4) had the highest predictive value; whereas in TPOAb positive women with insufficient UIC, the cutoff of 4.0 mIU/L(95%:CI 3.94-4) could better predict preterm birth. Cutoffs estimated in this study are close to the revised TSH value of 4.0mIU/L which is currently recommended by the American Thyroid Association.

CONCLUSION

Regardless of TPOAb status or iodine insufficiency, risk of preterm labor is increased in pregnant women with TSH value of > 3.92 mIU/L; these women may benefit from Levothyroxine (LT4) therapy for preventing preterm birth.

A Bayesian model to estimate the cutoff and the clinical utility of a biomarker assay

To enable targeted therapies and enhance medical decision-making, biomarkers are increasingly used as screening and diagnostic tests. When using quantitative biomarkers for classification purposes, this often implies that an appropriate cutoff for the biomarker has to be determined and its clinical utility must be assessed. In the context of drug development, it is of interest how the probability of response changes with increasing values of the biomarker. Unlike sensitivity and specificity, predictive values are functions of the accuracy of the test, depend on the prevalence of the disease and therefore are a useful tool in this setting. In this paper, we propose a Bayesian method to not only estimate the cutoff value using the negative and positive predictive values, but also estimate the uncertainty around this estimate. Using Bayesian inference allows us to incorporate prior information, and obtain posterior estimates and credible intervals for the cut-off and associated predictive values. The performance of the Bayesian approach is compared with alternative methods via simulation studies of bias, interval coverage and width and illustrations on real data with binary and time-to-event outcomes are provided.