How many longitudinal covariate measurements are needed for risk prediction?

The power-integrated discriminant improvement: An accurate measure of the incremental predictive value of additional biomarkers

The predictive performance of biomarkers is a central concern in biomedical research. This is often evaluated by comparing two statistical models: a "new" model incorporating additional biomarkers and an "old" model without them. In 2008, the integrated discrimination improvement (IDI) was proposed for cases when the response variable is binary, and it is now widely applied as a promising alternative to conventional measures, such as the difference of the area under the receiver operating characteristic curve. However, the IDI can erroneously identify a significant improvement in the new model even if no additional information has been provided by new biomarkers. In order to overcome problems with existing measures, in this study, we propose the power-IDI as a measure of incremental predictive value. Our study explains why the IDI cannot avoid false detection of apparent improvements in a new model and we show that our proposed measure is better able to capture improvements in prediction. Numerical simulations and examples using real empirical data reveal that the power-IDI is not only more powerful but also incurs fewer false detections of improvement.

Longitudinal immunosuppression data can minimize misclassification bias in solid organ transplantation cohorts

BACKGROUND

Most cohort studies investigating the effect of immunosuppression on transplant outcomes use drugs at first hospital discharge. We evaluated the extent of drug exposure misclassification and its impact on outcome prediction.

METHODS

We retrospectively collected longitudinal immunosuppression data, at discharge and at 1, 5, 10, and 15 years after transplantation, and outcomes for solid organ transplant recipients 1984-2006 (n = 3133). We compared the risk of death from exposure to individual immunosuppressive drugs (cyclosporine, tacrolimus, azathioprine, and mycophenolate) and dual therapies, as defined by discharge only vs longitudinal immunosuppression data, using adjusted Cox proportional hazards models.

RESULTS

During a median follow-up of 5.2 years, immunosuppressive drugs were altered for 947 (30%) recipients and 955 recipients died. Longitudinal receipt of cyclosporine and azathioprine were associated with an increased risk (HR 1.41, 95% CI 1.07-1.89, and HR 1.34, 95% CI 1.00-1.80), and mycophenolate with a reduced risk (HR 0.35, 0.16-0.78), of death. Recipients on mycophenolate and tacrolimus dual therapy had a lower risk of death compared to those on azathioprine and cyclosporine dual therapy (HR 0.30, 0.10-0.93). The increased risk of death associated with the receipt of cyclosporine or azathioprine was not shown in the analyses based on drugs allocated at discharge, and all of the associations between immunosuppressive regimens and death were strengthened in the analyses based on longitudinal immunosuppression data.

CONCLUSIONS

Cohort findings based on immunosuppressive drugs allocated at discharge should be interpreted with caution due to potential exposure misclassification. The use of granular, longitudinal data on immunosuppressive regimens could improve prediction.