Inference for reclassification statistics under nested and non-nested models for biomarker evaluation

Joint inference about the AUC and Youden index for paired biomarkers

It is common to compare biomarkers' diagnostic or prognostic performance using some summary ROC measures such as the area under the ROC curve (AUC) or the Youden index. We propose to compare two paired biomarkers using both the AUC and the Youden index since the two indices describe different aspects of the ROC curve. This comparison can be made by estimating the joint confidence region (an elliptical area) of the differences of the paired AUCs and the Youden indices. Furthermore, for deciding if one marker is better than the other in terms of both the A U C and the Youden index (J), we can test H 0 : A U C a ≤ A U C b or J a ≤ J b against H a : A U C a > A U C b and J a > J b using the paired differences. The construction of such a joint hypothesis is an example of the multivariate order-restricted hypotheses. For such a hypothesis, we propose and compare three testing procedures: (1) the intersection-union test ( I U T ); (2) the conditional test; and (3) the joint test. The performance of the proposed inference methods was evaluated and compared through simulations. The simulation results demonstrate that the proposed joint confidence region maintains the desired confidence level, and all three tests maintain the type I error under the null. Furthermore, among the three proposed testing methods, the conditional test is the preferred approach with markedly larger power consistently than the other two competing methods.

The role of pulse pressure in navigating the paradigm of chronic kidney disease progression in type 2 diabetes mellitus

BACKGROUND AND AIMS

Arterial stiffness is a risk factor for chronic kidney disease progression (CKD). Pulse pressure is a surrogate marker of arterial stiffness. It is unclear if pulse pressure predicts CKD progression in type 2 diabetes mellitus.

METHODS

This was prospective study involving 1494 patients with estimated glomerular filtration rate (eGFR) ≥ 15 ml/min/1.73 m². Carotid-femoral pulse wave velocity was measured using applanation tonometry. Pulse pressure was calculated as difference between systolic and diastolic blood pressures. CKD progression was defined as worsening of eGFR categories (stage 1, ≥ 90 ml/min/1.73 m²; stage 2, 60-89 ml/min/1.73 m²; stage 3a, 45-59 ml/min/1.73 m²; stage 3b, 30-44 ml/min/1.73 m²; stage 4; 15-29 ml/min/1.73 m²; and stage 5, < 15 ml/min/1.73 m²) with ≥ 25% decrease in eGFR from baseline.

RESULTS

After follow-up of up to 6 years, CKD progression occurred in 33.5% of subjects. Subjects in 2nd, 3rd and 4th quartiles of peripheral pulse pressure experienced higher risk of CKD progression with unadjusted hazard ratios (HRs) 1.55 [95% confidence interval (CI) 1.13-2.11; p = 0.006], 2.58 (1.93-3.45; p < 0.001) and 3.41 (2.58-4.52; p < 0.001). In the fully adjusted model, the association for 2nd, 3rd and 4th quartiles remained with HRs 1.40 (1.02-1.93; p = 0.038), 1.87 (1.37-2.56; p < 0.001) and 1.75 (1.25-2.44; p = 0.001) respectively. Similarly, 2nd, 3rd and 4th quartiles of aortic pulse pressure were associated with higher hazards of CKD progression with HRs 1.73 (1.25-2.40; p = 0.001), 1.65 (1.18-2.29; p = 0.003) and 1.81 (1.26-2.60; p = 0.001). Increasing urinary albumin-to-creatinine ratio accounted for 44.0% of the association between peripheral pulse pressure and CKD progression.

CONCLUSIONS

Individuals with high pulse pressure were more susceptible to deterioration of renal function. Pulse pressure could potentially be incorporated in clinical practice as an inexpensive and readily available biomarker of renal decline in type 2 diabetes mellitus. Graphic abstract.

Evidence for a Causal Role of the SH2B3-β2M Axis in Blood Pressure Regulation

Genetic variants at SH2B3 are associated with blood pressure and circulating β₂M (β-2 microglobulin), a well-characterized kidney filtration biomarker. We hypothesize that circulating β₂M is an independent risk predictor of hypertension and may causally contribute to its development. The study sample consisted of 7 065 Framingham Heart Study participants with measurements of plasma β₂M. Generalized estimating equations were used to test the association of β₂M with prevalent and new-onset hypertension. There were 2 145 (30%) cases of prevalent hypertension at baseline and 886 (21%) cases of incident hypertension during 6 years of follow-up. A 1-SD increase in baseline plasma β₂M was associated with a greater risk of prevalent (odds ratio 1.14, 95% CI 1.05-1.24) and new-onset (odds ratio 1.18, 95% CI 1.07-1.32) hypertension. Individuals within the top β₂M quartile had a greater risk than the bottom quartile for prevalent (odds ratio 1.29, 95% CI 1.05-1.57) and new-onset (odds ratio 1.59, 95% CI 1.20-2.11) hypertension. These associations remained essentially unchanged in analyses restricted to participants free of albuminuria and chronic kidney disease. Mendelian randomization demonstrated that lower SH2B3 expression is causal for increased circulating β₂M levels, and in a hypertensive mouse model, knockout of Sh2b3 increased β ₂ M gene expression. In a community-based study of healthy individuals, higher plasma β₂M levels are associated with increased risk of prevalent and incident hypertension independent of chronic kidney disease status. Overlapping genetic signals for hypertension and β₂M, in conjunction with mouse knockout experiments, suggest that the SH2B3-β₂M axis plays a causal role in hypertension.

Cardiac events within one year after a subarachnoid haemorrhage: The predictive value of troponin elevation after aneurysm occlusion

BACKGROUND

Patients who survive after an aneurysmal subarachnoid haemorrhage (ASAH) have an increased incidence of cardiovascular events compared with the general population. We assessed whether troponin elevation after aneurysm occlusion, as marker of myocardial injury, can predict long-term cardiac events.

METHODS

We analysed a prospectively collected cohort of 159 patients with ASAH and early aneurysm occlusion, in whom routine post-intervention troponin I (TnI) measurements were performed. With competing risk regression modelling we estimated the association between TnI elevation after aneurysm occlusion and major adverse cardiac events within one year. Secondary outcome measures were all-cause mortality and neurological condition within one year. The predictive value of post-intervention TnI was compared with the predictive value of pre-intervention characteristics using c-statistics and the integrated discrimination improvement index.

RESULTS

Subdistribution hazard ratios for TnI elevation and major adverse cardiac events at one year were 1.05 (95% confidence interval (CI) 1.03-1.07) per 10 ng/l increase in TnI and 7.91 (95% CI 1.46-43.0) for any TnI elevation. After adjustment for pre-intervention variables, the subdistribution hazard ratios were 1.47 (95% CI 0.81-2.67) per 10 ng/l and 9.00 (95% CI 1.62-50.1) for any elevation. The c-statistic was 0.71 for TnI elevation as a continuous measure and 0.69 for any TnI elevation. The integrated discrimination improvement index showed a minimum improvement in prediction of 0.08 (interquartile range 0.06 to 0.09) for TnI as a continuous measure and 0.003 (interquartile range -0.004 to 0.01) for any TnI elevation, when compared with pre-intervention characteristics.

CONCLUSION

TnI elevation after occlusion of a ruptured intracranial aneurysm predicts the occurrence of a major adverse cardiac event within one year after ASAH.

Reclassification calibration test for censored survival data: performance and comparison to goodness-of-fit criteria

BACKGROUND

The risk reclassification table assesses clinical performance of a biomarker in terms of movements across relevant risk categories. The Reclassification-Calibration (RC) statistic has been developed for binary outcomes, but its performance for survival data with moderate to high censoring rates has not been evaluated.

METHODS

We develop an RC statistic for survival data with higher censoring rates using the Greenwood-Nam-D'Agostino approach (RC-GND). We examine its performance characteristics and compare its performance and utility to the Hosmer-Lemeshow goodness-of-fit test under various assumptions about the censoring rate and the shape of the baseline hazard.

RESULTS

The RC-GND test was robust to high (up to 50%) censoring rates and did not exceed the targeted 5% Type I error in a variety of simulated scenarios. It achieved 80% power to detect better calibration with respect to clinical categories when an important predictor with a hazard ratio of at least 1.7 to 2.2 was added to the model, while the Hosmer-Lemeshow goodness of fit (gof) test had power of 5% in this scenario.

CONCLUSIONS

The RC-GND test should be used to test the improvement in calibration with respect to clinically-relevant risk strata. When an important predictor is omitted, the Hosmer-Lemeshow goodness-of-fit test is usually not significant, while the RC-GND test is sensitive to such an omission.