Clinical trial designs to assess treatment effects on glomerular filtration rate decline.

Glomerular filtration rate (GFR) decline is used as surrogate endpoint for kidney failure. Interventions that reduce chronic kidney disease (CKD) progression often exert acute GFR reductions which differ from their long-term benefits and complicate the estimation of long-term benefit. Here, we assessed the utility of two alternative trial designs (wash-out design and active run-in randomized withdrawal design) that attempt to exclude the impact of acute effects. Post-hoc analyses of two clinical trials that characterized the effect of an intervention with acute reductions in GFR were conducted. The two trials included a wash-out period (EMPA-REG Outcome testing empagliflozin vs placebo) or an active run-in period with a randomized withdrawal (SONAR testing atrasentan vs placebo). We compared the drug effect on GFR decline calculated from the first on-treatment visit to the end of treatment (chronic slope in a standard randomized trial design) with GFR change calculated from randomization to end of wash out, or GFR change from treatment-specific baseline GFR values (GFR at start-of-run-in for placebo and end-of-run-in for atrasentan) until end-of-treatment. The effect of empagliflozin versus placebo on chronic GFR slope was 1.72 (95% confidence interval 1.49-1.94) mL/min/1.73 m2/year, similar to total GFR decline from baseline to the end of wash-out period using a linear mixed model 1.64 (1.44-1.85) mL/min/1.73 m2/year). The effect of atrasentan versus placebo on chronic GFR slope was 0.72 (0.32-1.11) mL/min/1.73 m2/year, similar to total slope from a single slope model when estimated from treatment specific baseline GFR values 0.77 (0.39-1.14) mL/min/1.73 m2/year). Statistical power of the two designs outperformed the standard randomized design. Thus, wash-out and active-run-in randomized-withdrawal trial designs are appropriate models to compute treatment effects on GFR decline.

Multidimensional Thresholding for Individual-Level Preference Elicitation.

OBJECTIVES: Multiple methods are available for collecting health preference information. However, information on the design and analysis of novel methods is limited. This article aims to provide the first introduction into the design and analysis of multidimensional thresholding (MDT). METHODS: We introduce MDT as a 2-step approach: First, participants rank the largest possible improvements in all considered attributes by their importance. Second, participants complete a series of systematically combined trade-off questions. Hit-and-Run sampling is used for obtaining preference weights. We also use a computational experiment to compare different MDT designs. RESULTS: The outlined MDT can generate preference information suitable for specifying a multiattribute utility function at the individual level. The computational experiment demonstrates the method's ability to recover preference weights at a high level of precision. While all designs in the computation experiment perform comparably well on average, the design outlined in the paper stands out with a high level of precision even if differences in relative attribute importance are large. CONCLUSION: MDT is suitable for preference elicitation, in particular if sample sizes are small. Future research should help improve the methods (e.g., remove the need for an initial ranking) to increase the potential reach of MDT.

Effect of Dapagliflozin on Clinical Outcomes in Patients With Chronic Kidney Disease, With and Without Cardiovascular Disease.

BACKGROUND: Dapagliflozin reduces the risk of end-stage renal disease in patients with chronic kidney disease. We examined the relative risk of cardiovascular and renal events in these patients and the effect of dapagliflozin on either type of event, taking account of history of cardiovascular disease. METHODS: In the DAPA-CKD trial (Dapagliflozin and Prevention of Adverse Outcomes in Chronic Kidney Disease), 4304 participants with chronic kidney disease were randomly assigned to dapagliflozin 10 mg once daily or placebo. The primary end point was a composite of sustained decline in estimated glomerular filtration rate ≥50%, end-stage kidney disease, or kidney or cardiovascular death. The secondary end points were a kidney composite outcome (primary end point, minus cardiovascular death), the composite of hospitalization for heart failure or cardiovascular death, and all-cause death. In a prespecified subgroup analysis, we divided patients into primary and secondary prevention subgroups according to history of cardiovascular disease. RESULTS: Secondary prevention patients (n=1610; 37.4%) were older, were more often male, had a higher blood pressure and body mass index, and were more likely to have diabetes. Mean estimated glomerular filtration rate and median urinary albumin-to-creatinine ratio were similar in the primary and secondary prevention groups. The rates of adverse cardiovascular outcomes were higher in the secondary prevention group, but kidney failure occurred at the same rate in the primary and secondary prevention groups. Dapagliflozin reduced the risk of the primary composite outcome to a similar extent in both the primary (hazard ratio, 0.61 [95% CI, 0.48-0.78]) and secondary (0.61 [0.47-0.79]) prevention groups (P-interaction=0.90). This was also true for the composite of heart failure hospitalization or cardiovascular death (0.67 [0.40-1.13] versus 0.70 [0.52-0.94], respectively; P-interaction=0.88), and all-cause mortality (0.63 [0.41-0.98] versus 0.70 [0.51-0.95], respectively; P-interaction=0.71). Rates of adverse events were low overall and did not differ between patients with and without cardiovascular disease. CONCLUSIONS: Dapagliflozin reduced the risk of kidney failure, death from cardiovascular causes or hospitalization for heart failure, and prolonged survival in people with chronic kidney disease, with or without type 2 diabetes, independently of the presence of concomitant cardiovascular disease. Registration: URL: https://www.clinicaltrials.gov; Unique identifier: NCT03036150.

A pre-specified analysis of the Dapagliflozin and Prevention of Adverse Outcomes in Chronic Kidney Disease (DAPA-CKD) randomized controlled trial on the incidence of abrupt declines in kidney function.

This pre-specified analysis of DAPA-CKD assessed the impact of sodium-glucose cotransporter 2 inhibition on abrupt declines in kidney function in high-risk patients based on having chronic kidney disease (CKD) and substantial albuminuria. DAPA-CKD was a randomized, double-blind, placebo-controlled trial that had a median follow-up of 2.4 years. Adults with CKD (urinary albumin-to-creatinine ratio 200-5000 mg/g and estimated glomerular filtration rate 25-75 mL/min/1.73m2) were randomized to dapagliflozin 10 mg/day matched to placebo (2152 individuals each). An abrupt decline in kidney function was defined as a pre-specified endpoint of doubling of serum creatinine between two subsequent study visits. We also assessed a post-hoc analysis of investigator-reported acute kidney injury-related serious adverse events. Doubling of serum creatinine between two subsequent visits (median time-interval 100 days) occurred in 63 (2.9%) and 91 (4.2%) participants in the dapagliflozin and placebo groups, respectively (hazard ratio 0.68 [95% confidence interval 0.49, 0.94]). Accounting for the competing risk of mortality did not alter our findings. There was no heterogeneity in the effect of dapagliflozin on abrupt declines in kidney function based on baseline subgroups. Acute kidney injury-related serious adverse events were not significantly different and occurred in 52 (2.5%) and 69 (3.2%) participants in the dapagliflozin and placebo groups, respectively (0.77 [0.54, 1.10]). Thus, in patients with CKD and substantial albuminuria, dapagliflozin reduced the risk of abrupt declines in kidney function.

A simulated maximum likelihood procedure for analyzing imprecise trade-off thresholds between the benefits and harms of medicines.

Stated preference studies in which information on the willingness to trade-off between the benefits and harms of medicines is elicited from patients or other stakeholders are becoming increasingly mainstream. Such trade-offs can mathematically be represented by a weighted additive function, with the weights, whose ratios determine how much an individual is willing to trade-off between the treatment attributes, being the response vector for the statistical analysis. One way of eliciting trade-off information is through multi-dimensional thresholding (MDT), which is a bisection-based approach that results in increasingly tight bounds on the values of the weights ratios. While MDT is cognitively less demanding than other, more direct elicitation methods, its use complicates the statistical analysis as it results in weights data that are region censored. In this article, we present a simulated maximum likelihood (SML) procedure for fitting a Dirichlet population model directly to the region-censored weights data and perform a series of computational experiments to compare the proposed SML procedure to a naive approach in which a Dirichlet distribution is fitted to the centroids of the weights boundaries obtained with MDT. The results indicate that the SML procedure consistently outperformed the centroid-based approach, with the centroid-based approach requiring three bisection steps per trade-off to achieve a similar precision as the SML procedure with one bisection step per trade-off. Using the newly proposed SML procedure, MDT can be applied with smaller sample sizes or with fewer questions compared to the more naïve centroid-based approach that was applied in previous applications of MDT.

A heart failure phenotype stratified model for predicting 1-year mortality in patients admitted with acute heart failure: results from an individual participant data meta-analysis of four prospective European cohorts.

BACKGROUND: Prognostic models developed in general cohorts with a mixture of heart failure (HF) phenotypes, though more widely applicable, are also likely to yield larger prediction errors in settings where the HF phenotypes have substantially different baseline mortality rates or different predictor-outcome associations. This study sought to use individual participant data meta-analysis to develop an HF phenotype stratified model for predicting 1-year mortality in patients admitted with acute HF. METHODS: Four prospective European cohorts were used to develop an HF phenotype stratified model. Cox model with two rounds of backward elimination was used to derive the prognostic index. Weibull model was used to obtain the baseline hazard functions. The internal-external cross-validation (IECV) approach was used to evaluate the generalizability of the developed model in terms of discrimination and calibration. RESULTS: 3577 acute HF patients were included, of which 2368 were classified as having HF with reduced ejection fraction (EF) (HFrEF; EF < 40%), 588 as having HF with midrange EF (HFmrEF; EF 40-49%), and 621 as having HF with preserved EF (HFpEF; EF ≥ 50%). A total of 11 readily available variables built up the prognostic index. For four of these predictor variables, namely systolic blood pressure, serum creatinine, myocardial infarction, and diabetes, the effect differed across the three HF phenotypes. With a weighted IECV-adjusted AUC of 0.79 (0.74-0.83) for HFrEF, 0.74 (0.70-0.79) for HFmrEF, and 0.74 (0.71-0.77) for HFpEF, the model showed excellent discrimination. Moreover, there was a good agreement between the average observed and predicted 1-year mortality risks, especially after recalibration of the baseline mortality risks. CONCLUSIONS: Our HF phenotype stratified model showed excellent generalizability across four European cohorts and may provide a useful tool in HF phenotype-specific clinical decision-making.

Derivation and validation of a prognostic model for predicting in-hospital mortality in patients admitted with COVID-19 in Wuhan, China: the PLANS (platelet lymphocyte age neutrophil sex) model.

BACKGROUND: Previous published prognostic models for COVID-19 patients have been suggested to be prone to bias due to unrepresentativeness of patient population, lack of external validation, inappropriate statistical analyses, or poor reporting. A high-quality and easy-to-use prognostic model to predict in-hospital mortality for COVID-19 patients could support physicians to make better clinical decisions. METHODS: Fine-Gray models were used to derive a prognostic model to predict in-hospital mortality (treating discharged alive from hospital as the competing event) in COVID-19 patients using two retrospective cohorts (n = 1008) in Wuhan, China from January 1 to February 10, 2020. The proposed model was internally evaluated by bootstrap approach and externally evaluated in an external cohort (n = 1031). RESULTS: The derivation cohort was a case-mix of mild-to-severe hospitalized COVID-19 patients (43.6% females, median age 55). The final model (PLANS), including five predictor variables of platelet count, lymphocyte count, age, neutrophil count, and sex, had an excellent predictive performance (optimism-adjusted C-index: 0.85, 95% CI: 0.83 to 0.87; averaged calibration slope: 0.95, 95% CI: 0.82 to 1.08). Internal validation showed little overfitting. External validation using an independent cohort (47.8% female, median age 63) demonstrated excellent predictive performance (C-index: 0.87, 95% CI: 0.85 to 0.89; calibration slope: 1.02, 95% CI: 0.92 to 1.12). The averaged predicted cumulative incidence curves were close to the observed cumulative incidence curves in patients with different risk profiles. CONCLUSIONS: The PLANS model based on five routinely collected predictors would assist clinicians in better triaging patients and allocating healthcare resources to reduce COVID-19 fatality.

Urinary sulfate excretion and risk of late graft failure in renal transplant recipients - a prospective cohort study.

Hydrogen sulfide (H2 S), produced from metabolism of dietary sulfur-containing amino acids, is allegedly a renoprotective compound. Twenty-four-hour urinary sulfate excretion (USE) may reflect H2 S bioavailability. We aimed to investigate the association of USE with graft failure in a large prospective cohort of renal transplant recipients (RTR). We included 704 stable RTR, recruited at least 1 year after transplantation. We applied log-rank testing and Cox regression analyses to study association of USE, measured from baseline 24 h urine samples, with graft failure. Median age was 55 [45-63] years (57% male, eGFR was 45 ± 19 ml/min/1.73 m2 ). Median USE was 17.1 [13.1-21.1] mmol/24 h. Over median follow-up of 5.3 [4.5-6.0] years, 84 RTR experienced graft failure. RTR in the lowest sex-specific tertile of USE experienced a higher rate of graft failure during follow-up than RTR in the middle and highest sex-specific tertiles (18%, 13%, and 5%, respectively, log-rank P < 0.001). In Cox regression analyses, USE was inversely associated with graft failure [HR per 10 mmol/24 h: 0.37 (0.24-0.55), P < 0.001]. The association remained independent of adjustment for potential confounders, including age, sex, eGFR, proteinuria, time between transplantation and baseline, BMI, smoking, and high sensitivity C-reactive protein [HR per 10 mmol/24 h: 0.51 (0.31-0.82), P = 0.01]. In conclusion, this study demonstrates a significant inverse association of USE with graft failure in RTR, suggesting high H2 S bioavailability as a novel, potentially modifiable factor for prevention of graft failure in RTR.

Understanding Patients' Preferences: A Systematic Review of Psychological Instruments Used in Patients' Preference and Decision Studies.

BACKGROUND: Research has been mainly focused on how to elicit patient preferences, with less attention on why patients form certain preferences. OBJECTIVES: To assess which psychological instruments are currently used and which psychological constructs are known to have an impact on patients' preferences and health-related decisions including the formation of preferences and preference heterogeneity. METHODS: A systematic database search was undertaken to identify relevant studies. From the selected studies, the following information was extracted: study objectives, study population, design, psychological dimensions investigated, and instruments used to measure psychological variables. RESULTS: Thirty-three studies were identified that described the association between a psychological construct, measured using a validated instrument, and patients' preferences or health-related decisions. We identified 33 psychological instruments and 18 constructs, and categorized the instruments into 5 groups, namely, motivational factors, cognitive factors, individual differences, emotion and mood, and health beliefs. CONCLUSIONS: This review provides an overview of the psychological factors and related instruments in the context of patients' preferences and decisions in healthcare settings. Our results indicate that measures of health literacy, numeracy, and locus of control have an impact on health-related preferences and decisions. Within the category of constructs that could explain preference and decision heterogeneity, health locus of control is a strong predictor of decisions in several healthcare contexts and is useful to consider when designing a patient preference study. Future research should continue to explore the association of psychological constructs with preference formation and heterogeneity to build on these initial recommendations.

Individual Trade-Offs Between Possible Benefits and Risks of Cancer Treatments: Results from a Stated Preference Study with Patients with Multiple Myeloma.

BACKGROUND: The objectives of this study were to elicit the preferences of patients with multiple myeloma regarding the possible benefits and risks of cancer treatments and to illustrate how such data may be used to estimate patients' acceptance of new treatments. PATIENTS AND METHODS: Patients with multiple myeloma from the cancer charity Myeloma UK were invited to participate in an online survey based on multicriteria decision analysis and swing weighting to elicit individual stated preferences for the following attributes: (a) 1-year progression-free survival (PFS, ranging from 50% to 90%), (b) mild or moderate toxicity for 2 months or longer (ranging from 85% to 45%), and (c) severe or life-threatening toxicity (ranging from 80% to 20%). RESULTS: A total of 560 participants completed the survey. The average weight given to PFS was 0.54, followed by 0.32 for severe or life-threatening toxicity and 0.14 for mild or moderate chronic toxicity. Participants who ranked severe or life-threatening toxicity above mild or moderate chronic toxicity (56%) were more frequently younger, working, and looking after dependent family members and had more frequently experienced severe or life-threatening side effects. The amount of weight given to PFS did not depend on any of the collected covariates. The feasibility of using the collected preference data to estimate the patients' acceptance of specific multiple myeloma treatments was demonstrated in a subsequent decision analysis example. CONCLUSION: Stated preference studies provide a systematic approach to gain knowledge about the distribution of preferences in the population and about what this implies for patients' acceptance of specific treatments. IMPLICATIONS FOR PRACTICE: This study demonstrated how quantitative preference statements from a large group of participants can be collected through an online survey and how such information may be used to explore the acceptability of specific treatments based on the attributes studied. Results from such studies have the potential to become an important new tool for gathering patient views and studying heterogeneity in preferences in a systematic way, along with other methods, such as focus groups and expert opinions.

The early economic evaluation of novel biomarkers to accelerate their translation into clinical applications.

BACKGROUND: Translating prognostic and diagnostic biomarker candidates into clinical applications takes time, is very costly, and many candidates fail. It is therefore crucial to be able to select those biomarker candidates that have the highest chance of successfully being adopted in the clinic. This requires an early estimate of the potential clinical impact and commercial value. In this paper, we aim to demonstratively evaluate a set of novel biomarkers in terms of clinical impact and commercial value, using occurrence of cardiovascular disease (CVD) in type-2 diabetes (DM2) patients as a case study. METHODS: We defined a clinical application for the novel biomarkers, and subsequently used data from a large cohort study in The Netherlands in a modeling exercise to assess the potential clinical impact and headroom for the biomarkers. RESULTS: The most likely application of the biomarkers would be to identify DM2 patients with a low CVD risk and subsequently withhold statin treatment. As a result, one additional CVD event in every 75 patients may be expected. The expected downstream savings resulted in a headroom for a point-of-care device ranging from €119.09 at a willingness to accept of €0 for one additional CVD event, to €0 at a willingness to accept of €15,614 or more. CONCLUSION: It is feasible to evaluate novel biomarkers on outcomes directly relevant to technological development and clinical adoption. Importantly, this may be attained at the same point in time and using the same data as used for the evaluation of association with disease and predictive power.

Identifying Subpopulations with Distinct Response to Treatment Using Plasma Biomarkers in Acute Heart Failure: Results from the PROTECT Trial : Differential Response in Acute Heart Failure.

BACKGROUND: Over the last 50 years, clinical trials of novel interventions for acute heart failure (AHF) have, with few exceptions, been neutral or shown harm. We hypothesize that this might be related to a differential response to pharmacological therapy. METHODS: We studied the magnitude of treatment effect of rolofylline across clinical characteristics and plasma biomarkers in 2033 AHF patients and derived a biomarker-based responder sum score model. Treatment response was survival from all-cause mortality through day 180. RESULTS: In the overall study population, rolofylline had no effect on mortality (HR 1.03, 95% CI 0.82-1.28, p = 0.808). We found no treatment interaction across clinical characteristics, but we found interactions between several biomarkers and rolofylline. The biomarker-based sum score model included TNF-R1α, ST2, WAP four-disulfide core domain protein HE4 (WAP-4C), and total cholesterol, and the score ranged between 0 and 4. In patients with score 4 (those with increased TNF-R1α, ST2, WAP-4C, and low total cholesterol), treatment with rolofylline was beneficial (HR 0.61, 95% CI 0.40-0.92, p = 0.019). In patients with score 0, treatment with rolofylline was harmful (HR 5.52, 95% CI 1.68-18.13, p = 0.005; treatment by score interaction p < 0.001). Internal validation estimated similar hazard ratio estimates (0 points: HR 5.56, 95% CI 5.27-7-5.87; 1 point: HR 1.31, 95% CI 1.25-1.33; 2 points: HR 0.75, 95% CI 0.74-0.76; 3 points: HR 1.13, 95% CI 1.11-1.15; 4 points, HR 0.61, 95% CI 0.61-0.62) compared to the original data. CONCLUSION: Biomarkers are superior to clinical characteristics to study treatment heterogeneity in acute heart failure.

THE EARLY BIRD CATCHES THE WORM: EARLY COST-EFFECTIVENESS ANALYSIS OF NEW MEDICAL TESTS.

OBJECTIVES: There is little specific guidance on performing an early cost-effectiveness analysis (CEA) of medical tests. We developed a framework with general steps and applied it to two cases. METHODS: Step 1 is to narrow down the scope of analysis by defining the test's application, target population, outcome measures, and investigating current test strategies and test strategies if the new test were available. Step 2 is to collect evidence on the current test strategy. Step 3 is to develop a conceptual model of the current and new test strategies. Step 4 is to conduct the early-CEA by evaluating the potential (cost-)effectiveness of the new test in clinical practice. Step 5 involves a decision about the further development of the test. RESULTS: The first case illustrated the impact of varying the test performance on the headroom (maximum possible price) of an add-on test for patients with an intermediate-risk of having rheumatoid arthritis. Analyses showed that the headroom is particularly dependent on test performance. The second case estimated the minimum performance of a confirmatory imaging test to predict individual stroke risk. Different combinations of sensitivity and specificity were found to be cost-effective; if these combinations are attainable, the medical test developer can feel more confident about the value of further development of the test. CONCLUSIONS: A well-designed early-CEA methodology can improve the ability to develop (cost-)effective medical tests in an efficient manner. Early-CEAs should continuously integrate insights and evidence that arise through feedback, which may convince developers to return to earlier steps.

Hemoglobin levels and new-onset heart failure in the community.

BACKGROUND: In established cardiovascular disease and heart failure (HF), low hemoglobin levels are associated with unfavorable outcome. Whether hemoglobin levels are associated with the development of new-onset HF in the population is unclear. This study sought to investigate the relationship between hemoglobin levels and development of new-onset HF in the community. METHODS: In 6,744 patients from PREVEND, a prospective, community-based, cohort study, we analyzed the relationship between hemoglobin levels and the risk of new-onset HF. RESULTS: Mean age (±SD) was 53 ± 12 years, 49.8% was male, and mean hemoglobin level was 13.7 ± 1.2 g/dL. During a median follow-up of 8.3 years (interquartile range 7.8-8.9), 217 subjects (3.2%) were newly diagnosed with HF. The association between hemoglobin levels and the risk for new-onset HF was U shaped (P< .001), remaining significant after full adjustment in a multivariable model with established cardiovascular risk factors (P= .015). Furthermore, a increased annual HF incidence was already observed in subjects with high-normal hemoglobin levels (men >16 g/dL or women >15 g/dL; P= .041), whereas on the other side of the distribution, only severe anemia (men <11 g/dL or women <10 g/dL; P= .018) was associated with a higher annual incidence. CONCLUSIONS: The impact of hemoglobin level on the risk of new-onset HF in the community is best described as U shaped. Interestingly, higher hemoglobin levels, already within the high-reference range, are associated with an increased incidence. This in contrast to anemia, where a higher annual HF incidence was only observed for severe anemia.

Should non-cardiovascular mortality be considered in the SCORE model? Findings from the Prevention of Renal and Vascular End-stage Disease (PREVEND) cohort.

Competing non-cardiovascular related deaths were not accounted for in the Systematic COronary Risk Evaluation (SCORE) model. In this study we assessed the impact of non-cardiovascular related deaths on the prognostic performance and yield of the SCORE model. 5,752 participants from the Prevention of Renal and Vascular End stage Disease cohort aged 40 years and older who were free of atherosclerotic cardiovascular disease (CVD) at baseline were included. A cause-specific hazards (CSH) CVD-related mortality prediction model that accounted for non-CVD-related deaths was developed. The prognostic performance of this model was then compared with a refitted SCORE model. During a median follow-up period of 12.5 years, 139 CVD and 495 non-CVD-related deaths were reported. Discriminatory performance was comparable between the models (C-index = 0.64). The models showed good calibration although the CSH model underestimated risk in the highest decile while the refitted SCORE model showed overestimation. The CSH model classified more non-events into the low risk group compared to the refitted SCORE model (n = 51), yet it was accompanied by a misclassification of six events into the low risk group. The refitted SCORE model classified more individuals as high risk. However, the potential overtreatment that may result from utilizing the refitted SCORE model, when compared with the CSH model, still falls within acceptable limits. Our findings do not support the incorporation of non-cardiovascular mortality into the estimation of total cardiovascular risk in the SCORE model.

Graphical assessment of incremental value of novel markers in prediction models: From statistical to decision analytical perspectives.

New markers may improve prediction of diagnostic and prognostic outcomes. We aimed to review options for graphical display and summary measures to assess the predictive value of markers over standard, readily available predictors. We illustrated various approaches using previously published data on 3264 participants from the Framingham Heart Study, where 183 developed coronary heart disease (10-year risk 5.6%). We considered performance measures for the incremental value of adding HDL cholesterol to a prediction model. An initial assessment may consider statistical significance (HR = 0.65, 95% confidence interval 0.53 to 0.80; likelihood ratio p < 0.001), and distributions of predicted risks (densities or box plots) with various summary measures. A range of decision thresholds is considered in predictiveness and receiver operating characteristic curves, where the area under the curve (AUC) increased from 0.762 to 0.774 by adding HDL. We can furthermore focus on reclassification of participants with and without an event in a reclassification graph, with the continuous net reclassification improvement (NRI) as a summary measure. When we focus on one particular decision threshold, the changes in sensitivity and specificity are central. We propose a net reclassification risk graph, which allows us to focus on the number of reclassified persons and their event rates. Summary measures include the binary AUC, the two-category NRI, and decision analytic variants such as the net benefit (NB). Various graphs and summary measures can be used to assess the incremental predictive value of a marker. Important insights for impact on decision making are provided by a simple graph for the net reclassification risk.

Attitudes of healthcare professionals and drug regulators about progression-free survival as endpoint in the advanced cancer setting.

PURPOSE: To describe the attitudes of healthcare professionals and drug regulators about progression-free survival (PFS) as efficacy endpoint in clinical trials with patients with advanced cancer and to explore to what extent these attitudes influence the willingness to trade between PFS and toxicity. METHODS: Cross-sectional survey with regulators from the European Medicines Agency (EMA), and healthcare professionals (HCP) from the "Stichting Hemato-Oncologie voor Volwassenen Nederland" (HOVON) collaborative group and the European Organisation for Research and Treatment of Cancer (EORTC). Attitudes towards PFS were elicited using 5-point Likert items. The respondents' willingness to trade between PFS and grade 3 or 4 (G34) toxicity was assessed using the threshold technique and quantified in terms of their maximum acceptable risk (MAR). RESULTS: Responses were collected from 287 HCPs and 64 regulators with mainly clinical expertise. Attitudes towards PFS were often spread out in both groups and related to beliefs about PFS being a likely surrogate for clinical benefit, being an intrinsic benefit to be distinguished from OS, or on the importance given to OS. Being a regulator or holding stronger beliefs about PFS being a likely surrogate or an intrinsic benefit were associated with a higher MAR. Presence of a supportive trend in OS was stated as important but was not associated with MAR. There was agreement on the need to address bias in the adjudication of PFS and the need for improving communication to patients about meaning, strengths, and limitations of improvements in PFS. CONCLUSION: Attitudes towards PFS were spread out and were associated with individual differences in the willingness to trade between toxicity and PFS. There was agreement on the need to address bias in the adjudication of PFS and improving communication to patients.

Bayesian interim analysis for prospective randomized studies: reanalysis of the acute myeloid leukemia HOVON 132 clinical trial.

Randomized controlled trials (RCTs) are the gold standard to establish the benefit-risk ratio of novel drugs. However, the evaluation of mature results often takes many years. We hypothesized that the addition of Bayesian inference methods at interim analysis time points might accelerate and enforce the knowledge that such trials may generate. In order to test that hypothesis, we retrospectively applied a Bayesian approach to the HOVON 132 trial, in which 800 newly diagnosed AML patients aged 18 to 65 years were randomly assigned to a "7 + 3" induction with or without lenalidomide. Five years after the first patient was recruited, the trial was negative for its primary endpoint with no difference in event-free survival (EFS) between experimental and control groups (hazard ratio [HR] 0.99, p = 0.96) in the final conventional analysis. We retrospectively simulated interim analyses after the inclusion of 150, 300, 450, and 600 patients using a Bayesian methodology to detect early lack of efficacy signals. The HR for EFS comparing the lenalidomide arm with the control treatment arm was 1.21 (95% CI 0.81-1.69), 1.05 (95% CI 0.86-1.30), 1.00 (95% CI 0.84-1.19), and 1.02 (95% CI 0.87-1.19) at interim analysis 1, 2, 3 and 4, respectively. Complete remission rates were lower in the lenalidomide arm, and early deaths more frequent. A Bayesian approach identified that the probability of a clinically relevant benefit for EFS (HR < 0.76, as assumed in the statistical analysis plan) was very low at the first interim analysis (1.2%, 0.6%, 0.4%, and 0.1%, respectively). Similar observations were made for low probabilities of any benefit regarding CR. Therefore, Bayesian analysis significantly adds to conventional methods applied for interim analysis and may thereby accelerate the performance and completion of phase III trials.

Benefits and risks of clofarabine in adult acute lymphoblastic leukemia investigated in depth by multi-state modeling.

BACKGROUND: We recently reported results of the prospective, open-label HOVON-100 trial in 334 adult patients with acute lymphoblastic leukemia (ALL) randomized to first-line treatment with or without clofarabine (CLO). No improvement of event-free survival (EFS) was observed, while a higher proportion of patients receiving CLO obtained minimal residual disease (MRD) negativity. AIM: In order to investigate the effects of CLO in more depth, two multi-state models were developed to identify why CLO did not show a long-term survival benefit despite more MRD-negativity. METHODS: The first model evaluated the effect of CLO on going off-protocol (not due to refractory disease/relapse, completion or death) as a proxy of severe treatment-related toxicity, while the second model evaluated the effect of CLO on obtaining MRD negativity. The subsequent impact of these intermediate events on death or relapsed/refractory disease was assessed in both models. RESULTS: Overall, patients receiving CLO went off-protocol more frequently than control patients (35/168 [21%] vs. 18/166 [11%], p = 0.019; HR 2.00 [1.13-3.52], p = 0.02), especially during maintenance (13/44 [30%] vs. 6/56 [11%]; HR 2.85 [95%CI 1.08-7.50], p = 0.035). Going off-protocol was, however, not associated with more relapse or death. Patients in the CLO arm showed a trend towards an increased rate of MRD-negativity compared with control patients (HR MRD-negativity: 1.35 [0.95-1.91], p = 0.10), which did not translate into a significant survival benefit. CONCLUSION: We conclude that the intermediate states, i.e., going off-protocol and MRD-negativity, were affected by adding CLO, but these transitions were not associated with subsequent survival estimates, suggesting relatively modest antileukemic activity in ALL.