David van Klaveren
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
David van Klaveren has a total of 348 co-authors. Here are the top co-authors by number of shared publications:
- Ewout Willem Steyerberg
University Medical Center Utrecht
12 shared publications - David M. Kent
Predictive Analytics and Comparative Effectiveness Center, Tufts Medical Center, Boston, MA
11 shared publications - Carolien C. H. M. Maas
Department of Public Health Erasmus University Medical Center Rotterdam The Netherlands
7 shared publications - Hester Floor Lingsma
Department of Public Health, Erasmus MC University Medical Center, Rotterdam, The Netherlands
5 shared publications David M. Kent
Predictive Analytics and Comparative Effectiveness Center Institute for Clinical Research and Health Policy Studies, Tufts Medical Center Boston Massachusetts USA
4 shared publications- Andrew I. R. Maas
Antwerp University Hospital
3 shared publications
Latest Publications
- Open Access
- 17 October 2026
- Published by Wiley in Statistics in Medicine
- Vol. 45 (23-24) , e70783
Locally Estimated Scatterplot Smoothing for Censored Survival Data
In the absence of censoring, locally estimated scatterplot smoothing (LOESS) is a powerful tool to estimate the relationship between a continuous covariate and the probability of an event and is often used to visually assess calibration of prognostic models. Since follow‐up in prognostic studies is often incomplete, we aimed to study the extension of LOESS to right‐censored data. In line with other work, we first obtained inverse‐probability‐of‐censoring (IPC) weights through Kaplan–Meier estimation. When censoring varied by covariate levels, we used Cox regression with a spline function of the covariate to obtain IPC weights. We then applied LOESS to the weighted event indicators to estimate the relationship between the covariate and the probability of an event within a specific time horizon. We illustrated the application of LOESS for censored data (LOESScd) using patients with complex coronary artery disease. We used the LOESScd to estimate the association between continuous predictors and the probability of death within 10 years. LOESScd estimates closely followed grouped Kaplan–Meier estimates, especially in case of weak smoothing. Covariate‐dependent IPC weights improved LOESScd when censoring varied by covariate levels. We then examined the finite‐sample behavior of LOESScd in simulations motivated by four typical prognostic model validation scenarios. In 2000 samples of 1800 patients, bias was limited and coverage was close to nominal. In conclusion, LOESScd is a promising estimator of the relationship of a continuous covariate with the probability of an event in the presence of right‐censoring. LOESScd is especially useful for estimating calibration curves when validating prognostic models.
- 17 October 2026
- Published by Wiley in Statistics in Medicine
- Vol. 45 (23-24) , e70783
Open AccessLocally Estimated Scatterplot Smoothing for Censored Survival Data
In the absence of censoring, locally estimated scatterplot smoothing (LOESS) is a powerful tool to estimate the relationship between a continuous covariate and the probability of an event and is often used to visually assess calibration of prognostic models. Since follow‐up in prognostic studies is often incomplete, we aimed to study the extension of LOESS to right‐censored data. In line with other work, we first obtained inverse‐probability‐of‐censoring (IPC) weights through Kaplan–Meier estimation. When censoring varied by covariate levels, we used Cox regression with a spline function of the covariate to obtain IPC weights. We then applied LOESS to the weighted event indicators to estimate the relationship between the covariate and the probability of an event within a specific time horizon. We illustrated the application of LOESS for censored data (LOESScd) using patients with complex coronary artery disease. We used the LOESScd to estimate the association between continuous predictors and the probability of death within 10 years. LOESScd estimates closely followed grouped Kaplan–Meier estimates, especially in case of weak smoothing. Covariate‐dependent IPC weights improved LOESScd when censoring varied by covariate levels. We then examined the finite‐sample behavior of LOESScd in simulations motivated by four typical prognostic model validation scenarios. In 2000 samples of 1800 patients, bias was limited and coverage was close to nominal. In conclusion, LOESScd is a promising estimator of the relationship of a continuous covariate with the probability of an event in the presence of right‐censoring. LOESScd is especially useful for estimating calibration curves when validating prognostic models.
- Open Access
- 26 June 2026
- Published by Elsevier in Annals of Diagnostic Pathology
- Vol. 85, 152675
Improved prostate cancer grading by incorporating Gleason pattern quantification, invasive cribriform and intraductal carcinoma in the new QUICC-score
- 26 June 2026
- Published by Elsevier in Annals of Diagnostic Pathology
- Vol. 85, 152675
Open AccessImproved prostate cancer grading by incorporating Gleason pattern quantification, invasive cribriform and intraductal carcinoma in the new QUICC-score
- Open Access
- 22 June 2026
- Published by Tijdschrift voor Jeugdgezondheidszorg in JGZ Tijdschrift voor jeugdgezondheidszorg
Een kansrijke start voor alle kinderen: een populatie-perspectief. Ongelijkheid in geboorte-uitkomsten in Nederland op basis van landelijke registratiedata
Inleiding Effectief Kansrijke Start-beleid vereist inzicht in de verdeling van ongunstige geboorte-uitkomsten binnen de bevolking. Deze studie beschrijft ongelijkheid in ongunstige geboorte-uitkomsten naar opleidingsachtergrond van de moeder. Methode Landelijke registratiedata (Perined, Centraal Bureau van de Statistiek) van alle eenlinggeboorten in Nederland (2016–2019; n=639.007) werden gekoppeld. Ongunstige uitkomsten (doodgeboorte, neonatale sterfte, vroeggeboorte, te klein voor zwangerschapsduur, Apgar-score op 5 minuten <7, opname op neonatale intensive care, ernstige aangeboren afwijkingen) werden geanalyseerd voor vijf opleidingscategorieën. Resultaten Eén op de zes geboorten had een ongunstige uitkomst. Risico’s namen trapsgewijs toe in relatie tot genoten opleiding, van 13,5% (master) tot 21,1% (basisonderwijs). Het absolute aantal ongunstige geboorte-uitkomsten was het hoogst in de omvangrijke middengroep met een gematigd risico. Jaarlijks zouden 3703 (95%BI: 3357-4049) ongunstige uitkomsten (14,7% [95%BI: 13,3-16,0]) vermeden worden bij een populatie-breed risico gelijk aan dat van aan de hoogstopgeleiden. Neonatale sterfte was 2,25 keer (95%BI: 1,71-2,79) en doodgeboorten 2,94 keer (95%BI: 2,33-3,55) zo hoog in de laagste versus hoogste opleidingsgroep. Sterfte zou met circa een derde afnemen als de gehele bevolking de sterftecijfers had van de hoogstopgeleiden (populatie attributieve fractie [PAF]doodgeboorte=35,0% [24,4-45,6%]; PAFneonatale sterfte=27,1% [14,4-39,7%]). Ongelijkheden waren kleiner voor de overige uitkomsten (RR-bereik=1,32-1,77; PAF-bereik=13,8-17,7%). Beschouwing Ongelijkheid in geboorte-uitkomsten raakt de hele Nederlandse samenleving. Ondersteuning voor de groepen met het hoogste risico is belangrijk, maar niet genoeg. De grootste gezondheidswinst op populatieniveau is te behalen door het verkleinen van gezondheidsverschillen tussen álle opleidingsgroepen. Dit vraagt om verbreding van het Kansrijke Start programma, door inbedding van populatie-brede primaire preventie.
- 22 June 2026
- Published by Tijdschrift voor Jeugdgezondheidszorg in JGZ Tijdschrift voor jeugdgezondheidszorg
Open AccessEen kansrijke start voor alle kinderen: een populatie-perspectief. Ongelijkheid in geboorte-uitkomsten in Nederland op basis van landelijke registratiedata
Inleiding Effectief Kansrijke Start-beleid vereist inzicht in de verdeling van ongunstige geboorte-uitkomsten binnen de bevolking. Deze studie beschrijft ongelijkheid in ongunstige geboorte-uitkomsten naar opleidingsachtergrond van de moeder. Methode Landelijke registratiedata (Perined, Centraal Bureau van de Statistiek) van alle eenlinggeboorten in Nederland (2016–2019; n=639.007) werden gekoppeld. Ongunstige uitkomsten (doodgeboorte, neonatale sterfte, vroeggeboorte, te klein voor zwangerschapsduur, Apgar-score op 5 minuten <7, opname op neonatale intensive care, ernstige aangeboren afwijkingen) werden geanalyseerd voor vijf opleidingscategorieën. Resultaten Eén op de zes geboorten had een ongunstige uitkomst. Risico’s namen trapsgewijs toe in relatie tot genoten opleiding, van 13,5% (master) tot 21,1% (basisonderwijs). Het absolute aantal ongunstige geboorte-uitkomsten was het hoogst in de omvangrijke middengroep met een gematigd risico. Jaarlijks zouden 3703 (95%BI: 3357-4049) ongunstige uitkomsten (14,7% [95%BI: 13,3-16,0]) vermeden worden bij een populatie-breed risico gelijk aan dat van aan de hoogstopgeleiden. Neonatale sterfte was 2,25 keer (95%BI: 1,71-2,79) en doodgeboorten 2,94 keer (95%BI: 2,33-3,55) zo hoog in de laagste versus hoogste opleidingsgroep. Sterfte zou met circa een derde afnemen als de gehele bevolking de sterftecijfers had van de hoogstopgeleiden (populatie attributieve fractie [PAF]doodgeboorte=35,0% [24,4-45,6%]; PAFneonatale sterfte=27,1% [14,4-39,7%]). Ongelijkheden waren kleiner voor de overige uitkomsten (RR-bereik=1,32-1,77; PAF-bereik=13,8-17,7%). Beschouwing Ongelijkheid in geboorte-uitkomsten raakt de hele Nederlandse samenleving. Ondersteuning voor de groepen met het hoogste risico is belangrijk, maar niet genoeg. De grootste gezondheidswinst op populatieniveau is te behalen door het verkleinen van gezondheidsverschillen tussen álle opleidingsgroepen. Dit vraagt om verbreding van het Kansrijke Start programma, door inbedding van populatie-brede primaire preventie.
- Open Access
- 1 June 2026
- Published by Tijdschrift voor Jeugdgezondheidszorg in JGZ Tijdschrift voor jeugdgezondheidszorg
Het voorspellen van jeugdzorggebruik tussen 0 en 5 jaar: een studie met Nederlandse bevolkingsgegevens uit 2015-2019
Doel Jeugdzorg ondersteunt gezinnen met opvoedproblemen. Jeugdzorggebruik kan mogelijk deels voorkomen worden door vroegtijdige preventieve ondersteuning. Wij ontwikkelden risicomodellen voor het voorspellen van jeugdzorggebruik bij kinderen van 0-5 jaar ter ondersteuning van vroeg-signalering en gerichte preventie. Methoden We gebruikten registratiedata voor alle geboortes in Nederland (2015-2019). Predictoren omvatten buurt- en huishoudkenmerken, sociaaleconomische status, gezondheid en gedrag van ouders, en geboorte-uitkomsten. We ontwikkelden Cox-regressiemodellen voor vier beoordelingsmomenten: één jaar vóór geboorte, bij geboorte, één jaar en twee jaar na geboorte. Resultaten Van 776.559 geboortes, ontvingen 32.290 kinderen minstens één jeugdzorgtraject (4,2%). De modellen hadden een redelijk goed onderscheidend vermogen. Een jaar vóór geboorte was de oppervlakte onder de ROC-curve (AUC) 0,760. De AUC verbeterde minimaal bij elk volgend beoordelingsmoment, tot 0,798 op twee jaar na geboorte. De sterkste voorspellers waren eerder jeugdzorggebruik, opleidingsniveau (ouders), gebruik van psychiatrische medicatie (moeder), en arbeidsstatus (moeder). Wanneer 10% van het cohort met het hoogste risico op jeugdzorg als positief wordt geclassificeerd, was de negatief voorspellende waarde hoog (≥0,972) en de positief voorspellende waarde laag (≤0,164). Conclusie Vroeg-signalering op basis van onze modellen zou gezinnen kunnen ondersteunen alvorens de overbelaste jeugdzorg wordt ingeschakeld, mits de gevolgen van vals-positieve testen kunnen worden beperkt.
- 1 June 2026
- Published by Tijdschrift voor Jeugdgezondheidszorg in JGZ Tijdschrift voor jeugdgezondheidszorg
Open AccessHet voorspellen van jeugdzorggebruik tussen 0 en 5 jaar: een studie met Nederlandse bevolkingsgegevens uit 2015-2019
Doel Jeugdzorg ondersteunt gezinnen met opvoedproblemen. Jeugdzorggebruik kan mogelijk deels voorkomen worden door vroegtijdige preventieve ondersteuning. Wij ontwikkelden risicomodellen voor het voorspellen van jeugdzorggebruik bij kinderen van 0-5 jaar ter ondersteuning van vroeg-signalering en gerichte preventie. Methoden We gebruikten registratiedata voor alle geboortes in Nederland (2015-2019). Predictoren omvatten buurt- en huishoudkenmerken, sociaaleconomische status, gezondheid en gedrag van ouders, en geboorte-uitkomsten. We ontwikkelden Cox-regressiemodellen voor vier beoordelingsmomenten: één jaar vóór geboorte, bij geboorte, één jaar en twee jaar na geboorte. Resultaten Van 776.559 geboortes, ontvingen 32.290 kinderen minstens één jeugdzorgtraject (4,2%). De modellen hadden een redelijk goed onderscheidend vermogen. Een jaar vóór geboorte was de oppervlakte onder de ROC-curve (AUC) 0,760. De AUC verbeterde minimaal bij elk volgend beoordelingsmoment, tot 0,798 op twee jaar na geboorte. De sterkste voorspellers waren eerder jeugdzorggebruik, opleidingsniveau (ouders), gebruik van psychiatrische medicatie (moeder), en arbeidsstatus (moeder). Wanneer 10% van het cohort met het hoogste risico op jeugdzorg als positief wordt geclassificeerd, was de negatief voorspellende waarde hoog (≥0,972) en de positief voorspellende waarde laag (≤0,164). Conclusie Vroeg-signalering op basis van onze modellen zou gezinnen kunnen ondersteunen alvorens de overbelaste jeugdzorg wordt ingeschakeld, mits de gevolgen van vals-positieve testen kunnen worden beperkt.
- Open Access
Patient Versus Prediction-Level Evaluation of a Dynamic Clinical Prediction Model of Sepsis
The Epic Sepsis Model version 2 (ESMv2) is a prediction model embedded into the electronic medical record used to warn clinicians which hospitalized patients are at risk for sepsis. We conducted a retrospective cohort study of 31,951 hospitalizations of 25,760 patients to compare analyses conducted at the commonly used patient-level (where a maximum prediction prior to the onset of sepsis is used to measure performance) vs novel prediction-level (where each prediction is used to measure performance). Sepsis, defined by the Sepsis 3 criteria occurred during 1,049 hospitalizations (3.3%). Patient-level analyses suggested excellent discrimination AUC 0.86; [IQR 0.85, 0.87], whereas prediction-level analyses demonstrated lower performance AUC 0.62; [IQR 0.57, 0.65]. Low estimates of the positive predictive value (14.5% at the patient level vs 4% at the prediction level) imply a high number of false alerts. Common evaluation approaches may overstate the performance of dynamic prediction models and mislead clinical decision-making.
Open AccessPatient Versus Prediction-Level Evaluation of a Dynamic Clinical Prediction Model of Sepsis
The Epic Sepsis Model version 2 (ESMv2) is a prediction model embedded into the electronic medical record used to warn clinicians which hospitalized patients are at risk for sepsis. We conducted a retrospective cohort study of 31,951 hospitalizations of 25,760 patients to compare analyses conducted at the commonly used patient-level (where a maximum prediction prior to the onset of sepsis is used to measure performance) vs novel prediction-level (where each prediction is used to measure performance). Sepsis, defined by the Sepsis 3 criteria occurred during 1,049 hospitalizations (3.3%). Patient-level analyses suggested excellent discrimination AUC 0.86; [IQR 0.85, 0.87], whereas prediction-level analyses demonstrated lower performance AUC 0.62; [IQR 0.57, 0.65]. Low estimates of the positive predictive value (14.5% at the patient level vs 4% at the prediction level) imply a high number of false alerts. Common evaluation approaches may overstate the performance of dynamic prediction models and mislead clinical decision-making.
- Open Access
- 17 March 2026
- Published by Oxford University Press (OUP) in JNCI: Journal of the National Cancer Institute
- Vol. 118 (6) , 1108-1114
Restaging after neoadjuvant FOLFIRINOX for localized pancreatic cancer: a clinical calculator from the Trans-Atlantic Pancreatic Surgery consortium
Background: Restaging after neoadjuvant chemotherapy aims to assess treatment response, revise prognosis, and guide further management. This study investigated independent prognostic factors for overall survival after restaging in patients with localized pancreatic adenocarcinoma. Methods: In this retrospective international study, consecutive patients with localized pancreatic adenocarcinoma who received at least 1 cycle of (m)FOLFIRINOX as first-line therapy were identified. Multivariable Cox regression analysis was performed with a web-based calculator to predict individualized overall survival. Results: The Trans-Atlantic Pancreatic Surgery cohort included 2338 patients with localized pancreatic adenocarcinoma of whom 22.6% were potentially resectable, 30.7% borderline resectable, and 46.7% locally advanced at initial staging. Several baseline characteristics remained independent prognostic factors for overall survival after restaging: borderline resectable (hazard ratio [HR] = 1.31, 95% confidence interval [CI] = 1.14 to 1.51), locally advanced (HR = 1.78, 95% CI = 1.55 to 2.05), body or tail tumor (HR = 0.79, 95% CI = 0.68 to 0.90), and baseline World Health Organization performance status of 1 (HR = 1.18, 95% CI = 1.07 to 1.30) or at least 2 (HR = 1.54, 95% CI = 1.20 to 1.98). Additional independent factors were metastatic disease at restaging (HR = 1.57, 95% CI = 1.31 to 1.87), postinduction carbohydrate antigen 19-9 (HR = 1.47, 95% CI = 1.37 to 1.57), Δcarbohydrate antigen 19-9 (HR = 0.83, 95% CI = 0.77 to 0.89), postinduction tumor size (HR = 1.22, 95% CI = 1.11 to 1.35), and Δtumor size (HR = 0.92, 95% CI = 0.87 to 0.98). Patients were stratified into 4 risk groups, with 3-year overall survival after restaging ranging from 6.0% to 65.8%. Conclusion: Survival at restaging after neoadjuvant chemotherapy of patients with localized pancreatic adenocarcinoma is determined by 8 patient, tumor, and treatment response characteristics. A web-based calculator can inform clinicians and patients about individualized prognosis and guide further management.
- 17 March 2026
- Published by Oxford University Press (OUP) in JNCI: Journal of the National Cancer Institute
- Vol. 118 (6) , 1108-1114
Open AccessRestaging after neoadjuvant FOLFIRINOX for localized pancreatic cancer: a clinical calculator from the Trans-Atlantic Pancreatic Surgery consortium
Background: Restaging after neoadjuvant chemotherapy aims to assess treatment response, revise prognosis, and guide further management. This study investigated independent prognostic factors for overall survival after restaging in patients with localized pancreatic adenocarcinoma. Methods: In this retrospective international study, consecutive patients with localized pancreatic adenocarcinoma who received at least 1 cycle of (m)FOLFIRINOX as first-line therapy were identified. Multivariable Cox regression analysis was performed with a web-based calculator to predict individualized overall survival. Results: The Trans-Atlantic Pancreatic Surgery cohort included 2338 patients with localized pancreatic adenocarcinoma of whom 22.6% were potentially resectable, 30.7% borderline resectable, and 46.7% locally advanced at initial staging. Several baseline characteristics remained independent prognostic factors for overall survival after restaging: borderline resectable (hazard ratio [HR] = 1.31, 95% confidence interval [CI] = 1.14 to 1.51), locally advanced (HR = 1.78, 95% CI = 1.55 to 2.05), body or tail tumor (HR = 0.79, 95% CI = 0.68 to 0.90), and baseline World Health Organization performance status of 1 (HR = 1.18, 95% CI = 1.07 to 1.30) or at least 2 (HR = 1.54, 95% CI = 1.20 to 1.98). Additional independent factors were metastatic disease at restaging (HR = 1.57, 95% CI = 1.31 to 1.87), postinduction carbohydrate antigen 19-9 (HR = 1.47, 95% CI = 1.37 to 1.57), Δcarbohydrate antigen 19-9 (HR = 0.83, 95% CI = 0.77 to 0.89), postinduction tumor size (HR = 1.22, 95% CI = 1.11 to 1.35), and Δtumor size (HR = 0.92, 95% CI = 0.87 to 0.98). Patients were stratified into 4 risk groups, with 3-year overall survival after restaging ranging from 6.0% to 65.8%. Conclusion: Survival at restaging after neoadjuvant chemotherapy of patients with localized pancreatic adenocarcinoma is determined by 8 patient, tumor, and treatment response characteristics. A web-based calculator can inform clinicians and patients about individualized prognosis and guide further management.
- Open Access
- 18 February 2026
- Published by Elsevier in SSRN Electronic Journal
Regional variations in molecular testing and treatment of patients with metastatic NSCLC in the Netherlands
Background: Molecular testing combined with targeted therapy has led to a rapid improvement of patient outcomes in metastatic non-small cell lung cancer (NSCLC). However, previous reports suggested that the use of these treatment modalities might vary between geographic regions within countries. We aimed to examine variation in diagnostics and treatment of non-squamous metastatic NSCLC.Methods: We compared five Dutch regions using population-based data from the Netherlands Cancer Registry (2020-2023). Variation in molecular testing and first-line targeted therapy across regions was assessed with the likelihood-ratio test (LRT) of a multivariable logistic regression model. This model included region as a variable, and accounted for patient characteristics such as sex, age, PDL-1 status, year of diagnosis and World Health Organization performance status. Variation in treatment was also assessed with the LRT in patients with actionable genetic aberrations (i.e. EGFR, ALK, and BRAF V600E).Results: A total of 16,655 patients (median age 69 [62, 75]; 52% male) were diagnosed with metastatic non-squamous NSCLC. Of these patients,14,297 (86%) underwent molecular testing, ranging from 82% to 91% between regions (p<.001). 1,470 patients had an actionable genetic aberration, of which 1,252 (85%) received targeted therapy, ranging from 82% to 89% (p=.17).Conclusion: The application of molecular testing in metastatic non-squamous NSCLC varies across regions in the Netherlands, while minimal variation was observed in the application of targeted therapy. These findings highlight the need for better collaboration between regions to optimize guidelines and ensure uniform care.
- 18 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRegional variations in molecular testing and treatment of patients with metastatic NSCLC in the Netherlands
Background: Molecular testing combined with targeted therapy has led to a rapid improvement of patient outcomes in metastatic non-small cell lung cancer (NSCLC). However, previous reports suggested that the use of these treatment modalities might vary between geographic regions within countries. We aimed to examine variation in diagnostics and treatment of non-squamous metastatic NSCLC.Methods: We compared five Dutch regions using population-based data from the Netherlands Cancer Registry (2020-2023). Variation in molecular testing and first-line targeted therapy across regions was assessed with the likelihood-ratio test (LRT) of a multivariable logistic regression model. This model included region as a variable, and accounted for patient characteristics such as sex, age, PDL-1 status, year of diagnosis and World Health Organization performance status. Variation in treatment was also assessed with the LRT in patients with actionable genetic aberrations (i.e. EGFR, ALK, and BRAF V600E).Results: A total of 16,655 patients (median age 69 [62, 75]; 52% male) were diagnosed with metastatic non-squamous NSCLC. Of these patients,14,297 (86%) underwent molecular testing, ranging from 82% to 91% between regions (p<.001). 1,470 patients had an actionable genetic aberration, of which 1,252 (85%) received targeted therapy, ranging from 82% to 89% (p=.17).Conclusion: The application of molecular testing in metastatic non-squamous NSCLC varies across regions in the Netherlands, while minimal variation was observed in the application of targeted therapy. These findings highlight the need for better collaboration between regions to optimize guidelines and ensure uniform care.
- 14 October 2025
- Published by Oxford University Press (OUP) in International Journal of Epidemiology
- Vol. 54 (6)
Diverging conclusions from risk difference and difference in restricted mean survival time in quantifying absolute treatment effect heterogeneity
Background: Risk-based analyses are increasingly popular for understanding heterogeneous treatment effects (HTEs) in clinical trials. For time-to-event analyses, the assumption that high-risk patients benefit most on the clinically important absolute scale when hazard ratios (HRs) are constant across risk strata might not hold. Absolute treatment effects can be measured as either the risk difference (RD) at a given time point or the difference in the restricted mean survival time (ΔRMST), which aligns more closely with utilitarian medical decision-making frameworks. We examined risk-based HTE analyses strata in time-to-event analyses to identify the patterns of absolute HTE across risk strata and whether the ΔRMST may lead to better treatment decisions than the RD. Methods: Using artificial and empirical time-to-event data, we compared the RD—the difference between Kaplan–Meier estimates at a certain time point—and the ΔRMST—the area between the Kaplan–Meier curves—across risk strata and show how these metrics can prioritize different subgroups for treatment. We explored scenarios involving constant HRs while varying both the overall event rates and the discrimination of the risk models. Results: When event rates and discrimination were low, the RD and the ΔRMST increased monotonically, with high-risk patients benefitting more than low-risk patients. As the event rate increased and/or discrimination increased, a ‘sweet spot’ pattern emerged: intermediate-risk patients benefit more than low-risk and high-risk patients. When the RD was used, the ‘sweet spot’ pattern emerged, even in circumstances in which the ΔRMST increased across the risk groups, thus understating the benefit for higher-risk patients and potentially leading to treatment mistargeting. Conclusion: The pattern of HTE characterized by the RD may diverge substantially from the ΔRMST, potentially leading to treatment mistargeting. Therefore, we recommend the ΔRMST for assessing the absolute HTE in time-to-event data.
- 14 October 2025
- Published by Oxford University Press (OUP) in International Journal of Epidemiology
- Vol. 54 (6)
Diverging conclusions from risk difference and difference in restricted mean survival time in quantifying absolute treatment effect heterogeneity
Background: Risk-based analyses are increasingly popular for understanding heterogeneous treatment effects (HTEs) in clinical trials. For time-to-event analyses, the assumption that high-risk patients benefit most on the clinically important absolute scale when hazard ratios (HRs) are constant across risk strata might not hold. Absolute treatment effects can be measured as either the risk difference (RD) at a given time point or the difference in the restricted mean survival time (ΔRMST), which aligns more closely with utilitarian medical decision-making frameworks. We examined risk-based HTE analyses strata in time-to-event analyses to identify the patterns of absolute HTE across risk strata and whether the ΔRMST may lead to better treatment decisions than the RD. Methods: Using artificial and empirical time-to-event data, we compared the RD—the difference between Kaplan–Meier estimates at a certain time point—and the ΔRMST—the area between the Kaplan–Meier curves—across risk strata and show how these metrics can prioritize different subgroups for treatment. We explored scenarios involving constant HRs while varying both the overall event rates and the discrimination of the risk models. Results: When event rates and discrimination were low, the RD and the ΔRMST increased monotonically, with high-risk patients benefitting more than low-risk patients. As the event rate increased and/or discrimination increased, a ‘sweet spot’ pattern emerged: intermediate-risk patients benefit more than low-risk and high-risk patients. When the RD was used, the ‘sweet spot’ pattern emerged, even in circumstances in which the ΔRMST increased across the risk groups, thus understating the benefit for higher-risk patients and potentially leading to treatment mistargeting. Conclusion: The pattern of HTE characterized by the RD may diverge substantially from the ΔRMST, potentially leading to treatment mistargeting. Therefore, we recommend the ΔRMST for assessing the absolute HTE in time-to-event data.
- Open Access
- 2 September 2025
- Published by Wiley in Clinical Transplantation
- Vol. 39 (9) , e70299
A Comprehensive Assessment of Plasma CXCL9 and CXCL10 in Improving Clinical Prediction Models for Kidney Allograft Rejection
Chemokine levels may predict kidney graft rejection. This study evaluated whether adding early plasma chemokines C‐X‐C motif ligand 9 (CXCL9) or chemokines C‐X‐C motif ligand 10 (CXCL10) measurements to a standard‐of‐care model improves the prediction of the need for antirejection treatment and helps guide biopsy decisions. The benchmark model used recipient and donor age, human leukocyte antigen mismatches, and dialysis need in the first 3 days after transplantation. Plasma CXCL9 or CXCL10 was added, and the extended models were evaluated using likelihood ratio tests (LRTs), area under the receiver operating characteristic curve (ROC‐AUC), flexible calibration curves, and net benefit analysis. Model internal validation was performed through bootstrapping. Among 163 consecutively transplanted recipients on standard immunosuppression, 43 (26.4%) required antirejection therapy between Days 3 and 21 posttransplant. The chemokine‐extended models outperformed the benchmark (LRT p < 0.01), increasing discriminative ability (delta ROC‐AUC of 0.02) and improving calibration. Across the range of risk thresholds for biopsy, the extended models provided better clinical utility, resulting in up to 17 fewer unnecessary biopsies per 100 patients. These findings suggest that adding plasma CXCL9 or CXCL10 to a benchmark model can improve patient care by reducing the number of biopsies in individuals unlikely to require antirejection therapy.
- 2 September 2025
- Published by Wiley in Clinical Transplantation
- Vol. 39 (9) , e70299
Open AccessA Comprehensive Assessment of Plasma CXCL9 and CXCL10 in Improving Clinical Prediction Models for Kidney Allograft Rejection
Chemokine levels may predict kidney graft rejection. This study evaluated whether adding early plasma chemokines C‐X‐C motif ligand 9 (CXCL9) or chemokines C‐X‐C motif ligand 10 (CXCL10) measurements to a standard‐of‐care model improves the prediction of the need for antirejection treatment and helps guide biopsy decisions. The benchmark model used recipient and donor age, human leukocyte antigen mismatches, and dialysis need in the first 3 days after transplantation. Plasma CXCL9 or CXCL10 was added, and the extended models were evaluated using likelihood ratio tests (LRTs), area under the receiver operating characteristic curve (ROC‐AUC), flexible calibration curves, and net benefit analysis. Model internal validation was performed through bootstrapping. Among 163 consecutively transplanted recipients on standard immunosuppression, 43 (26.4%) required antirejection therapy between Days 3 and 21 posttransplant. The chemokine‐extended models outperformed the benchmark (LRT p < 0.01), increasing discriminative ability (delta ROC‐AUC of 0.02) and improving calibration. Across the range of risk thresholds for biopsy, the extended models provided better clinical utility, resulting in up to 17 fewer unnecessary biopsies per 100 patients. These findings suggest that adding plasma CXCL9 or CXCL10 to a benchmark model can improve patient care by reducing the number of biopsies in individuals unlikely to require antirejection therapy.
- 28 August 2025
- Published by Elsevier in The Lancet
- Vol. 406 (10506) , 913-914
Estimating the benefits of personalised glycaemic control
- 28 August 2025
- Published by Elsevier in The Lancet
- Vol. 406 (10506) , 913-914
Estimating the benefits of personalised glycaemic control