Camille Maringe
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Camille Maringe has a total of 157 co-authors. Here are the top co-authors by number of shared publications:
- Bernard Rachet
London School of Hygiene and Tropical Medicine
14 shared publications Bernard Rachet
Inequalities in Cancer Outcomes Network, Department of Health Services Research and Policy, Faculty of Public Health and Policy London School of Hygiene and Tropical Medicine London UK
7 shared publications- Aurélien Belot
London School of Hygiene & Tropical Medicine
6 shared publications - Miguel Angel Luque-Fernandez
Inequalities in Cancer Outcomes Network London School of Hygiene and Tropical Medicine London UK
6 shared publications - Clémence Leyrat
Department of Medical Statistics, London School of Hygiene and Tropical Medicine, UK; and
5 shared publications - Francisco Javier Rubio
Department of Statistical Science, University College London, London, UK
5 shared publications
Latest Publications
- Open Access
Beyond Signal Detection: Sequential Target Trial Emulations to Confirm Previously Detected Adverse Drug Event Signals for Atorvastatin in Older Medicare Beneficiaries
Background and objective: High dimensional, hypothesis free signal detection in claims data can identify adverse drug events (ADEs), yet these exploratory signals remain vulnerable to false positives and residual confounding. Rigorous internal confirmation using causal inference methods is therefore a critical intermediate step in a staged pharmacovigilance paradigm before external validation. We aimed to confirm or refute previously detected ADE signals associated with atorvastatin initiation among older adults after myocardial or cerebral infarction. Methods: Internal confirmatory sequential target trial emulations were conducted using Medicare fee-for-service claims (2017–2019). Eligible participants were statin naïve beneficiaries aged ≥ 65 years hospitalized for myocardial or cerebral infarction. Up to 14 nested daily trials were constructed beginning on the discharge date. Pragmatic treatment strategies compared atorvastatin initiation with initiation of a different new outpatient medication. Per protocol effects were estimated using inverse probability weighted Fine-Gray models treating death as a competing risk. Previously detected ADE signals and clinically coherent alternatives within the same outcome families were evaluated according to prespecified dual criteria controlling for the false discovery rate and residual bias. Absolute risks and numbers needed to harm (NNH) were reported. A prespecified sensitivity analysis restricted inference to the first two trials with optimal covariate balance; a post hoc analysis examined effect modification by concomitant high risk antithrombotic therapy. Results: Of 70,130 eligible patients, 39,948 initiated atorvastatin (81% high intensity) and 19,182 initiated a different new medication. After weighting, baseline characteristics were closely balanced. Associations meeting dual confirmatory criteria formed a gradient of support. The strongest findings were early (days 1 to 30) acute hemorrhagic cerebrovascular disease (sHR 2.20, 95% CI 1.35-3.58; NNH 351) and acute hepatic failure (sHR 1.72, 1.16-2.55; NNH 468), both robust to high risk antithrombotic therapy and to the restricted trial set with optimal covariate balance. Intermediate support was observed for biliary tract disease (women; sHR 1.45, 1.14-1.85; NNH 102) and musculoskeletal injuries (men; sHR 1.66, 1.24-2.21; NNH 71). Tentative evidence was observed for cardiac valve disorders (attenuated in the absence of high risk antithrombotic therapy) and sensory symptoms (non-White patients). Prediabetes and posthemorrhagic anemia were not confirmed. Conclusions: Several previously detected ADE associations with atorvastatin initiation met dual confirmatory criteria and formed a gradient of confirmatory support. Strongest evidence supported early hemorrhagic cerebrovascular disease, hepatic dysfunction, and musculoskeletal injuries. Absolute excess risks were modest yet clinically relevant to a high risk post-infarction population (NNH 71-468) and must be interpreted alongside the established benefits of high intensity statin therapy (number needed to treat to prevent one event (NNT) 28-93). These findings support a two stage active pharmacovigilance paradigm (signal detection followed by rigorous confirmation) and justify heightened clinical vigilance for the confirmed events, while underscoring the need for external validation in independent populations and data sources. KEY POINTS: Question: Can previously detected signals of adverse drug events associated with atorvastatin initiation in older adults be confirmed under more rigorous dual confirmatory criteria? Findings: In this internal confirmatory Medicare claims study using sequential target trial emulation, associations meeting dual confirmatory criteria formed a gradient of support. The strongest findings were early acute hemorrhagic cerebrovascular disease (sHR 2.20) and acute hepatic failure (sHR 1.72), both robust to high risk antithrombotic therapy and to the optimally balanced early trials. Intermediate support was observed for biliary tract disease and musculoskeletal injuries. Weaker evidence was observed for cardiac valve disorders (attenuated in the absence of high risk antithrombotic therapy) and sensory symptoms. Prediabetes and posthemorrhagic anemia were not confirmed. Meaning: Although atorvastatin remains foundational for secondary cardiovascular prevention (NNT 28–93 to prevent one event), these findings support heightened clinical vigilance for early hemorrhagic, hepatic, and musculoskeletal events in older adults after myocardial or cerebral infarction, while underscoring the need for external validation. PLAIN LANGUAGE SUMMARY: Older adults who start atorvastatin after a heart attack or stroke may face a higher risk of certain serious side effects, according to this large study of Medicare insurance records. Researchers previously identified possible safety signals for atorvastatin using advanced methods that mimic clinical trials with real world data. This follow up study carefully reexamined those signals with stricter eligibility rules, more precise outcome definitions, and statistical techniques designed to reduce false findings. The analysis confirmed a range of associations that varied in strength. The strongest evidence pointed to higher rates of bleeding in the brain (especially in the first 30 days) and acute liver failure. Intermediate evidence supported bile duct problems and muscle strains or joint injuries. Weaker evidence was found for certain heart valve problems and for dizziness or sensory symptoms. Some earlier possible signals, such as prediabetes, were not confirmed. Although atorvastatin remains an important treatment to prevent future heart attacks and strokes, these findings suggest that older adults starting the drug may benefit from close monitoring for early bleeding in the brain, liver problems, and muscle or joint injuries, and that suspicion for these...
Open AccessBeyond Signal Detection: Sequential Target Trial Emulations to Confirm Previously Detected Adverse Drug Event Signals for Atorvastatin in Older Medicare Beneficiaries
Background and objective: High dimensional, hypothesis free signal detection in claims data can identify adverse drug events (ADEs), yet these exploratory signals remain vulnerable to false positives and residual confounding. Rigorous internal confirmation using causal inference methods is therefore a critical intermediate step in a staged pharmacovigilance paradigm before external validation. We aimed to confirm or refute previously detected ADE signals associated with atorvastatin initiation among older adults after myocardial or cerebral infarction. Methods: Internal confirmatory sequential target trial emulations were conducted using Medicare fee-for-service claims (2017–2019). Eligible participants were statin naïve beneficiaries aged ≥ 65 years hospitalized for myocardial or cerebral infarction. Up to 14 nested daily trials were constructed beginning on the discharge date. Pragmatic treatment strategies compared atorvastatin initiation with initiation of a different new outpatient medication. Per protocol effects were estimated using inverse probability weighted Fine-Gray models treating death as a competing risk. Previously detected ADE signals and clinically coherent alternatives within the same outcome families were evaluated according to prespecified dual criteria controlling for the false discovery rate and residual bias. Absolute risks and numbers needed to harm (NNH) were reported. A prespecified sensitivity analysis restricted inference to the first two trials with optimal covariate balance; a post hoc analysis examined effect modification by concomitant high risk antithrombotic therapy. Results: Of 70,130 eligible patients, 39,948 initiated atorvastatin (81% high intensity) and 19,182 initiated a different new medication. After weighting, baseline characteristics were closely balanced. Associations meeting dual confirmatory criteria formed a gradient of support. The strongest findings were early (days 1 to 30) acute hemorrhagic cerebrovascular disease (sHR 2.20, 95% CI 1.35-3.58; NNH 351) and acute hepatic failure (sHR 1.72, 1.16-2.55; NNH 468), both robust to high risk antithrombotic therapy and to the restricted trial set with optimal covariate balance. Intermediate support was observed for biliary tract disease (women; sHR 1.45, 1.14-1.85; NNH 102) and musculoskeletal injuries (men; sHR 1.66, 1.24-2.21; NNH 71). Tentative evidence was observed for cardiac valve disorders (attenuated in the absence of high risk antithrombotic therapy) and sensory symptoms (non-White patients). Prediabetes and posthemorrhagic anemia were not confirmed. Conclusions: Several previously detected ADE associations with atorvastatin initiation met dual confirmatory criteria and formed a gradient of confirmatory support. Strongest evidence supported early hemorrhagic cerebrovascular disease, hepatic dysfunction, and musculoskeletal injuries. Absolute excess risks were modest yet clinically relevant to a high risk post-infarction population (NNH 71-468) and must be interpreted alongside the established benefits of high intensity statin therapy (number needed to treat to prevent one event (NNT) 28-93). These findings support a two stage active pharmacovigilance paradigm (signal detection followed by rigorous confirmation) and justify heightened clinical vigilance for the confirmed events, while underscoring the need for external validation in independent populations and data sources. KEY POINTS: Question: Can previously detected signals of adverse drug events associated with atorvastatin initiation in older adults be confirmed under more rigorous dual confirmatory criteria? Findings: In this internal confirmatory Medicare claims study using sequential target trial emulation, associations meeting dual confirmatory criteria formed a gradient of support. The strongest findings were early acute hemorrhagic cerebrovascular disease (sHR 2.20) and acute hepatic failure (sHR 1.72), both robust to high risk antithrombotic therapy and to the optimally balanced early trials. Intermediate support was observed for biliary tract disease and musculoskeletal injuries. Weaker evidence was observed for cardiac valve disorders (attenuated in the absence of high risk antithrombotic therapy) and sensory symptoms. Prediabetes and posthemorrhagic anemia were not confirmed. Meaning: Although atorvastatin remains foundational for secondary cardiovascular prevention (NNT 28–93 to prevent one event), these findings support heightened clinical vigilance for early hemorrhagic, hepatic, and musculoskeletal events in older adults after myocardial or cerebral infarction, while underscoring the need for external validation. PLAIN LANGUAGE SUMMARY: Older adults who start atorvastatin after a heart attack or stroke may face a higher risk of certain serious side effects, according to this large study of Medicare insurance records. Researchers previously identified possible safety signals for atorvastatin using advanced methods that mimic clinical trials with real world data. This follow up study carefully reexamined those signals with stricter eligibility rules, more precise outcome definitions, and statistical techniques designed to reduce false findings. The analysis confirmed a range of associations that varied in strength. The strongest evidence pointed to higher rates of bleeding in the brain (especially in the first 30 days) and acute liver failure. Intermediate evidence supported bile duct problems and muscle strains or joint injuries. Weaker evidence was found for certain heart valve problems and for dizziness or sensory symptoms. Some earlier possible signals, such as prediabetes, were not confirmed. Although atorvastatin remains an important treatment to prevent future heart attacks and strokes, these findings suggest that older adults starting the drug may benefit from close monitoring for early bleeding in the brain, liver problems, and muscle or joint injuries, and that suspicion for these...
- Open Access
- 14 July 2026
- Published by Wiley in British Journal of Haematology
- Vol. 209 (3) , 1128-1135
Socioeconomic inequality in treatment of acute myeloid leukaemia ( AML ) in England, 2015–2022: A population‐based study
Summary: The socioeconomic inequality in acute myeloid leukaemia (AML) survival may be driven by differential access to intensive chemotherapy (ICT). We aimed to quantify socioeconomic inequality in the receipt of ICT among patients with AML in England and identify its key drivers. We evaluated 10 495 patients with de novo AML identified in the national cancer registry of England (2015–2022) along with their area‐deprivation status, as a measure of socioeconomic inequality, and treatment information based on linkage databases. We related prevalence of the ICT receipt to deprivation with multivariable‐adjusted generalised regression and mixed‐effects probit regression. Overall, 5190 patients (49%) received ICT. The patients in the most deprived area were 6.1% less likely to receive ICT than those in the least deprived area. The differences in the prevalence measures varied across the 125 administrative clusters of health service units. If all patients had access to ICT similar to the least deprived patients, 247 more patients would receive ICT. Of the deprivation effect, 94% was estimated to be attributable to the variability among administrative clusters. Despite universal healthcare coverage, socioeconomic inequality exists in AML treatment in England. The inequality may be driven by cluster‐level variations such as institutional culture and resource capacity.
- 14 July 2026
- Published by Wiley in British Journal of Haematology
- Vol. 209 (3) , 1128-1135
Open AccessSocioeconomic inequality in treatment of acute myeloid leukaemia ( AML ) in England, 2015–2022: A population‐based study
Summary: The socioeconomic inequality in acute myeloid leukaemia (AML) survival may be driven by differential access to intensive chemotherapy (ICT). We aimed to quantify socioeconomic inequality in the receipt of ICT among patients with AML in England and identify its key drivers. We evaluated 10 495 patients with de novo AML identified in the national cancer registry of England (2015–2022) along with their area‐deprivation status, as a measure of socioeconomic inequality, and treatment information based on linkage databases. We related prevalence of the ICT receipt to deprivation with multivariable‐adjusted generalised regression and mixed‐effects probit regression. Overall, 5190 patients (49%) received ICT. The patients in the most deprived area were 6.1% less likely to receive ICT than those in the least deprived area. The differences in the prevalence measures varied across the 125 administrative clusters of health service units. If all patients had access to ICT similar to the least deprived patients, 247 more patients would receive ICT. Of the deprivation effect, 94% was estimated to be attributable to the variability among administrative clusters. Despite universal healthcare coverage, socioeconomic inequality exists in AML treatment in England. The inequality may be driven by cluster‐level variations such as institutional culture and resource capacity.
- Open Access
Integrating Causal Inference into Pharmacovigilance: Target Trial Emulations for Proactive Signal Detection of Atorvastatin Initiation in Medicare Beneficiaries
Importance: Adverse drug events (ADEs) in older adults are a substantial public health burden, yet spontaneous reporting systems detect them poorly owing to underreporting and the lack of a defined population. These limitations are of particular concern for older adults, who are underrepresented in pre-approval trials yet at elevated ADE risk owing to polypharmacy, multimorbidity, and age-related changes in drug metabolism. Objective: To develop and apply an active, claims-based pharmacovigilance framework using sequential target trial emulation to detect ADE signals in older adults, with atorvastatin as the initial application. Methods: Using Medicare fee-for-service claims (2017-2019), we studied statin-naïve beneficiaries aged 65 years or older following hospitalization for myocardial or cerebral infarction. We emulated up to 14 daily sequential trials from the discharge date, classifying patients as initiating atorvastatin (A1), initiating a different medication (A2), or no new medication (A0); the primary contrast was A1 versus A2. For each trial, incident ADEs were ascertained and classified into 552 outcomes based on the Clinical Classifications Software Refined categories. Per-protocol effect estimates were measured over a 6-month follow-up period using Fine-Gray regression weighted by the inverse probability of treatment (IPT) and censoring, treating death as a competing risk, with the false discovery rate controlled via the Benjamini-Hochberg procedure. A signal was declared when the q-value was ≤ 0.10 and the subdistribution hazard ratio (sHR) was ≥ 1.20 in any prespecified analytic stratum (sensitivity analyses used thresholds of q ≤ 0.20 and sHR ≥ 1.20). Results: Of 70,130 eligible patients, 39,948 initiated atorvastatin (A1), 19,182 initiated another new medication (A2); after IPT-weighting, baseline characteristics were closely balanced. After excluding outcomes with sparse cell counts, 295 outcomes were analyzed; five met the primary signal detection criteria: valve disorders (sHR 1.71, 1.20-2.43); sprains and strains (sHR 1.79, 1.26-2.54); general sensation/perception symptoms (sHR 1.23, 95% CI 1.11-1.36); abnormal findings without diagnosis (sHR 1.55, 1.18-2.05); and prediabetes (sHR 1.71, 1.24-2.36). In the sensitivity analysis, we additionally detected: posthemorrhagic anemia, hemorrhagic stroke, varicose veins, other circulatory and skin conditions. Conclusions: A claims-based framework using sequential target trial emulations detected both expected and previously unrecognized ADE signals following atorvastatin initiation in older adults, offering a systematic alternative to passive surveillance that can be extended to other commonly prescribed medications. Confirmatory analyses with disaggregated outcomes and narrower comparators are underway and will be reported separately. KEY POINTS: Question: Can an active, claims-based pharmacovigilance framework using sequential target trial emulation detect adverse drug event signals among older Medicare beneficiaries who initiated atorvastatin? Findings: In this cohort study of 59,130 Medicare beneficiaries discharged after myocardial or cerebral infarction, a hypothesis-free scan across hundreds of prespecified outcomes detected five primary signals (valve disorders, sprains and strains, sensory symptoms including dizziness, abnormal laboratory findings, and prediabetes) and additional signals in sensitivity analyses (including hemorrhagic stroke), several of which align with known or labeled statin adverse effects (serving as positive controls that validate the framework) while others remain uncertain. Meaning: This framework offers a systematic, quantitative, and proactive alternative to spontaneous reporting for detecting medication safety signals in high-risk older adults. PLAIN LANGUAGE SUMMARY: Older adults use more prescription drugs than any other age group, yet they are often left out of clinical trials conducted before a drug is approved. As a result, the safety of many medications in older patients is poorly understood until the drugs are already in wide use. The main system for detecting new drug safety problems in the United States, the FDA Adverse Event Reporting System, relies on voluntary reports and captures only a small fraction of harms, which limits its usefulness. Using Medicare reimbursement records, the researchers developed an active monitoring method and tested it on a commonly used cholesterol-lowering medication (atorvastatin) – that is frequently prescribed after a heart attack or stroke. They followed thousands of patients for six months, statistically balancing those who started atorvastatin with those who started a different new drug, and screened for hundreds of possible side effects. Evidence of potential medication harm emerged along a clear range of how likely the drug caused each problem. This ranged from well-known side effects (such as diabetes and sprains and strains) and outcomes already noted in the product labeling or prior research (such as hemorrhagic stroke and dizziness) to more uncertain associations (such as cardiac valve disorders) that are difficult to interpret. Overall, atorvastatin appeared largely safe in this older population, and the same method can now be used to study other widely used drugs.
Open AccessIntegrating Causal Inference into Pharmacovigilance: Target Trial Emulations for Proactive Signal Detection of Atorvastatin Initiation in Medicare Beneficiaries
Importance: Adverse drug events (ADEs) in older adults are a substantial public health burden, yet spontaneous reporting systems detect them poorly owing to underreporting and the lack of a defined population. These limitations are of particular concern for older adults, who are underrepresented in pre-approval trials yet at elevated ADE risk owing to polypharmacy, multimorbidity, and age-related changes in drug metabolism. Objective: To develop and apply an active, claims-based pharmacovigilance framework using sequential target trial emulation to detect ADE signals in older adults, with atorvastatin as the initial application. Methods: Using Medicare fee-for-service claims (2017-2019), we studied statin-naïve beneficiaries aged 65 years or older following hospitalization for myocardial or cerebral infarction. We emulated up to 14 daily sequential trials from the discharge date, classifying patients as initiating atorvastatin (A1), initiating a different medication (A2), or no new medication (A0); the primary contrast was A1 versus A2. For each trial, incident ADEs were ascertained and classified into 552 outcomes based on the Clinical Classifications Software Refined categories. Per-protocol effect estimates were measured over a 6-month follow-up period using Fine-Gray regression weighted by the inverse probability of treatment (IPT) and censoring, treating death as a competing risk, with the false discovery rate controlled via the Benjamini-Hochberg procedure. A signal was declared when the q-value was ≤ 0.10 and the subdistribution hazard ratio (sHR) was ≥ 1.20 in any prespecified analytic stratum (sensitivity analyses used thresholds of q ≤ 0.20 and sHR ≥ 1.20). Results: Of 70,130 eligible patients, 39,948 initiated atorvastatin (A1), 19,182 initiated another new medication (A2); after IPT-weighting, baseline characteristics were closely balanced. After excluding outcomes with sparse cell counts, 295 outcomes were analyzed; five met the primary signal detection criteria: valve disorders (sHR 1.71, 1.20-2.43); sprains and strains (sHR 1.79, 1.26-2.54); general sensation/perception symptoms (sHR 1.23, 95% CI 1.11-1.36); abnormal findings without diagnosis (sHR 1.55, 1.18-2.05); and prediabetes (sHR 1.71, 1.24-2.36). In the sensitivity analysis, we additionally detected: posthemorrhagic anemia, hemorrhagic stroke, varicose veins, other circulatory and skin conditions. Conclusions: A claims-based framework using sequential target trial emulations detected both expected and previously unrecognized ADE signals following atorvastatin initiation in older adults, offering a systematic alternative to passive surveillance that can be extended to other commonly prescribed medications. Confirmatory analyses with disaggregated outcomes and narrower comparators are underway and will be reported separately. KEY POINTS: Question: Can an active, claims-based pharmacovigilance framework using sequential target trial emulation detect adverse drug event signals among older Medicare beneficiaries who initiated atorvastatin? Findings: In this cohort study of 59,130 Medicare beneficiaries discharged after myocardial or cerebral infarction, a hypothesis-free scan across hundreds of prespecified outcomes detected five primary signals (valve disorders, sprains and strains, sensory symptoms including dizziness, abnormal laboratory findings, and prediabetes) and additional signals in sensitivity analyses (including hemorrhagic stroke), several of which align with known or labeled statin adverse effects (serving as positive controls that validate the framework) while others remain uncertain. Meaning: This framework offers a systematic, quantitative, and proactive alternative to spontaneous reporting for detecting medication safety signals in high-risk older adults. PLAIN LANGUAGE SUMMARY: Older adults use more prescription drugs than any other age group, yet they are often left out of clinical trials conducted before a drug is approved. As a result, the safety of many medications in older patients is poorly understood until the drugs are already in wide use. The main system for detecting new drug safety problems in the United States, the FDA Adverse Event Reporting System, relies on voluntary reports and captures only a small fraction of harms, which limits its usefulness. Using Medicare reimbursement records, the researchers developed an active monitoring method and tested it on a commonly used cholesterol-lowering medication (atorvastatin) – that is frequently prescribed after a heart attack or stroke. They followed thousands of patients for six months, statistically balancing those who started atorvastatin with those who started a different new drug, and screened for hundreds of possible side effects. Evidence of potential medication harm emerged along a clear range of how likely the drug caused each problem. This ranged from well-known side effects (such as diabetes and sprains and strains) and outcomes already noted in the product labeling or prior research (such as hemorrhagic stroke and dizziness) to more uncertain associations (such as cardiac valve disorders) that are difficult to interpret. Overall, atorvastatin appeared largely safe in this older population, and the same method can now be used to study other widely used drugs.
- Open Access
Can Predictive Modeling Inform the Selection of Time Zero for Target Trial Emulations? An Empirical Study of Atorvastatin Initiation in Medicare Beneficiaries
Purpose: When emulating trials of medication initiation using real-world data, there may be ambiguity regarding the most suitable time zero event for the research question of interest. The time zero event must be strongly associated with the clinical indication for treatment, confer a reasonably high probability of actual treatment initiation, and be measurable with sufficient temporal precision in the source data. When it is uncertain whether a candidate event will satisfy these three conditions simultaneously, empirical identification of predictors of medication initiation can provide valuable guidance. The objective of this study was to empirically identify predictors of incident atorvastatin initiation to inform the definition of time zero for future target trial emulations. Methods: A retrospective cohort study was conducted using Medicare claims data from January 1, 2018 – December 31, 2019 (study period). The study population included beneficiaries: age ≥ 65 years with ≥ 12 months of continuous enrollment and no statin medication for ≥ 12 months as of the study period start date. The index date was the date of the first pharmacy dispensing claim for a newly initiated medication during the study period, and classified as atorvastatin or non-atorvastatin (non-atorvastatin initiators randomly sampled at 10%). Candidate predictor variables were ascertained within 6 months pre-index and included demographics, comorbidities, healthcare utilization, and pharmacotherapy. A prespecified eight-step procedure was used to identify significant, independent predictors of incident atorvastatin initiation (vs initiating a medication other than atorvastatin). Results: The study cohort comprised 117,069 incident atorvastatin initiators and 393,083 non-atorvastatin initiators. The strongest predictors of atorvastatin initiation were having an inpatient admission for cerebral infarction (OR 11.29, 95% CI 10.47-12.17) and myocardial infarction (OR 6.56, 95% CI 5.92–7.27). For example, a White male with a recent hospitalization for cerebral infarction had a predicted probability of atorvastatin initiation of 69% (95% CI 68-70). Conclusion: The empirically identified predictors of atorvastatin initiation among statin naïve Medicare beneficiaries align with ACC/AHA guidelines recommending prompt statin therapy for secondary prevention. These predictors satisfy the three key requirements for a valid time zero event and should mitigate channeling bias and residual confounding in future target trial emulations. KEY POINTS: Findings : Acute myocardial infarction and cerebral infarction recorded during an inpatient admission were the strongest predictors of incident atorvastatin initiation among statin-naïve Medicare beneficiaries age 65 years and older. Clinical Context: These findings align with current American College of Cardiology/American Heart Association guidelines that recommend prompt statin therapy for secondary prevention after these acute cardiovascular events. Implications for Future Research: Anchoring the time zero event to an inpatient admission for myocardial/cerebral infarction satisfies the three key requirements for a valid time zero event when studying medication initiation: it is strongly associated with the clinical indication for treatment, carries a high probability of actual statin initiation, and can be identified with sufficient temporal precision in administrative data. This approach should reduce channeling bias, selection bias (e.g., immortal time bias) and residual confounding in future target trial emulations. Broader Significance: The study provides an empirically derived, high-probability time zero event that can strengthen future target trial emulations using real-world data to assess the safety of commonly used medicines in older adults, a population often underrepresented in randomized trials to obtain regulatory approval. PLAIN LANGUAGE SUMMARY: This study aimed to identify a clear starting point for future research on the safety of atorvastatin in older adults. Using Medicare claims data from 2018–2019, researchers examined more than 1.3 million beneficiaries age 65 and older who had not previously taken statins in the last year. They developed a predictive model to determine which patient characteristics were most strongly linked to starting atorvastatin. The strongest predictors were a recent hospital admission for heart attack (myocardial infarction) or stroke (cerebral infarction). These events were associated with a much higher chance of promptly receiving atorvastatin, which aligns with American College of Cardiology and American Heart Association guidelines recommending statin therapy soon after such events for secondary prevention. By using hospital discharge after these acute events as the starting point for future studies, researchers can create comparisons that reduce bias and allow more reliable estimates of atorvastatin’s effects on potential harms in this vulnerable elderly population.
Open AccessCan Predictive Modeling Inform the Selection of Time Zero for Target Trial Emulations? An Empirical Study of Atorvastatin Initiation in Medicare Beneficiaries
Purpose: When emulating trials of medication initiation using real-world data, there may be ambiguity regarding the most suitable time zero event for the research question of interest. The time zero event must be strongly associated with the clinical indication for treatment, confer a reasonably high probability of actual treatment initiation, and be measurable with sufficient temporal precision in the source data. When it is uncertain whether a candidate event will satisfy these three conditions simultaneously, empirical identification of predictors of medication initiation can provide valuable guidance. The objective of this study was to empirically identify predictors of incident atorvastatin initiation to inform the definition of time zero for future target trial emulations. Methods: A retrospective cohort study was conducted using Medicare claims data from January 1, 2018 – December 31, 2019 (study period). The study population included beneficiaries: age ≥ 65 years with ≥ 12 months of continuous enrollment and no statin medication for ≥ 12 months as of the study period start date. The index date was the date of the first pharmacy dispensing claim for a newly initiated medication during the study period, and classified as atorvastatin or non-atorvastatin (non-atorvastatin initiators randomly sampled at 10%). Candidate predictor variables were ascertained within 6 months pre-index and included demographics, comorbidities, healthcare utilization, and pharmacotherapy. A prespecified eight-step procedure was used to identify significant, independent predictors of incident atorvastatin initiation (vs initiating a medication other than atorvastatin). Results: The study cohort comprised 117,069 incident atorvastatin initiators and 393,083 non-atorvastatin initiators. The strongest predictors of atorvastatin initiation were having an inpatient admission for cerebral infarction (OR 11.29, 95% CI 10.47-12.17) and myocardial infarction (OR 6.56, 95% CI 5.92–7.27). For example, a White male with a recent hospitalization for cerebral infarction had a predicted probability of atorvastatin initiation of 69% (95% CI 68-70). Conclusion: The empirically identified predictors of atorvastatin initiation among statin naïve Medicare beneficiaries align with ACC/AHA guidelines recommending prompt statin therapy for secondary prevention. These predictors satisfy the three key requirements for a valid time zero event and should mitigate channeling bias and residual confounding in future target trial emulations. KEY POINTS: Findings : Acute myocardial infarction and cerebral infarction recorded during an inpatient admission were the strongest predictors of incident atorvastatin initiation among statin-naïve Medicare beneficiaries age 65 years and older. Clinical Context: These findings align with current American College of Cardiology/American Heart Association guidelines that recommend prompt statin therapy for secondary prevention after these acute cardiovascular events. Implications for Future Research: Anchoring the time zero event to an inpatient admission for myocardial/cerebral infarction satisfies the three key requirements for a valid time zero event when studying medication initiation: it is strongly associated with the clinical indication for treatment, carries a high probability of actual statin initiation, and can be identified with sufficient temporal precision in administrative data. This approach should reduce channeling bias, selection bias (e.g., immortal time bias) and residual confounding in future target trial emulations. Broader Significance: The study provides an empirically derived, high-probability time zero event that can strengthen future target trial emulations using real-world data to assess the safety of commonly used medicines in older adults, a population often underrepresented in randomized trials to obtain regulatory approval. PLAIN LANGUAGE SUMMARY: This study aimed to identify a clear starting point for future research on the safety of atorvastatin in older adults. Using Medicare claims data from 2018–2019, researchers examined more than 1.3 million beneficiaries age 65 and older who had not previously taken statins in the last year. They developed a predictive model to determine which patient characteristics were most strongly linked to starting atorvastatin. The strongest predictors were a recent hospital admission for heart attack (myocardial infarction) or stroke (cerebral infarction). These events were associated with a much higher chance of promptly receiving atorvastatin, which aligns with American College of Cardiology and American Heart Association guidelines recommending statin therapy soon after such events for secondary prevention. By using hospital discharge after these acute events as the starting point for future studies, researchers can create comparisons that reduce bias and allow more reliable estimates of atorvastatin’s effects on potential harms in this vulnerable elderly population.
- Open Access
- 2 February 2026
- Published by Elsevier in eClinicalMedicine
- Vol. 92
Minimally invasive surgical resection reduces one-year mortality, especially in high-risk colon cancer patients: an emulated trial
- 2 February 2026
- Published by Elsevier in eClinicalMedicine
- Vol. 92
Open AccessMinimally invasive surgical resection reduces one-year mortality, especially in high-risk colon cancer patients: an emulated trial
- Open Access
- 2 February 2026
- Published by Elsevier in The Lancet Oncology
- Vol. 27 (2) , 212-222
Estimated effect of correcting inequalities in minimally invasive surgical resection in patients with colon cancer in England: a population-based study
- 2 February 2026
- Published by Elsevier in The Lancet Oncology
- Vol. 27 (2) , 212-222
Open AccessEstimated effect of correcting inequalities in minimally invasive surgical resection in patients with colon cancer in England: a population-based study
- Open Access
- 17 July 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (15-17) , e70185
Performance of Cross‐Validated Targeted Maximum Likelihood Estimation
Background: Advanced methods for causal inference, such as targeted maximum likelihood estimation (TMLE), require specific convergence rates and the Donsker class condition for valid statistical estimation and inference. In situations where there is no differentiability due to data sparsity or near‐positivity violations, the Donsker class condition is violated. In such instances, the bias of the targeted estimand is inflated, and its variance is anti‐conservative, leading to poor coverage. Cross‐validation of the TMLE algorithm (CVTMLE) is a straightforward, yet effective way to ensure efficiency, especially in settings where the Donsker class condition is violated, such as random or near‐positivity violations. We aim to investigate the performance of CVTMLE compared to TMLE in various settings. Methods: We utilized the data‐generating mechanism described in Leger et al. (2022) to run a Monte Carlo experiment under different Donsker class violations. Then, we evaluated the respective statistical performances of TMLE and CVTMLE with different super learner libraries, with and without regression tree methods. Results: We found that CVTMLE vastly improves confidence interval coverage without adversely affecting bias, particularly in settings with small sample sizes and near‐positivity violations. Furthermore, incorporating regression trees using standard TMLE with ensemble super learner‐based initial estimates increases bias and reduces variance, leading to invalid statistical inference. Conclusions: We show through simulations that CVTMLE is much less sensitive to the choice of the super learner library and thereby provides better estimation and inference in cases where the super learner library uses more flexible candidates and is prone to overfitting.
- 17 July 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (15-17) , e70185
Open AccessPerformance of Cross‐Validated Targeted Maximum Likelihood Estimation
Background: Advanced methods for causal inference, such as targeted maximum likelihood estimation (TMLE), require specific convergence rates and the Donsker class condition for valid statistical estimation and inference. In situations where there is no differentiability due to data sparsity or near‐positivity violations, the Donsker class condition is violated. In such instances, the bias of the targeted estimand is inflated, and its variance is anti‐conservative, leading to poor coverage. Cross‐validation of the TMLE algorithm (CVTMLE) is a straightforward, yet effective way to ensure efficiency, especially in settings where the Donsker class condition is violated, such as random or near‐positivity violations. We aim to investigate the performance of CVTMLE compared to TMLE in various settings. Methods: We utilized the data‐generating mechanism described in Leger et al. (2022) to run a Monte Carlo experiment under different Donsker class violations. Then, we evaluated the respective statistical performances of TMLE and CVTMLE with different super learner libraries, with and without regression tree methods. Results: We found that CVTMLE vastly improves confidence interval coverage without adversely affecting bias, particularly in settings with small sample sizes and near‐positivity violations. Furthermore, incorporating regression trees using standard TMLE with ensemble super learner‐based initial estimates increases bias and reduces variance, leading to invalid statistical inference. Conclusions: We show through simulations that CVTMLE is much less sensitive to the choice of the super learner library and thereby provides better estimation and inference in cases where the super learner library uses more flexible candidates and is prone to overfitting.
- Open Access
- 26 May 2025
- Published by Wiley in International Journal of Cancer
- Vol. 157 (6) , 1043-1054
Characteristics of interventions aimed at reducing inequalities along the cancer continuum: A scoping review
Cancer inequalities are wide and enduring, within countries between socio‐demographic groups and between countries. These are generated and sustained throughout the key phases of the cancer pathway, from investigation, clinical assessment, decision and access to treatment, and follow‐up care. We aimed to describe the characteristics of implemented interventions, evaluated in published controlled experiments in the medical literature, specifically designed to target reductions in inequalities along the cancer pathway. We searched the Ovid Medline and Embase databases from January 2005 to April 2024 for controlled experiments reporting on interventions tackling inequalities. We extracted information on the publication, the aim and type of intervention, its setting, the characteristics of the sample and of the interventions, and summarised their results and limitations. We identified 56 articles reporting on 57 interventions. Of these, 51 (89.5%) focused on access to screening; 56 (98.2%) focused on colorectal, breast, and cervical cancers; 37 (64.9%) concentrated on ethnic inequalities and 48 (84.2%) were based in the USA. In addition, the majority of interventions sought to change individual knowledge, beliefs, and behaviour rather than issues at the system‐level. The importance of addressing how healthcare is delivered equitably to all individuals is widely recognised, and there is evidence that individual factors account for only a small part of cancer pathway inequalities. Yet, this scoping review reports a lack of diversity in the implementation of interventions addressing cancer inequalities, and a minority of them target health system issues.
- 26 May 2025
- Published by Wiley in International Journal of Cancer
- Vol. 157 (6) , 1043-1054
Open AccessCharacteristics of interventions aimed at reducing inequalities along the cancer continuum: A scoping review
Cancer inequalities are wide and enduring, within countries between socio‐demographic groups and between countries. These are generated and sustained throughout the key phases of the cancer pathway, from investigation, clinical assessment, decision and access to treatment, and follow‐up care. We aimed to describe the characteristics of implemented interventions, evaluated in published controlled experiments in the medical literature, specifically designed to target reductions in inequalities along the cancer pathway. We searched the Ovid Medline and Embase databases from January 2005 to April 2024 for controlled experiments reporting on interventions tackling inequalities. We extracted information on the publication, the aim and type of intervention, its setting, the characteristics of the sample and of the interventions, and summarised their results and limitations. We identified 56 articles reporting on 57 interventions. Of these, 51 (89.5%) focused on access to screening; 56 (98.2%) focused on colorectal, breast, and cervical cancers; 37 (64.9%) concentrated on ethnic inequalities and 48 (84.2%) were based in the USA. In addition, the majority of interventions sought to change individual knowledge, beliefs, and behaviour rather than issues at the system‐level. The importance of addressing how healthcare is delivered equitably to all individuals is widely recognised, and there is evidence that individual factors account for only a small part of cancer pathway inequalities. Yet, this scoping review reports a lack of diversity in the implementation of interventions addressing cancer inequalities, and a minority of them target health system issues.
- Open Access
Relative Survival Analysis Using Bayesian Decision Tree Ensembles
In cancer epidemiology, the \emph{relative survival framework} is used to quantify the hazard associated with cancer by comparing the all-cause mortality hazard in cancer patients to that of the general population. This framework assumes that an individual's hazard function is the sum of a known population hazard and an excess hazard associated with the cancer. Several estimands are derived from the excess hazard, including the \emph{net survival}, which are used to inform decisions and to assess the effectiveness of interventions on cancer management. In this paper, we introduce a Bayesian machine learning approach to estimating the excess hazard and identifying vulnerable subgroups, with a higher excess risk, using Bayesian additive regression trees (BART). We first develop a proportional hazards extension of the BART model to the relative survival setting, and then extend this model to non-proportional hazards. We develop tools for model interpretation and posterior summarization and then present an application using colon cancer data from England, highlighting the insights our proposed methodology offers when paired with state-of-the-art data linkage methods. This application demonstrates how these methods can be used to identify drivers of inequalities in cancer survival through variable importance quantification.
Open AccessRelative Survival Analysis Using Bayesian Decision Tree Ensembles
In cancer epidemiology, the \emph{relative survival framework} is used to quantify the hazard associated with cancer by comparing the all-cause mortality hazard in cancer patients to that of the general population. This framework assumes that an individual's hazard function is the sum of a known population hazard and an excess hazard associated with the cancer. Several estimands are derived from the excess hazard, including the \emph{net survival}, which are used to inform decisions and to assess the effectiveness of interventions on cancer management. In this paper, we introduce a Bayesian machine learning approach to estimating the excess hazard and identifying vulnerable subgroups, with a higher excess risk, using Bayesian additive regression trees (BART). We first develop a proportional hazards extension of the BART model to the relative survival setting, and then extend this model to non-proportional hazards. We develop tools for model interpretation and posterior summarization and then present an application using colon cancer data from England, highlighting the insights our proposed methodology offers when paired with state-of-the-art data linkage methods. This application demonstrates how these methods can be used to identify drivers of inequalities in cancer survival through variable importance quantification.
- Open Access
- 14 July 2024
- Published by BMJ in Family Medicine and Community Health
- Vol. 12 (3) , e002834-11
Effectiveness of post-COVID-19 primary care attendance in improving survival in very old patients with multimorbidity: a territory-wide target trial emulation
Objectives: Older individuals with multimorbidity are at an elevated risk of infection and complications from COVID-19. Effectiveness of post-COVID-19 interventions or care models in reducing subsequent adverse outcomes in these individuals have rarely been examined. This study aims to examine the effectiveness of attending general outpatient within 30 days after discharge from COVID-19 on 1-year survival among older adults aged 85 years or above with multimorbidity.Design: Retrospective cohort study emulating a randomised target trial using electronic health records.Setting: We used data from the Hospital Authority and the Department of Health in Hong Kong, which provided comprehensive electronic health records, COVID-19 confirmed case data, population-based vaccination records and other individual characteristics for the study.Participants: Adults aged 85 years or above with multimorbidity who were discharged after hospitalisation for COVID-19 between January 2020 and August 2022.Interventions: Attending a general outpatient within 30 days of last COVID-19 discharge defined the exposure, compared to no outpatient visit.Main outcome measures: Primary outcome was all-cause mortality within one year. Secondary outcomes included mortality from respiratory, cardiovascular and cancer causes.Results: A total of 6183 eligible COVID-19 survivors were included in the analysis. The all-cause mortality rate following COVID-19 hospitalisation was lower in the general outpatient visit group (17.1 deaths per 100 person-year) compared with non-visit group (42.8 deaths per 100 person-year). After adjustment, primary care consultations within 30 days after discharge were associated with a significantly greater 1-year survival (difference in 1-year survival: 11.2%, 95% CI 8.1% to 14.4%). We also observed significantly better survival from respiratory diseases in the general outpatient visit group (difference in 1-year survival: 6.3%, 95% CI 3.5% to 8.9%). In a sensitivity analysis for different grace period lengths, we found that the earlier participants had a general outpatient visit after COVID-19 discharge, the better the survival.Conclusions: Timely primary care consultations after COVID-19 hospitalisation may improve survival following COVID-19 hospitalisation among older adults aged 85 or above with multimorbidity. Expanding primary care services and implementing follow-up mechanisms are crucial to support this vulnerable population’s recovery and well-being.
- 14 July 2024
- Published by BMJ in Family Medicine and Community Health
- Vol. 12 (3) , e002834-11
Open AccessEffectiveness of post-COVID-19 primary care attendance in improving survival in very old patients with multimorbidity: a territory-wide target trial emulation
Objectives: Older individuals with multimorbidity are at an elevated risk of infection and complications from COVID-19. Effectiveness of post-COVID-19 interventions or care models in reducing subsequent adverse outcomes in these individuals have rarely been examined. This study aims to examine the effectiveness of attending general outpatient within 30 days after discharge from COVID-19 on 1-year survival among older adults aged 85 years or above with multimorbidity.Design: Retrospective cohort study emulating a randomised target trial using electronic health records.Setting: We used data from the Hospital Authority and the Department of Health in Hong Kong, which provided comprehensive electronic health records, COVID-19 confirmed case data, population-based vaccination records and other individual characteristics for the study.Participants: Adults aged 85 years or above with multimorbidity who were discharged after hospitalisation for COVID-19 between January 2020 and August 2022.Interventions: Attending a general outpatient within 30 days of last COVID-19 discharge defined the exposure, compared to no outpatient visit.Main outcome measures: Primary outcome was all-cause mortality within one year. Secondary outcomes included mortality from respiratory, cardiovascular and cancer causes.Results: A total of 6183 eligible COVID-19 survivors were included in the analysis. The all-cause mortality rate following COVID-19 hospitalisation was lower in the general outpatient visit group (17.1 deaths per 100 person-year) compared with non-visit group (42.8 deaths per 100 person-year). After adjustment, primary care consultations within 30 days after discharge were associated with a significantly greater 1-year survival (difference in 1-year survival: 11.2%, 95% CI 8.1% to 14.4%). We also observed significantly better survival from respiratory diseases in the general outpatient visit group (difference in 1-year survival: 6.3%, 95% CI 3.5% to 8.9%). In a sensitivity analysis for different grace period lengths, we found that the earlier participants had a general outpatient visit after COVID-19 discharge, the better the survival.Conclusions: Timely primary care consultations after COVID-19 hospitalisation may improve survival following COVID-19 hospitalisation among older adults aged 85 or above with multimorbidity. Expanding primary care services and implementing follow-up mechanisms are crucial to support this vulnerable population’s recovery and well-being.