Miguel A Hernán’s research while affiliated with Karolinska Institutet and other places

... This study has several limitations • As mentioned above, the absence of a randomized control group and use of external control will raise the possibility of confounding bias. This is a problem common to all intervention studies in rare diseases when randomization is not practically or ethically feasible [13]. • The sample size is small for a Phase 1 study. ...

Reference:

Safety and Efficacy of Vesemnogene lantuparvovec, an AAV-based Gene therapy for Spinal Muscular Atrophy in Children Less Than 24 Months of Age in Low-Middle Income Countries. First interim report as of April 2025

... 16 As such the eligibility criteria must characterise the target population for which inference is ultimately sought, and the ideal trial must also have perfect adherence, and no missing data or measurement error (e.g. the outcome measure must be the gold standard measure that would be used in practice, even in the absence of data mapping to these protocol components). This is different to a recently recommended approach to target trials, 29 but is necessary to systematically assess all potential sources of bias. 16 We operationalized the final step of the road map using a recent simultaneous biasadjustment approach. ...

Reference:

A roadmap for systematic identification and analysis of multiple biases in causal inference

... As Pepe et al. (2004) demonstrated for odds ratios in diagnostic accuracy, a measure of association such as the hazard ratio does not directly capture how well a biomarker can discriminate patients at higher risk of experiencing the event within a given time-horizon. Moreover, the proportional hazards assumption is often violated, making hazard ratios unreliable if applied by default (Stensrud and Hernàn 2025). Relying solely on this approach risks overlooking important time-varying and covariate-specific differences in prognostic accuracy. ...

Reference:

Nonparanormal Modeling Framework for Prognostic Biomarker Assessment with Application to Amyotrophic Lateral Sclerosis

... These issues, and methods that can address selection problems beyond misallocation of person time, are discussed elsewhere. [20] A challenge in applying IMMORTOOL is that study authors often do not provide full data on the timings of the intervention or endpoints. However, readers will often have some idea of the expected distribution of time to death and can test a number of plausible time to intervention distributions. ...

Reference:

Quantifying potential immortal time bias in observational studies in acute severe infection

... The National Cancer Screening Programs provided by the National Health Insurance Service (NHIS) include screening for five types of cancer (stomach, colorectum, liver, breast, and uterine cervix) but do not include prostate cancer screening [5]. The NHIS does not provide routine screening for prostate cancer because it is unclear whether prostate-specific antigen (PSA)-based screening can reduce prostate cancer-related or overall mortality and be cost-effective [6,7]. ...

Reference:

A 5-Year Mortality Prediction Model for Prostate Cancer Patients Based on the Korean Nationwide Health Insurance Claims Database

... M adenci et al. 1 used the target trial emulation framework to estimate the causal effect of bariatric surgery on the 7-year risk of cardiovascular disease (CVD). They highlighted the pitfalls in the design and analysis of previous observational studies conducted on this topic, which found overall that bariatric surgery lowers the risk of CVD. ...

Reference:

Bariatric surgery and CVD: The target trial emulation framework provides transparency in articulating the limits of observational studies

... 8,9 Benchmarking can involve anything from formal pre-specified criteria for comparison to informal post hoc deep dives into assessment of the potential impact of design differences, bias, and clinical context. [10][11][12][13] However, prior work on benchmarking and transportability of results between RCTs and database studies has required several assumptions that may be untenable in many practical settings. For example, in addition to the usual assumptions of conditional exchangeability (e.g., no unmeasured confounding) between treatment arms within the RCT and database study, consistency, and positivity, existing benchmarking and transportability methods have required an assumption that trial participation does not have an effect on the outcome other than through treatment. ...

Reference:

A Benchmark, Expand, and Calibration (BenchExCal) Trial Emulation Approach for Using Real-World Evidence to Support Indication Expansions: Design and Process for a Planned Empirical Evaluation

... Currently, the treatment of SCZ is primarily based on a multifaceted and comprehensive strategy that aims to alleviate symptoms, enhance patients' quality of life, and reduce relapse rates. Common first-line treatments include first-generation antipsychotics such as chlorpromazine and sulpiride, as well as second-and thirdgeneration medications such as clozapine, risperidone, aripiprazole, and lurasidone, which are less likely to cause extrapyramidal reactions (most common extrapyramidal adverse effects include acute dystonia, akathisia, parkinsonism, and tardive dyskinesia) (Leucht et al., 2023;Szmulewicz et al., 2024). In addition, psychotherapeutic approaches such as cognitive-behavioral therapy and family therapy are also important components of treatment (O'Driscoll et al., 2014;Daruvala et al., 2021). ...

Reference:

Aberrant glycosylation in schizophrenia: insights into pathophysiological mechanisms and therapeutic potentials

... • Gathered data dictionaries and/or other documentation (when available) • Met with data vendors/providers to understand essential information such as data provenance and completeness of fields needed for most research questions (e.g., age) • Compiled key information for each data source and organized data characterization into newly developed standard template ( • Used existing frameworks and templates to: ○ Guide design choices and rationale, and identify minimal criteria for real-world data to meet the needs of the study [14,26,27] ○ Illustrate initial study design and key assessment windows (e.g., baseline, exposure, follow-up) [37] ○ Identify potential confounders (comparative studies only) [28,29] ○ Identify fit-for-purpose data source(s) (i.e., reliable and relevant) to meet the needs of a specific research question [15] ○ For efficiency, in future would use newly published template with combined study design and data fitness assessment steps [34] (Key Lesson #5) Explore selected data source(s) • While carefully avoiding connecting treatments and outcome of inter est (e.g., outcome blinded,) explored data and documented findings to maintain objectivity, but allowed evidence based protocol decisions such as: ○ Completeness and general trends of key variables (e.g., inclusion/ exclusion criteria, subgroups, primary outcome) ○ Considered alternate definitions/time windows for key variables (e.g., confirmed COVID-19 or mortality) ○ Compared general trends of key variables to internal and external benchmarks (as available) to inform final variable definitions and time windows ○ Identified potential threats to validity that warrant analysis considerations, should be further explored via planned sensitivity analyses, and/or should be documented as potential limitations of the data source ○ Data explorations, maintaining blinding, were necessary to yield valid and interpretable study findings. In future, would consider also applying newly published process that incorporates diagnostic steps as a standard part of protocol design [36] (Key Lesson #6) Develop, finalize, and post protocol ...

Reference:

Increasing the Utility of Real-World Data to Inform Public Health Decision Making Through a US-based Private–Public Partnership: 10 Lessons Learned from a Principled Approach to Rapid Pandemic RWE Generation

... We know of no relevant framework that would have been appropriate to implement in order to test specific hypotheses within the biomedical community. Further, null hypothesis testing is not appropriate for determining beliefs and when hypothesis testing is used you need a formal sample size calculation [21,22]. ...

Reference:

Biomedical researchers’ perspectives on the reproducibility of research