Using leave-one-out cross validation (LOO) in a multilevel regression and poststratification (MRP) workflow: A cautionary tale.

In recent decades, multilevel regression and poststratification (MRP) has surged in popularity for population inference. However, the validity of the estimates can depend on details of the model, and there is currently little research on validation. We explore how leave-one-out cross validation (LOO) can be used to compare Bayesian models for MRP. We investigate two approximate calculations of LOO: Pareto smoothed importance sampling (PSIS-LOO) and a survey-weighted alternative (WTD-PSIS-LOO). Using two simulation designs, we examine how accurately these two criteria recover the correct ordering of model goodness at predicting population and small-area estimands. Focusing first on variable selection, we find that neither PSIS-LOO nor WTD-PSIS-LOO correctly recovers the models' order for an MRP population estimand, although both criteria correctly identify the best and worst model. When considering small-area estimation, the best model differs for different small areas, highlighting the complexity of MRP validation. When considering different priors, the models' order seems slightly better at smaller-area levels. These findings suggest that, while not terrible, PSIS-LOO-based ranking techniques may not be suitable to evaluate MRP as a method. We suggest this is due to the aggregation stage of MRP, where individual-level prediction errors average out. We validate these results by applying to the real world National Health and Nutrition Examination Survey (NHANES) data in the United States. Altogether, these results show that PSIS-LOO-based model validation tools need to be used with caution and might not convey the full story when validating MRP as a method.

A nonparametric relative treatment effect for direct comparisons of censored paired survival outcomes.

A frequently addressed issue in clinical trials is the comparison of censored paired survival outcomes, for example, when individuals were matched based on their characteristics prior to the analysis. In this regard, a proper incorporation of the dependence structure of the paired censored outcomes is required and, up to now, appropriate methods are only rarely available in the literature. Moreover, existing methods are not motivated by the strive for insights by means of an easy-to-interpret parameter. Hence, we seek to develop a new estimand-driven method to compare the effectiveness of two treatments in the context of right-censored survival data with matched pairs. With the help of competing risks techniques, the so-called relative treatment effect is estimated. This estimand describes the probability that individuals under Treatment 1 have a longer lifetime than comparable individuals under Treatment 2. We derive hypothesis tests and confidence intervals based on a studentized version of the estimator, where resampling-based inference is established by means of a randomization method. In a simulation study, we demonstrate for numerous sample sizes and different amounts of censoring that the developed test exhibits a good power. Finally, we apply the methodology to a well-known benchmark data set from a trial with patients suffering from diabetic retinopathy.

End points in clinical trials in diffuse large B-cell lymphoma: time for more dialogue?

We observed lack of clarity and consistency in end point definitions of large randomized clinical trials in diffuse large B-cell lymphoma. These inconsistencies are such that trials might, in fact, address different clinical questions. They complicate interpretation of results, including comparisons across studies. Problems arise from different ways to account for events occurring after randomization including absence of improvement in disease status, treatment discontinuation or the initiation of new therapy. We call for more dialogue between stakeholders to define with clarity the questions of interest and corresponding end points. We illustrate that assessing different end point rules across a range of plausible patient journeys can be a powerful tool to facilitate such a discussion and contribute to better understanding of patient-relevant end points.

What is this article about? This article talks about the lack of clarity and consistency in the definitions of outcomes used in clinical trials that investigate new treatments for diffuse large B-cell lymphoma. This is mainly due to how these different outcome definitions handle events such as absence of improvement in disease status, treatment discontinuation or initiation of new treatment. The authors discuss how these inconsistencies make it hard to interpret the results of individual clinical trials and to compare results across clinical trials.Why is it important? Defining the above events and consequently defining outcomes affects what we can learn from the trials and can lead to different results. Some approaches may not reflect good and bad outcomes for patients appropriately. This makes it challenging for patients, physicians, health authorities and payors to understand the true benefit of treatments under investigation and which one is better.What are the key take-aways? This article serves as a call-to-action for more dialogue among all stakeholders involved in drug development and the decision-making process related to drug evaluations. There is an urgent need for clinical trials to be designed with more clarity and consistency on what is being measured so that relevant questions for patients and prescribing physicians are addressed. Understanding patient journeys will be key to successfully understand what truly matters to patients and how to measure the benefit of new treatments. Such discussions will contribute toward more clarity and consistency in the evaluation of new treatments.

Scaling and interpreting treatment effects in clinical trials using restricted mean survival time.

BACKGROUND: Restricted mean survival time is the expected duration of survival up to a chosen time of restriction τ. For comparison studies, the difference in restricted mean survival times between two groups provides a summary measure of the treatment effect that is free of assumptions regarding the relative shape of the two survival curves, such as proportional hazards. However, it can be difficult to judge the magnitude of the effect from a comparison of restricted means due to the truncation of observation at time τ. METHODS: In this article, we describe additional ways of expressing the treatment effect based on restricted means that can be helpful in this regard. These include the ratio of restricted means, the ratio of life-years (or time) lost, and the average integrated difference between the survival curves, equal to the difference in restricted means divided by τ. These alternative metrics are straightforward to calculate and provide a means for scaling the effect size as an aid to interpretation. Examples from two randomized, multicenter clinical trials in prostate cancer, NRG/RTOG 0521 and NRG/RTOG 0534, with primary endpoints of overall survival and biochemical/radiological progression-free survival, respectively, are presented to illustrate the ideas. RESULTS: The four effect measures (restricted mean survival time difference, restricted mean survival time ratio, time lost ratio, and average survival rate difference) were 0.45 years, 1.05, 0.81, and 0.038 for RTOG 0521 and 1.36 years, 1.17, 0.56, and 0.12 for RTOG 0534 with τ = 12 and 11 years, respectively. Thus, for example, the 0.45-year difference in the first trial translates into a 19% reduction in time lost and a 3.8% average absolute difference between the survival curves over the 12-year horizon, a modest effect size, whereas the 1.36-year difference in the second trial corresponds to a 44% reduction in time lost and a 12% absolute survival difference, a rather large effect. CONCLUSIONS: In addition to the difference in restricted mean survival times, these alternative measures can be helpful in determining whether the magnitude of the treatment effect is clinically meaningful.

A safety estimand for late phase clinical trials where the analysis period varies over the subjects.

BACKGROUND/AIMS: Evaluating safety is as important as evaluating efficacy in a clinical trial, yet the tradition for safety analysis is rudimentary. This article explores more complex methodologies for safety evaluation, with the aim of improving the interpretability, as well as generalizability, of the results. METHODS: For studies where the analysis periods vary over the subjects, using the International Council for Harmonisation estimand framework, we construct a formal estimand that could be used in the setting of safety surveillance that answers the clinical question of 'What is the magnitude of the increase in risk of experiencing an adverse event if the treatment is taken, as prescribed, for a specific period of time?'. Estimation methodologies for this estimand are also discussed. RESULTS: The proposed estimand is similar to that found in the efficacy analyses of time to event data (e.g. in outcome studies), with the key difference of utilization of hypothetical intercurrent event strategy for the intercurrent event of treatment discontinuation. This is motivated by what we perceive to be a key difference for the safety objective compared to efficacy objectives, namely a desire for sensitivity (i.e. greater possibility of detecting a negative impact of the drug, if such exists) as opposed to the need to prove a positive effect of the drug in a conservative manner. CONCLUSION: It is valuable, and possible, to use the International Council for Harmonisation estimand framework not only for efficacy but also for safety evaluation, with the estimand driven by an interpretable, and relevant, clinical question.

Causal interpretation of the hazard ratio in randomized clinical trials.

BACKGROUND: Although the hazard ratio has no straightforward causal interpretation, clinical trialists commonly use it as a measure of treatment effect. METHODS: We review the definition and examples of causal estimands. We discuss the causal interpretation of the hazard ratio from a two-arm randomized clinical trial, and the implications of proportional hazards assumptions in the context of potential outcomes. We illustrate the application of these concepts in a synthetic model and in a model of the time-varying effects of COVID-19 vaccination. RESULTS: We define causal estimands as having either an individual-level or population-level interpretation. Difference-in-expectation estimands are both individual-level and population-level estimands, whereas without strong untestable assumptions the causal rate ratio and hazard ratio have only population-level interpretations. We caution users against making an incorrect individual-level interpretation, emphasizing that in general a hazard ratio does not on average change each individual's hazard by a factor. We discuss a potentially valid interpretation of the constant hazard ratio as a population-level causal effect under the proportional hazards assumption. CONCLUSION: We conclude that the population-level hazard ratio remains a useful estimand, but one must interpret it with appropriate attention to the underlying causal model. This is especially important for interpreting hazard ratios over time.

[Special issue PRO] Considering endpoints for comparative tolerability of cancer treatments using patient report given the estimand framework.

Regulatory agencies are advancing the use of systematic approaches to collect patient experience data, including patient-reported outcomes (PROs), in cancer clinical trials to inform regulatory decision-making. Due in part to clinician under-reporting of symptomatic adverse events, there is a growing recognition that evaluation of cancer treatment tolerability should include the patient experience, both in terms of the overall side effect impact and symptomatic adverse events. Methodologies around implementation, analysis, and interpretation of "patient" reported tolerability are under development, and current approaches are largely descriptive. There is robust guidance for use of PROs as efficacy endpoints to compare cancer treatments, but it is unclear to what extent this can be relied-upon to develop tolerability endpoints. An important consideration when developing endpoints to compare tolerability between treatments is the linkage of trial design, objectives, and statistical analysis. Despite interest in and frequent collection of PRO data in oncology trials, heterogeneity in analyses and unclear PRO objectives mean that design, objectives, and analysis may not be aligned, posing substantial challenges for the interpretation of results. The recent ICH E9 (R1) estimand framework represents an opportunity to help address these challenges. Efforts to apply the estimand framework in the context of PROs have primarily focused on efficacy outcomes. In this paper, we discuss considerations for comparing the patient-reported tolerability of different treatments in an oncology trial context.

Application of estimand framework to the design and analysis of multi-regional clinical trials.

With the increasing globalization of drug development and the publication of the International Council for Harmonisation (ICH) E17 guideline (ICH International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use 2017), multi-regional clinical trials (MRCTs) have become a preferred option to accelerate the availability of new medical products by design, execution and simultaneous submission under one protocol. MRCTs, with the participation of all major regions including countries from both developed and emerging markets, surely make new drug development more efficient. Even though the proposed estimand framework (ICH E9 (R1) (2019), came later in 2019 and was not mentioned in ICH E17, the application of the estimand framework has the potential to enhance the design, execution, and analysis in MRCTs. Defining an estimand within the regional context in MRCTs is an important issue that requires careful consideration. Given that consistency evaluation of treatment effects across regions is critical in MRCTs, the utilization of the estimand framework for regional consistency evaluation is also worth discussion. This paper aims to address these two questions. The five attributes of the estimand definition are discussed within a multi-regional context. It is imperative to thoroughly consider regional intrinsic/extrinsic factors when planning the estimand and estimation of MRCTs. A holistic approach is summarized to conduct consistency evaluation. When a regional inconsistency is observed, the possible reasons need to be further explored under five attributes of the estimand framework. Two real case studies are discussed to illustrate the application of the estimand framework in the consistency evaluation.

Considerations for master protocols using external controls.

There has been an increasing use of master protocols in oncology clinical trials because of its efficiency to accelerate cancer drug development and flexibility to accommodate multiple substudies. Depending on the study objective and design, a master protocol trial can be a basket trial, an umbrella trial, a platform trial, or any other form of trials in which multiple investigational products and/or subpopulations are studied under a single protocol. Master protocols can use external data and evidence (e.g. external controls) for treatment effect estimation, which can further improve efficiency of master protocol trials. This paper provides an overview of different types of external controls and their unique features when used in master protocols. Some key considerations in master protocols with external controls are discussed including construction of estimands, assessment of fit-for-use real-world data, and considerations for different types of master protocols. Similarities and differences between regular randomized controlled trials and master protocols when using external controls are discussed. A targeted learning-based causal roadmap is presented which constitutes three key steps: (1) define a target statistical estimand that aligns with the causal estimand for the study objective, (2) use an efficient estimator to estimate the target statistical estimand and its uncertainty, and (3) evaluate the impact of causal assumptions on the study conclusion by performing sensitivity analyses. Two illustrative examples for master protocols using external controls are discussed for their merits and possible improvement in causal effect estimation.

Duration of and time to response in oncology clinical trials from the perspective of the estimand framework.

Duration of response (DOR) and time to response (TTR) are typically evaluated as secondary endpoints in early-stage clinical studies in oncology when efficacy is assessed by the best overall response and presented as the overall response rate. Despite common use of DOR and TTR in particular in single-arm studies, the definition of these endpoints and the questions they are intended to answer remain unclear. Motivated by the estimand framework, we present relevant scientific questions of interest for DOR and TTR and propose corresponding estimand definitions. We elaborate on how to deal with relevant intercurrent events which should follow the same considerations as implemented for the primary response estimand. A case study in mantle cell lymphoma illustrates the implementation of relevant estimands of DOR and TTR. We close the paper with practical recommendations to implement DOR and TTR in clinical study protocols.

Current developments of the estimand concept.

Since the introduction of the estimand in therapeutical studies, several adaptions have been developed. This short article highlights the important aspects of the estimand concept. A literature research was conducted to identify different extensions to this framework. Different modified strategies for intercurrent events are presented, as well as examples of methods to implement the estimand in clinical studies. The article reflects that the estimand is an ongoing research field with further exploration.

Application of hypothetical strategies in acute pain.

Since the publication of ICH E9 (R1), "Addendum to statistical principles for clinical trials: on choosing appropriate estimands and defining sensitivity analyses in clinical trials," there has been a lot of debate about the hypothetical strategy for handling intercurrent events. Arguments against the hypothetical strategy are twofold: (1) the clinical question has limited clinical/regulatory interest; (2) the estimation may need strong statistical assumptions. In this article, we provide an example of a hypothetical strategy handling use of rescue medications in the acute pain setting. We argue that the treatment effect of a drug that is attributable to the treatment alone is the clinical question of interest and is important to regulators. The hypothetical strategy is important when developing non-opioid treatment as it estimates the treatment effect due to treatment during the pre-specified evaluation period whereas the treatment policy strategy does not. Two widely acceptable and non-controversial clinical inputs are required to construct a reasonable estimator. More importantly, this estimator does not rely on additional strong statistical assumptions and is considered reasonable for regulatory decision making. In this article, we point out examples where estimators for a hypothetical strategy can be constructed without any strong additional statistical assumptions besides acceptable clinical inputs. We also showcase a new way to obtain estimation based on disease specific clinical knowledge instead of strong statistical assumptions. In the example presented, we clearly demonstrate the advantages of the hypothetical strategy compared to alternative strategies including the treatment policy strategy and a composite variable strategy.

Why estimands are needed to define treatment effects in clinical trials.

BACKGROUND: The estimand for a clinical trial is a precise definition of the treatment effect to be estimated. Traditionally, estimates of treatment effects are based on either an ITT analysis or a per-protocol analysis. However, there are important clinical questions which are not addressed by either of these analyses. For example, consider a trial where patients take a rescue medication. The ITT analysis includes data after use of rescue, while the per-protocol analysis excludes these patients altogether. Neither of these analyses addresses the important question of what the treatment effect would have been if patients did not take rescue medication. MAIN TEXT: Trial estimands provide a broader perspective compared to the limitations of ITT and per-protocol analysis. Trial treatment effects depend on how events occurring after treatment initiation such as use of alternative medication or discontinuation of the intervention are included in the definition. These events can be accounted for in different ways, depending on the clinical question of interest. CONCLUSION: The estimand framework is an important step forward in improving the clarity and transparency of clinical trials. The centrality of estimands to clinical trials is currently not reflected in methods recommended by the Cochrane group or the CONSORT statement, the current standard for reporting clinical trials in medical journals. We encourage revisions to these guidelines.

Contrasting principal stratum and hypothetical strategy estimands in multi-period crossover trials with incomplete data.

Complete case analyses of complete crossover designs provide an opportunity to make comparisons based on patients who can tolerate all treatments. It is argued that this provides a means of estimating a principal stratum strategy estimand, something which is difficult to do in parallel group trials. While some trial users will consider this a relevant aim, others may be interested in hypothetical strategy estimands, that is, the effect that would be found if all patients completed the trial. Whether these estimands differ importantly is a question of interest to the different users of the trial results. This paper derives the difference between principal stratum strategy and hypothetical strategy estimands, where the former is estimated by a complete-case analysis of the crossover design, and a model for the dropout process is assumed. Complete crossover designs, that is, those where all treatments appear in all sequences, and which compare t treatments over p periods with respect to a continuous outcome are considered. Numerical results are presented for Williams designs with four and six periods. Results from a trial of obstructive sleep apnoea-hypopnoea (TOMADO) are also used for illustration. The results demonstrate that the percentage difference between the estimands is modest, exceeding 5% only when the trial has been severely affected by dropouts or if the within-subject correlation is low.

Interpretation of active-control randomised trials: the case for a new analytical perspective involving averted events.

Active-control trials, where an experimental treatment is compared with an established treatment, are performed when the inclusion of a placebo control group is deemed to be unethical. For time-to-event outcomes, the primary estimand is usually the rate ratio, or the closely-related hazard ratio, comparing the experimental group with the control group. In this article we describe major problems in the interpretation of this estimand, using examples from COVID-19 vaccine and HIV pre-exposure prophylaxis trials. In particular, when the control treatment is highly effective, the rate ratio may indicate that the experimental treatment is clearly statistically inferior even when it is worthwhile from a public health perspective. We argue that it is crucially important to consider averted events as well as observed events in the interpretation of active-control trials. An alternative metric that incorporates this information, the averted events ratio, is proposed and exemplified. Its interpretation is simple and conceptually appealing, namely the proportion of events that would be averted by using the experimental treatment rather than the control treatment. The averted events ratio cannot be directly estimated from the active-control trial, and requires an additional assumption about either: (a) the incidence that would have been observed in a hypothetical placebo arm (the counterfactual incidence) or (b) the efficacy of the control treatment (relative to no treatment) that pertained in the active-control trial. Although estimation of these parameters is not straightforward, this must be attempted in order to draw rational inferences. To date, this method has been applied only within HIV prevention research, but has wider applicability to treatment trials and other disease areas.

The Trial within Cohorts (TwiCs) study design in oncology: experience and methodological reflections.

A Trial within Cohorts (TwiCs) study design is a trial design that uses the infrastructure of an observational cohort study to initiate a randomized trial. Upon cohort enrollment, the participants provide consent for being randomized in future studies without being informed. Once a new treatment is available, eligible cohort participants are randomly assigned to the treatment or standard of care. Patients randomized to the treatment arm are offered the new treatment, which they can choose to refuse. Patients who refuse will receive standard of care instead. Patients randomized to the standard of care arm receive no information about the trial and continue receiving standard of care as part of the cohort study. Standard cohort measures are used for outcome comparisons. The TwiCs study design aims to overcome some issues encountered in standard Randomized Controlled Trials (RCTs). An example of an issue in standard RCTs is the slow patient accrual. A TwiCs study aims to improve this by selecting patients using a cohort and only offering the intervention to patients in the intervention arm. In oncology, the TwiCs study design has gained increasing interest during the last decade. Despite its potential advantages over RCTs, the TwiCs study design has several methodological challenges that need careful consideration when planning a TwiCs study. In this article, we focus on these challenges and reflect on them using experiences from TwiCs studies initiated in oncology. Important methodological challenges that are discussed are the timing of randomization, the issue of non-compliance (refusal) after randomization in the intervention arm, and the definition of the intention-to-treat effect in a TwiCs study and how this effect is related to its counterpart in standard RCTs.

The utility of the self-controlled study design for pharmacoepidemiological studies without an active comparator medication using a medical information database: An application to assess the risk of varenicline on cardiovascular outcomes.

PURPOSE: To illustrate the utility of the self-controlled study design for studies without an active comparator, we compared the results of a cohort design study with a non-user comparator with those of a self-controlled design study in evaluating the risk of varenicline on cardiovascular outcomes, using a Japanese medical claims database. METHODS: The participating smokers were identified from health-screening results collected between May 2008 and April 2017. Using a non-user-comparator cohort study design, we estimated the hazard ratios (HRs) and 95% confidence intervals (CIs) of varenicline on initial hospitalization with cardiovascular outcomes using Cox's model adjusted for patients' sex, age, medical history, medication history, and health-screening results. Using a self-controlled study design, the within-subject HR was estimated using a stratified Cox's model adjusted for medical history, medication history, and health-screening results. The estimate from a recent meta-analysis was considered the gold standard (risk ratio: 1.03). RESULTS: We identified 460 464 smokers (398 694 males [86.6%]; mean (standard deviation) age: 42.9 [10.8] years) in the database. Of these, 11 561 had been dispensed varenicline at least once, and 4511 had experienced cardiovascular outcomes. The estimate of the non-user-comparator cohort study design exceeded the gold standard (HR [95% CI]: 2.04 [1.22-3.42]), whereas that of the self-controlled study design was close to the gold standard (within-subject HR [95% CI]: 1.12 [0.27-4.70]). CONCLUSIONS: The self-controlled study design is useful alternative to a non-user-comparator cohort design when evaluating the risk of medications relative to their non-use, based on a medical information database.

Tying research question and analytical strategy when variables are affected by medication use.

Ill-defined research questions could be particularly problematic in an epidemiological setting where measurements fluctuate over time due to intercurrent events, such as medication use. When a research question fails to specify how medication use should be handled methodologically, arbitrary decisions may be made during the analysis phase, which likely leads to a mismatch between the intended question and the performed analysis. The mismatch can result in vastly different or meaningless interpretations of estimated effects. Thus, a research question such as "what is the effect of X on Y?" requires further elaboration, and it should consider whether and how medication use has affected the measurements of interest. In our study, we will discuss how well-defined questions can be formulated when medication use is involved in observational studies. We will distinguish between a situation where an exposure is affected by medication use and where the outcome of interest is affected by medication use. For each setting, we will give examples of different research questions that could be asked depending on how medication use is considered in the estimand and discuss methodological considerations under each question.

Tell me what you want, what you really really want: Estimands in observational pharmacoepidemiologic comparative effectiveness and safety studies.

PURPOSE: Ideally, the objectives of a pharmacoepidemiologic comparative effectiveness or safety study should dictate its design and data analysis. This paper discusses how defining an estimand is instrumental to this process. METHODS: We applied the ICH-E9 (Statistical Principles for Clinical Trials) R1 addendum on estimands - which originally focused on randomized trials - to three examples of observational pharmacoepidemiologic comparative effectiveness and safety studies. Five key elements specify the estimand: the population, contrasted treatments, endpoint, intercurrent events, and population-level summary measure. RESULTS: Different estimands were defined for case studies representing three types of pharmacological treatments: (1) single-dose treatments using a case study about the effect of influenza vaccination versus no vaccination on mortality risk in an adult population of ≥60 years of age; (2) sustained-treatments using a case study about the effect of dipeptidyl peptidase 4 inhibitor versus glucagon-like peptide-1 agonist on hypoglycemia risk in treatment of uncontrolled diabetes; and (3) as needed treatments using a case study on the effect of nitroglycerin spray as-needed versus no nitroglycerin on syncope risk in treatment of stabile angina pectoris. CONCLUSIONS: The case studies illustrated that a seemingly clear research question can still be open to multiple interpretations. Defining an estimand ensures that the study targets a treatment effect that aligns with the treatment decision the study aims to inform. Estimand definitions further help to inform choices regarding study design and data-analysis and clarify how to interpret study findings.

Evaluating whether the proportional odds models to analyse ordinal outcomes in COVID-19 clinical trials is providing clinically interpretable treatment effects: A systematic review.

BACKGROUND: After an initial recommendation from the World Health Organisation, trials of patients hospitalised with COVID-19 often include an ordinal clinical status outcome, which comprises a series of ordered categorical variables, typically ranging from 'Alive and discharged from hospital' to 'Dead'. These ordinal outcomes are often analysed using a proportional odds model, which provides a common odds ratio as an overall measure of effect, which is generally interpreted as the odds ratio for being in a higher category. The common odds ratio relies on the assumption of proportional odds, which implies an identical odds ratio across all ordinal categories; however, there is generally no statistical or biological basis for which this assumption should hold; and when violated, the common odds ratio may be a biased representation of the odds ratios for particular categories within the ordinal outcome. In this study, we aimed to evaluate to what extent the common odds ratio in published COVID-19 trials differed to simple binary odds ratios for clinically important outcomes. METHODS: We conducted a systematic review of randomised trials evaluating interventions for patients hospitalised with COVID-19, which used a proportional odds model to analyse an ordinal clinical status outcome, published between January 2020 and May 2021. We assessed agreement between the common odds ratio and the odds ratio from a standard logistic regression model for three clinically important binary outcomes: 'Alive', 'Alive without mechanical ventilation', and 'Alive and discharged from hospital'. RESULTS: Sixteen randomised clinical trials, comprising 38 individual comparisons, were included in this study; of these, only 6 trials (38%) formally assessed the proportional odds assumption. The common odds ratio differed by more than 25% compared to the binary odds ratios in 55% of comparisons for the outcome 'Alive', 37% for 'Alive without mechanical ventilation', and 24% for 'Alive and discharged from hospital'. In addition, the common odds ratio systematically underestimated the odds ratio for the outcome 'Alive' by -16.8% (95% confidence interval: -28.7% to -2.9%, p = 0.02), though differences for the other outcomes were smaller and not statistically significant (-8.4% for 'Alive without mechanical ventilation' and 3.6% for 'Alive and discharged from hospital'). The common odds ratio was statistically significant for 18% of comparisons, while the binary odds ratio was significant in 5%, 16%, and 3% of comparisons for the outcomes 'Alive', 'Alive without mechanical ventilation', and 'Alive and discharged from hospital', respectively. CONCLUSION: The common odds ratio from proportional odds models often differs substantially to odds ratios from clinically important binary outcomes, and similar to composite outcomes, a beneficial common OR from a proportional odds model does not necessarily indicate a beneficial effect on the most important categories within the ordinal outcome.