Joint models inform the longitudinal assessment of patient-reported outcomes in clinical trials: a simulation study and secondary analysis of the restrictive Vs. liberal fluid therapy for major abdominal surgery (RELIEF) randomized controlled trial
Bibliographic record
Abstract
OBJECTIVES: Evaluate the utility of a joint model when analysing a patient-reported endpoint as part of a randomized controlled trial (RCT) in which censoring occurs when patients die during follow-up. STUDY DESIGN AND SETTING: The present study comprises two parts as follows: first we reanalyzed data from a previously published RCT comparing two fluid regimens in the first 24 hours of major abdomino-pelvic surgery ('Restrictive versus Liberal Fluid Therapy for Major Abdominal Surgery [RELIEF]' trial). In this trial, patient-reported disability was measured at multiple timepoints before and after surgery. Next, we conducted a simulation study to jointly emulate patient-reported disability and survival, similar to the RELIEF trial, under nine treatment-outcome scenarios. In both parts, we compared a joint model analysis to a linear mixed-effect model combined with one of the several traditional methods of handling longitudinal missingness as follows: available data analysis, complete case analysis, last observation carried forward, and worst-case assumption. RESULTS: In part one, the joint model revealed no between-group differences in patient-reported disability at 1, 3, 6, and 12 months after surgery. The worst-case approach consistently resulted in the largest deviation from the joint model estimates, although in this particular setting none of the approaches materially changed the study's conclusions. In part two, the simulations revealed that across all treatment-outcome scenarios, the joint model expectedly produced unbiased estimates of patient-reported disability. Similarly, employing an approach based on all available data (ie, relying on the maximum likelihood estimator for handling missingness) yielded disability estimates close to the simulated values, albeit with slight bias across some scenarios. The last observation carried forward approach that mirrored the joint model's estimates except when the treatment had a nonnull effect on patient-reported disability. The worst-case analysis resulted in high bias, which was particularly evident when the treatment had a large effect on survival. The complete case analysis resulted in high bias across all scenarios. CONCLUSION: In randomized trials that employ a patient-reported outcome as one of their endpoints, a joint model can address bias arising from informative missingness related to death. Methods for handling missingness based on all available data appear to be a reasonable alternative to joint models, with only slight bias across some simulated scenarios. PLAIN LANGUAGE SUMMARY: 'Patient-centered research' focuses on outcomes that are prioritized by patients. This approach often involves asking patients to complete questionnaires about their health experiences. However, if a patient does not finish a study, dealing with their missing answers can pose significant challenges. Joint models are a recent statistical method that may help address this issue. In this study, we used joint models in a real-world clinical trial, and in a series of simulated trials, to determine how well they handle missing questionnaire data from patients. We found that joint models offer significant benefits over most traditional methods used to analyze clinical trials.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.233 | 0.328 |
| Meta-epidemiology (narrow) | 0.003 | 0.002 |
| Meta-epidemiology (broad) | 0.007 | 0.010 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.005 | 0.005 |
| Open science | 0.004 | 0.004 |
| Research integrity | 0.005 | 0.007 |
| Insufficient payload (model declined to judge) | 0.009 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".