Joint modeling of multiple longitudinal patient-reported outcomes and survival

Our plain-English summary of Joint modeling of multiple longitudinal patient-reported outcomes and survival, by L.A. Hatfield, M.E. Boye, B.P. Carlin (Journal of Biopharmaceutical Statistics, 2011).

The study

• Authors: L.A. Hatfield, M.E. Boye, B.P. Carlin
• Journal: Journal of Biopharmaceutical Statistics
• Published: 2011;21(5):971-91
• Cited by: 30 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Researchers often include patient-reported outcomes (PROs) in Phase III clinical trials to demonstrate the value of treatment from the patient's perspective. These data are collected as longitudinal repeated measures and are often censored by occurrence of a clinical event that defines a survival time. Hierarchical Bayesian models having latent individual-level trajectories provide a flexible approach to modeling such multiple outcome types simultaneously. We consider the case of many zeros in the longitudinal data motivating a mixture model, and demonstrate several approaches to modeling multiple longitudinal PROs with survival in a cancer clinical trial. These joint models may enhance Phase III analyses and better inform health care decision makers.

In plain English

This was a statistics methods paper, not a study of asbestos exposure. The authors built Bayesian 'joint models' that analyze patients' repeated symptom ratings together with progression-free survival, meaning the time until the cancer worsened or the patient died. They tested the models on data from a Phase III trial of 243 patients with locally advanced or metastatic malignant pleural mesothelioma. That trial compared pemetrexed plus best supportive care with best supportive care alone as second-line therapy. The authors modeled two symptoms from the Lung Cancer Symptom Scale, cough and dyspnea (shortness of breath). Of the model versions compared, they favored two, called UV2I and MV2KI. In most models, treatment effects on symptoms were not statistically different from zero. Only the single-symptom models showed trends toward a treatment benefit on progression-free survival, and those intervals narrowly included no effect. The authors concluded that the data offered little support for their expectation that treatment would reduce symptoms in a way linked to longer progression-free survival. They suggested this might reflect the small sample and the gaps of about 21 days between assessments.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Statistical methods paper developing hierarchical Bayesian joint models for repeated patient-reported symptoms and progression-free survival, demonstrated on data from an existing Phase III clinical trial
• Who was studied: 243 patients with locally advanced or metastatic malignant pleural mesothelioma in a second-line therapy trial (Jassem et al., 2008). Each patient contributed 1 to 14 symptom assessments (median 3), and 7% of survival times were censored. Only the LCSS cough and dyspnea items were modeled.
• Exposure: Asbestos exposure was not studied. The trial compared pemetrexed plus best supportive care with best supportive care alone.
• Main results:
  • Based on the model comparison statistics, the authors favored model UV2I (one symptom at a time) and model MV2KI (both symptoms together) for these data.
  • For the probability of reporting a nonzero symptom score, treatment effects in most models were not significantly different from zero, although they trended toward improved symptoms in all models. The treatment-by-day effect reached significance for cough in the MV2KI model and for both symptoms in the MV2I model.
  • For progression-free survival, only the univariate (single-symptom) models showed trends toward a treatment benefit, and their credible intervals narrowly included zero (no effect).
  • In the MV2KI model, each patient's tendency to report a symptom and that symptom's severity were strongly positively correlated: 0.61 (CI .36 to .80) for cough and 0.70 (CI .51 to .83) for dyspnea.
  • The parameters linking individual symptom levels to progression-free survival were positive, with intervals excluding zero, in all cases except one parameter in the MV2KI model.
  • Patients who had responded to their prior chemotherapy tended to have lower cough severity in all of the multivariate models.
• The authors' conclusion: The authors concluded that the data offered little support for their expectation that pemetrexed plus best supportive care would reduce symptom burden, and that any such reduction would be linked to longer progression-free survival, compared with best supportive care alone.
• Limitations the authors state:
  • The lack of observed treatment differences in symptoms may be due to the study's relatively small sample size.
  • The questionnaire asked about the previous 24 hours but was given about every 21 days, so the assessments may not represent patients' true symptom experience.
  • A key assumption of the survival model (proportional hazards) may have been violated.

Topics

Medical terms in this study, explained in our Lexicon: Clinical Trial.

Related studies

• Hierarchical regression for multiple comparisons in a case-control study of occupational risks for lung cancer — M. Corbin, L. Richiardi, et al. (2012)
• Longitudinal pattern of reported respiratory symptoms and accelerated ventilatory loss in asbestos-exposed workers — C.A. Brodkin, S. Barnhart, et al. (1996)
• Second-line gemcitabine-based chemotherapy regimens improve overall 3-year survival rate in patients with malignant pleural mesothelioma: a multicenter retrospective study — H. Mutlu, S. Gündüz, et al. (2014)
• Outcome of asbestos exposure (lung fibrosis and antinuclear antibodies) with respect to skin reactivity: an 8-year longitudinal study — A. Lange, D. Garncarek, et al. (1986)
• Exposures from chrysotile-containing joint compound: evaluation of new model relating respirable dust to fiber concentrations — G.P. Brorby, P.J. Sheehan, et al. (2013)
• Modeling Variables With a Spike at Zero: Examples and Practical Recommendations — E. Lorenz, C. Jenkner, et al. (2017)

Browse return to medical research.