Cluster clarity? Distinct symptom clusters predict impaired quality of life

See the article by Coomans et al. in this issue, pp. 1447–1457.

Cancer patients often report symptoms (classically, such as fatigue) which tend to co-occur with other symptoms (such as physical impairment, pain, and depression).^1,2 Such co-occurring, statistically correlated symptoms can be termed “symptom clusters.” ³ This term is more than just an elegant way to condense multiple symptoms: the underlying hypothesis is that specific symptom clusters may represent common mechanistic elements that could be identified and targeted by novel treatments.

In this issue of Neuro-Oncology, Coomans and colleagues⁴ apply a powerful analytic technique to ask whether distinct symptom clusters associate with quality of life in glioma. Their study is a publication from the European Organisation for Research and Treatment of Cancer (EORTC) CODAGLIO project, which aims to combine clinical and health-related quality of life data collected from randomized trials in glioma patients.⁵

Coomans et al analyzed baseline EORTC questionnaire data from 4307 patients with newly diagnosed glioma across 11 international trials. They identified 4 distinct clusters of co-occurring symptoms. These clusters were categorized as “motor,” “fatigue,” “pain,” and a composite “gastrointestinal/seizures/bladder control.” Two of the clusters (motor and fatigue) associated with self-reported functional impairment. Specifically, having symptoms in the motor cluster was associated with significantly decreased physical, role, and social functioning (P < 0.001), independent of other factors. Similarly, having symptoms in the fatigue cluster independently associated with poorer role functioning (P < 0.001).

Their study has several features of appeal to methodology geeks: a sample size that is an order of magnitude higher than usual, a stellar and collaborative team sharing data from international randomized controlled trials, and notably, first use in the neuro-oncology quality of life field of the powerful methodology of Individual Patient Data (IPD) analysis. Whereas conventional meta-analyses use aggregate data reported in the original publications, IPD analyses extract individual-level data from each raw study dataset and combine them for re-analysis. The resulting power can provide more detailed results than conventional methods of aggregate review, driving forward both clinical and research agendas.⁶

With a clinical hat on, this rigorous analysis implies two things. The first is that fatigue remains one of, if not the, most troublesome symptoms that our patients experience. We know from authoritative primary studies that fatigue is a problem of high prevalence.^2,7 This timely work provides level 1a evidence that it is also a problem of high impact: a bad combination. Worse, as yet, no interventions are known to be effective.⁸ The imperative to study novel treatments for fatigue is obvious.

The second thing that these results imply is that where effective treatments exist, broad-based symptom screening could benefit patients. As the authors note, comprehensive assessment may enable clinical teams to address symptoms in a timely manner. Early neuro-rehabilitation and physical therapy might help alleviate the impact of motor impairment on patients’ quality of life. It is possible to foresee multidisciplinary teams which specialize in managing symptoms; indeed we have piloted something similar in Edinburgh in recent years. The potential benefits of symptom screening and intervention and perhaps which screening tools may be the best to use would be attractive questions for future trials.

Cautiously swapping the clinical hat for an academic helmet, it may not escape the notice of sharp-eyed readers that the classic example of a symptom cluster proffered at the start—namely fatigue, physical impairment, pain, and depression—involves multiple and arguably relatively diverse symptoms. By contrast, some of the clusters described by the CODAGLIO authors (such as that of fatigue with drowsiness, or pain with headache) involve only 2 relatively similar symptoms. Are these ways of defining clusters equivalent?

This question is hard to answer confidently, which in turn reflects more fundamental and as yet unresolved questions for this field. What exactly is a cluster? Is it right or wrong to cluster fatigue with drowsiness when each is inherent to the other? Can we argue, like Coomans et al, that they must measure different things because their intercorrelation was only modest? Either stance involves assumptions: which are the correct ones? How many symptoms make a cluster anyway?

Future conceptual research will help to answer such questions. Yet to be determined are: the best statistical methods to identify symptom clusters, their stability over time, the meaning patients attach to them, and putative underlying mechanisms.^1,9,10 Until then we must interpret symptom cluster research while accepting that a lot remains to be settled.

A few caveats apply more specifically to the current work. For instance, it cannot inform us about the relative impact of symptoms which were not measured in detail, neurocognitive dysfunction and depression being two relevant examples of this. And extrapolating from baseline pretrial reports in well-functioning patients to those on toxic treatments or in long-term follow-up is difficult. As acknowledged, these caveats signpost natural future studies.

Taken as a whole, the novel and rigorous analysis by Coomans et al introduces a powerful analytic method to the neuro-oncology quality of life field. The results provide level 1a evidence that fatigue impairs quality of life in patients with newly diagnosed glioma. Clinicians may consider the potential value of broad-based symptom assessment, as academics debate esoteric points of methodology.

Funding

The text is the sole product of the author and no third party had input or gave support to its writing. The author’s salary is supported by the Royal College of Physicians of Edinburgh JMAS Sim post-doctoral fellowship. There are no conflicts of interest to declare.