For hundreds of thousands of cancer survivors, the cruelest legacy of treatment is not the tumor but the fog that follows: misplaced words, fragmented attention, a mind that no longer works at its old speed. The scientific field built to understand this condition—known as cancer-related cognitive impairment, or “chemo-brain”—has a problem of its own, according to a sweeping new analysis: the clinical trials investigating it are so methodologically fragmented that their findings may be nearly impossible to compare, combine, or trust at scale. Published in Supportive Care in Cancer, the study is the first scoping review to systematically dissect how cognition is measured and statistically analyzed in longitudinal randomized controlled trials in oncology. Led by François Christy and colleagues at the Centre François Baclesse and INSERM Unit 1086 in Caen, France, the review scrutinized 113 trials published between 2010 and 2026. Its verdict: a patchwork of neuropsychological tests, scoring conventions, and statistical models so divergent that the true trajectory of treatment-related cognitive decline—and which interventions genuinely protect the brain—remains obscured.
Once dismissed as an anecdotal complaint, cancer-related cognitive impairment is increasingly recognized as a significant consequence of cancer and its treatments, striking patients with tumors far outside the brain as well as those with central nervous system malignancies. Scientists attribute it to a web of interacting forces: the direct effects of the disease itself, systemic therapies such as chemotherapy, endocrine and targeted treatments, surgical interventions, and psychological factors including anxiety and depression. Its footprint spans memory, attention, executive function, and processing speed, ranging from subtle shifts detectable only on neuropsychological testing to deficits that erode the capacity to work, sustain relationships, and live independently. Clinicians often overlook these symptoms in routine care. Longitudinal randomized controlled trials—in which patients are followed over time while randomly assigned to different treatments or interventions—remain the most robust framework for establishing when cognitive decline emerges and what might prevent it. That is why the methodological disorder documented in the new review carries weight: trials that cannot be compared, or that are statistically underpowered, cannot deliver the answers survivors are waiting for.
To map the field, the Caen team followed PRISMA reporting guidelines and searched PubMed, Scopus, and the Cochrane Library for longitudinal randomized trials assessing cognition in adult cancer patients, covering January 2010 through June 2026. Of 418 records initially retrieved, 350 unique studies were screened, and after exclusions—most often for lacking any cognitive endpoint—113 trials remained. Two reviewers independently extracted data on each trial’s objectives, sample sizes, assessment tools, testing timepoints, statistical methods, and handling of missing data, and classified each study’s conclusion as a reported cognitive change or a null finding. Reporting quality was scored against a ten-item subset of the CONSORT 2025 statement, the newly updated international checklist for randomized trial reporting, and adherence was measured against the International Cognition and Cancer Task Force recommendations, which since 2011 have urged oncology researchers to adopt standardized neuropsychological test batteries and consistent criteria for defining impairment and change. The corpus ranged from pilot and feasibility studies to phase III trials, though 42 percent never clearly stated their phase.
The landscape the reviewers uncovered was dominated by breast cancer, the focus of 50.4 percent of trials, followed by central nervous system tumors at 21.2 percent. The overwhelming majority—82 of the 113 trials, roughly 73 percent—tested interventions intended to prevent or manage cognitive impairment, spanning cognitive training, aerobic exercise, mindfulness, yoga, qigong, acupuncture, and drugs such as donepezil, nicotine, and esketamine. The remaining 27 percent tracked cognitive change caused by cancer treatments themselves, including chemotherapy and endocrine therapy. Notably, cognition was not treated as a peripheral measure: it served as the primary endpoint in 63 of the 113 trials, a prominence that exceeds most existing guidance, which typically frames cognitive outcomes as secondary add-ons to survival or tumor response. But the field’s structural fragility was equally apparent. The median trial enrolled just 81 participants, with a median of 37 per randomization arm, and sample sizes ranged from 8 to 867—dramatically smaller than typical oncology drug trials, which frequently enroll hundreds of patients.
Measurement emerged as the field’s most glaring fault line. Across the 113 trials, researchers deployed no fewer than 76 different neuropsychological instruments. The Trail Making Test, a timed paper-and-pencil measure of attention and mental flexibility, led among trials using objective measures at 42 percent, followed by the Wechsler Adult Intelligence Scale battery at 39 percent, the Hopkins Verbal Learning Test at 28 percent, and the Controlled Oral Word Association Test at 26 percent. Yet the task force’s recommended core battery, designed to harmonize cognitive research in oncology, was fully implemented in only 19 percent of trials that objectively assessed cognition in patients with tumors outside the central nervous system. Just 37 percent of studies specified criteria for defining cognitive impairment, decline, or improvement—most commonly thresholds of one, one-and-a-half, or two standard deviations below normative means, or the Reliable Change Index, a statistic that flags score changes exceeding measurement error. Scoring diverged as well: 72 percent of trials reported raw rather than standardized scores, and only 40 percent of multi-test studies built composite scores for domains such as attention, verbal memory, and executive function.
The patient-reported side of the field was equally fractured. While 78 percent of trials included at least one objective, performance-based measure, 22 percent relied exclusively on subjective self-reports, and about two-thirds of all trials captured patients’ perceptions of their own cognition. The Functional Assessment of Cancer Therapy–Cognitive Function questionnaire, or FACT-Cog, dominated subjective measurement, appearing in just over half of studies that assessed self-reported symptoms. This divide is far from trivial: accumulating evidence shows that subjective complaints and objective test performance frequently dissociate, reflecting partially distinct dimensions of the same condition. A patient can score within normal limits on every neuropsychological test yet feel profoundly impaired at work—or the reverse. Validated patient-reported instruments such as FACT-Cog and the PROMIS Cognitive Function scale are increasingly woven into oncology trials precisely because they capture the everyday functional impact of cognitive symptoms that clinic-based testing can miss. Strikingly, in every one of the 24 trials involving brain tumors, cognition was measured with at least one objective test, even as assessment conventions stayed similar across cancer types.
The statistical machinery behind the trials varied just as widely. Eleven percent of studies relied solely on descriptive statistics, reporting mean changes from baseline without any inferential testing. Twenty percent applied simple paired comparisons, such as paired t-tests or Wilcoxon signed-rank tests, while 24 percent used baseline-adjusted fixed-effect regression models, including analysis of covariance, which statistically controls for patients’ starting scores. The most common choice, at 57 percent of trials, was repeated-measures modeling: linear mixed-effects models and repeated-measures analysis of variance that track each individual’s trajectory across multiple timepoints. These methods are favored for solid technical reasons: they explicitly model intra-individual change, tolerate irregular assessment schedules, and handle unbalanced or missing data under standard statistical assumptions. Notably, the field outperforms its neighbor in quality-of-life research, where comparable reviews found mixed models in only about 30 to 47 percent of oncology studies. Yet only 29 percent of trials reported effect sizes alongside cognitive outcomes, and the choice of analytical approach appeared unrelated to sample size, with median group sizes nearly identical across studies using descriptive, regression, or mixed-model designs.
Adherence to reporting standards exposed deeper vulnerabilities. While 89 percent of trials clearly stated cognitive objectives and prespecified outcomes, only 56 percent reported a sample size calculation—the formal justification that a study can detect the effect it seeks—and only 26 percent described how missing data were handled. Just 42 percent defined the sample analyzed for each endpoint, and a mere 22 percent provided sufficiently detailed results for their primary and secondary outcomes. These omissions are not cosmetic. In longitudinal cognitive research, patients drop out, miss assessment visits, or deteriorate, and unaddressed missingness can bias estimates of cognitive change, erode statistical power, and inflate the risk of misleading conclusions. A further blind spot is the practice effect: simply retaking the same tests can raise scores independent of any treatment, and the reviewers could not determine whether trials adjusted for it. Small samples compound everything, priming underpowered studies for type II errors—falsely concluding that an intervention does nothing. Thirty-nine trials openly acknowledged their sample size as a limitation, the most frequently cited weakness in the entire corpus.
The review also tallied what the 113 trials ultimately concluded: 72 reported a statistically significant cognitive change or between-group difference, while 41 reported null findings. Positive cognitive results clustered in particular settings. When cognition was the primary objective, 75 percent of trials reported change versus 25 percent null findings; non-pharmacological intervention trials claimed change in 75 percent of cases, while pharmacological trials split nearly evenly at 54 versus 46 percent. Conversely, studies evaluating the cognitive impact of cancer treatments reported null findings 58 percent of the time. The authors urge restraint in reading these patterns: as a descriptive synthesis, the review cannot establish whether design features cause such differences, and given the heterogeneity of tools and methods, a positive finding in one trial may not be equivalent to a positive finding in another. The inconsistencies themselves, they argue, make the case for transparent reporting and standardized assessment before the field can speak with one voice.
The team’s prescriptions are concrete. Future trials, they argue, should ground sample size calculations in expected effect sizes and repeated measures; adopt linear mixed-effects or generalized linear mixed models as the analytical default; explicitly declare the assumed missing data mechanism, typically missing at random, and probe its robustness through sensitivity analyses; adjust for baseline cognitive performance; and report the analysis sample, estimands, and model assumptions. Standardized or composite cognitive scores should become the norm to make results comparable across studies, and, at minimum, trials should embed the task force’s core neuropsychological battery while reliably capturing both objective and patient-reported outcomes. The authors concede that the 2011 recommendations themselves now need updating to reflect newer thinking on patient-reported measures and subtle cognitive change. For the growing population of survivors living with cognitive fog, the stakes are direct: the difference between a fragmented field and a rigorous one is the difference between scattered, uninterpretable findings and real answers about which treatments harm the mind—and which strategies protect it.
Subject of Research: Methodological practices for measuring and statistically analyzing cancer-related cognitive impairment in longitudinal randomized controlled trials in oncology, including neuropsychological assessment tools, statistical methods, sample sizes, and adherence to CONSORT 2025 and International Cognition and Cancer Task Force reporting recommendations.
Subject of Research: Cancer
Article Title: Methodology of analysis of cognitive impairment in longitudinal randomized clinical trials in oncology: a scoping review
Article References: Christy, F., Joly, F., Lange, M., Duivon, M., & Lequesne, J. (2026). Methodology of analysis of cognitive impairment in longitudinal randomized clinical trials in oncology: a scoping review. Supportive Care in Cancer, 34(9), Article 905. https://doi.org/10.1007/s00520-026-11059-1
Image Credits: AI Generated
DOI: 10.1007/s00520-026-11059-1
Keywords: cancer-related cognitive impairment, chemo-brain, chemotherapy-related cognitive impairment, oncology, randomized controlled trials, longitudinal studies, neuropsychological assessment, CONSORT reporting guidelines, statistical methods, missing data, clinical trials
Cite Scienmag News
APA MLA Chicago
Nathaniel Bowman. (August 30, 2026). Review maps methods for analyzing cognitive impairment in cancer trials. Scienmag. https://scienmag.com/review-maps-methods-for-analyzing-cognitive-impairment-in-cancer-trials/
Nathaniel Bowman. “Review maps methods for analyzing cognitive impairment in cancer trials.” Scienmag, 30 August 2026, https://scienmag.com/review-maps-methods-for-analyzing-cognitive-impairment-in-cancer-trials/. Accessed 30 August 2026.
Nathaniel Bowman. “Review maps methods for analyzing cognitive impairment in cancer trials.” Scienmag. August 30, 2026. https://scienmag.com/review-maps-methods-for-analyzing-cognitive-impairment-in-cancer-trials/
Copy citation Download RIS
Tags: analysis of cognitive outcomes in cancer trialsassessment of cognitive function in cancer survivorscancer-related cognitive impairmentchallenges in measuring cancer-related cognitionchallenges in measuring cancer-related cognitive impairmentchemo-brainchemo-brain analysisheterogeneity in cognitive testing protocolsimpact of cancer treatments on brain functionlongitudinal randomized controlled trials in oncologymethodological diversity in cancer cognition studiesmethods of cognitive assessment in cancer trialsneuropsychological scoring conventions in cancer researchneuropsychological scoring conventions in oncologyneuropsychological testing in cancer survivorsneuropsychological testing methodsresearch methods in oncology cognitive studiesresearch standards in cancer cognitive impairment studiesstatistical analysis of cognitive declinesystematic review of cognitive assessment toolssystematic review of cognitive measurement tools
