Fatal outcomes appeared in 20.1% of reports of muscle toxicity involving lung-cancer patients and immune checkpoint inhibitors, or about one in five records, in the analysis. It found positive reporting signals for nivolumab, pembrolizumab and atezolizumab, while durvalumab did not show a significant signal. The figures describe what was reported in the database, not the chance that a patient will develop the problem or die from it, and they do not show that a drug caused an event.
A signal is not a rate
The analysis set out to compare muscle-toxicity reporting across four immune checkpoint inhibitors, or ICIs, using lung-cancer reports in FAERS. The researchers reviewed reports from the first quarter of 2015 through the fourth quarter of 2024, focusing on nivolumab, pembrolizumab, atezolizumab and durvalumab. The final analysis kept reports in which the ICI was coded as the primary suspect.
To look for a reporting signal, the team used four measures: the reporting odds ratio (ROR), proportional reporting ratio (PRR), information component (IC) and EBGM. ROR and PRR are frequentist measures, while IC and EBGM are Bayesian measures. In practical terms, they look for events reported more often with a drug than the expected reporting pattern would suggest. The study required, among other criteria, a lower 95% confidence limit above 1 for ROR and a PRR of at least 2. These rules identify disproportionate reporting, not the rate at which patients develop muscle toxicity.
The full extract contained 16,144,939 adverse-event reports. Of these, 45,266 were identified as ICI-associated myotoxicity reports, and 1,515 reports with the ICI listed as the primary suspect entered the final analysis. Among the included cases, 904 (59.7%) were male and 508 (33.5%) female. The median age was 69.4 years, the middle half ranged from 49.9 to 77.6 years, and 46.2% were aged 65 to 85.
For the drug-level comparison, the reporting odds ratio was 1.20 for nivolumab, with a 95% confidence interval of 1.11 to 1.31; 1.28 for pembrolizumab, with an interval of 1.17 to 1.39; and 1.25 for atezolizumab, with an interval of 1.12 to 1.40. Durvalumab's ratio was 0.94, with an interval of 0.81 to 1.10, and did not meet the study's signal criterion. A positive reporting odds ratio should not be read as a higher clinical risk without a denominator for drug exposure.
The labels varied sharply
The study coded muscle events using MedDRA Preferred Terms, standardized labels for adverse events. Of the eight terms listed, myalgia had the most reports, 366, followed by muscular weakness with 317 and myositis with 267. The highest reporting odds ratios were seen with polymyositis, 8.99, and dermatomyositis, 7.36, while myositis had a ratio of 4.66. Across the eight terms, the ratios ranged from 1.25 to 8.99. These are disproportional-reporting values, not measures of clinical risk.
Fatal-outcome reporting also varied by term. The reported proportions were 10.93% for myalgia, 28.8% for myositis, 28.6% for dermatomyositis, 42.86% for polymyositis and 38.83% for rhabdomyolysis. Those percentages are proportions within spontaneous reports and should not be interpreted as mortality rates.
A small overlap group had the highest fatal share
Among the 267 myositis reports, isolated myositis was the largest group, with 165 reports (61.8%). The other coded overlap groups included 39 reports (14.6%) in myositis plus MG, 48 (18.0%) in myositis plus MC, and 15 (5.6%) in Triple M. The study treated these report-based overlap groups as exploratory.
Reported death ranged from 20.0% in isolated myositis to 46.7% in Triple M, and the reported outcome patterns differed across the phenotype groups, with a p-value below 0.001. Triple M, however, was based on only 15 reports. The study's authors say that the small number limits precision and statistical power, so the result is a signal for further study rather than a measure of the syndrome's true mortality.
Onset often appeared early
Timing was another feature of the reports. A time to onset could be calculated for only 611 reports. The median was 31 days, with the middle half between 14 and 82 days. Onset was recorded within 30 days in 303 reports (49.6%), between 31 and 60 days in 117 (19.1%), and after six months in 11.1%.
Median onset differed by ICI: 21 days for atezolizumab, 33 for nivolumab, 34 for pembrolizumab and 36 for durvalumab. The reported onset distributions also differed across the categories studied, with a p-value of 0.003. Because dates were available for only a subset of reports, the pattern cannot be used as a timetable for individual patients.
The limits matter
FAERS has no denominator showing how many people received each drug, so the study cannot estimate incidence or absolute mortality. Spontaneous-reporting bias, incomplete documentation, missing or incomplete clinical variables and residual confounding can all affect the pattern. Serious or fatal outcomes in a report cannot be assigned to the ICI alone on the basis of this analysis.
The authors say the patterns should not be interpreted as incidence estimates or evidence of causality, and they call for prospective and mechanistic validation. The study provides a map of reported patterns, not a comparative safety estimate for the four drugs.
Paper data and sources
Original title: Immune checkpoint inhibitor-related myotoxicity in lung cancer: a decade-long real-world pharmacovigilance analysis of the FAERS database
Authors: Liu C, Wei S, Qi X
Journal/Repository: Frontiers in pharmacology
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text