- Widespread adoption of chimeric antigen receptor (CAR) T-cell cancer therapy depends on the ability to prognosticate and mitigate adverse effects, which notably include neurotoxicity
- In this study, researchers constructed a multivariable model to determine the probability of neurotoxicity in 126 patients with relapsed or refractory lymphoma who received axicabtagene ciloleucel
- At a cutoff value of 6, the scoring system was 77% accurate in a validation cohort of 78 patients
- With improved prognostication of neurotoxicity after CAR T-cell therapy, higher-risk patients could be triaged and treated early, while lower-risk patients could be spared unnecessarily prolonged hospitalization
Neurotoxicity is one of the two principal treatment-limiting effects of chimeric antigen receptor (CAR) T-cell therapy. It's less common than the other main toxicity, cytokine release syndrome (CRS), but the time of onset is more variable. In many institutions, patients treated with CAR T-cell therapy are routinely kept in the hospital for seven to 14 days to be observed for neurologic symptoms.
Neurologist Daniel B. Rubin, MD, PhD, critical care neurologist in the Department of Neurology at Massachusetts General Hospital, and colleagues have created a prognostic tool to determine which patients are likely to develop neurotoxicity and need prolonged hospitalization. Their report appears in JAMA Neurology.
To derive the multivariable model, the researchers used data on 126 patients who received axicabtagene ciloleucel for relapsed or refractory lymphoma between April 2015 and April 2019. To validate the model, they studied 78 such patients who were treated between May 2019 and February 2020.
Deriving the Tool
A clinical score was calculated as follows:
- Age ≥52—1 point
- C-reactive protein ≥8.95 mg/dL—1 point
- White blood cell count ≤790/microliter—1 point
- CRS onset prior to day 3—1 point
- Aggressive histologic subtype—2 points
- Fever ≥38.5 °C—2 points
- CRS of any grade up to day 5—2 points
- Tocilizumab treatment of CRS before or on day 5—3 points
The possible score ranges from 0 to 14. When tested on the entire derivation cohort, the tool had an accuracy of 86%.
Validating the Tool
The researchers calculated the predictive score for each patient in the validation cohort. When a score of 6 was used as a cutoff, the model correctly predicted the occurrence of neurotoxicity with an accuracy of 77%, sensitivity of 82% and specificity of 70%.
In the validation cohort, the predictive score was also able to detect patients who were unlikely to develop moderate neurotoxicity (grade ≥2) or severe neurotoxicity (grade ≥3).
Eight patients in the validation cohort were misclassified as unlikely to develop neurotoxicity (false-negatives). Five of them had only grade 1 symptoms, and seven had delayed onset of CRS before developing neurotoxicity.
Advantages of the Tool
The prognostic score is calculated from objective data during the first five days of hospitalization, so high-risk patients can be triaged to higher levels of care. For patients unlikely to develop neurotoxicity, earlier hospital discharge may be possible.
Delayed onset of CRS should alert clinicians to the possibility of delayed neurotoxicity, even in patients previously deemed to be at low risk. It's worth noting that most patients with false-negative results in this study had only grade 1 symptoms, which can be treated in the outpatient setting.
Learn more about the Department of Neurology at Mass General
Learn more about neurology research at Mass General