Jiping Zhou, MD, MPH
PGY-3 resident in Neurology
Wayne State University/Detroit Medical Center
Wazim Mohamed, MD
Associate Professor of Neurology and Neurosurgery
Wayne State University
Roberg, Lars Egil, et al. "Prediction of Long-term Survival After Status Epilepticus Using the ACD Score." JAMA neurology (2022).
Status epilepticus (SE) is a neurological emergency with a high in-hospital and long-term mortality. However, evidence regarding risk factors associated with post-SE neurological outcomes and mortality is scarce. The current study sought to investigate the risk factors associated with long-term outcomes after SE and developed a score (the ACD score) that predicted long-term mortality after SE in non-devastating brain-injured patients. This score was further validated in two independent cohorts.
This multinational study consisted of patient data from both national and linked local civil registries in Denmark, Germany, and Norway over nine years. The same inclusion (age >18 years; meets the diagnosis criteria of SE) and exclusion criteria (anoxic or hypoglycemic injury, previous episodes of SE) apply to all 3 cohorts of patients. The primary outcome was long-term survival, which was defined as survival 2 years after SE.
The Danish cohort (exploratory cohort) was used for developmemt of the ACD score. A multivariate logistic regression model was initially built to identify risk factors associated with 2-year survival. The output revealed an association between new neurological deficits (incremental change of NIHSS ≥5 between the baseline and the first follow-up) and 2-year mortality. Subsequently, a linear regression model was made to explore the predictors of developing new neurological deficits at the first follow-up. The ACD score was then developed out of a logistic regression model with the least absolute shrinkage and selection operator (LASSO) method, a technique to increase model accuracy and interpretability with a reduced number of predictors. The ACD score was eventually validated in the German and Norwegian cohorts by the estimation of the area under the curve (AUC).
Among 261 patients in the exploratory cohort, the post-discharge mortality was 47.1% with an average 2.2-year follow-up period, which was almost 3-fold the rate of in-hospital mortality (17.6%). Both age of onset and new neurological deficits during hospitalization significantly contributed to mortality in patients with SE who survived their in-hospital stay. Specifically, the development of substantial new neurological deficits (NIHSS increase ≥5) was associated with 5-fold higher odds (adjusted OR, 5.1; 95% CI, 2.2-11.8) of fatal outcome in the post-discharge period after SE. The estimated median survival for patients with substantial new neurological deficits was 34.2 (95% CI, 8.1-60.3) months for those with a 5-10 point NIHSS increase, and 3.4 (95% CI, 2.3-4.5) months for those with a ≥10 point NIHSS increase. Those who did not sustain substantial new neurological deficits had an estimated 85.2 (95% CI, 30.0-107.2) months of median survival. Further analysis revealed that the mean duration of coma in patients with both convulsive and nonconvulsive status epilepticus had a strong correlation with worsening of neurological function during the hospitalization and subsequent new neurological deficits.
The initial ACD score identified 5 prognostic factors—age, duration of status epilepticus, time to diagnosis, treatment in the intensive care unit, and level of consciousness at admission. However, time to diagnosis and treatment in the intensive care unit were excluded given their marginal contribution to the score. On a scale of 0-15 points, the 3-factor ACD score (age, consciousness level at admission, and duration of status epilepticus) greater than 10 was used as a cut-off value for poor prognosis, with a sensitivity of 0.63 and a specificity of 0.82 for being alive without new neurological deficits at first follow-up.
The validation cohorts consisted of patients from Norway (n = 139) and Germany (n = 906). The ACD score performed reasonably well in predicting 2-year survival of all the patients (Norway: AUC, 0.700 [95% CI, 0.601-0.799]; Germany: AUC, 0.689 [95% CI, 0.649- 0.728]). The prediction accuracy of the ACD score improved in the subgroup of the patients with SE from non-devastating brain injuries (Norway: AUC, 0.763 [95% CI, 0.636-0.839]; Germany: AUC, 0.733 [95% CI, 0.685- 0.781]). In the same subgroup, from the combined exploratory and validation cohorts, the ACD score showed a strong correlation with the mortality 2 years post-discharge (r2= 0.848).
This study suggests that the development of new neurological deficits has the most important impact on post-discharge mortality in patients with SE. The authors also concluded that the ACD score could be a reliable prognostic tool in predicting long-term survival post-SE. This is among the few studies that propose a robust scoring system in long-term prognostication after SE.
This study has many strengths. The two step approach used a cohort to develop the score followed by a multicenter validation. The Scandinavian patient survival data were robust because of their linkage between patients’ medical records and their civil registry data. This likely reduced potential misclassification bias attributable to missing data. Although the study Compared to other studies, which use the modified Rankin Score as an indicator of neurological outcome, the authors used the NIHSS, a more detailed measurement of specific neurological deficits. Even though the NIHSS is not commonly used in quantifying neurological deficits in SE patients, it may provide a more meaningful estimate of the patients' condition before and after discharge.
However, this study does have several limitations. Although the ACD score only has three factors, the grading system for each factor is very complex. The duration of status epilepticus has seven grades with irregularly distributed cut-off values. Moreover, the duration of SE was an estimation based on the treating physicians’ judgement due to the limited availability of continuous electroencephalography. There were multiple new definitions coined for this study, including “SE causes that were not damaging or less damaging to the brain” and “incremental change of NIHSS ≥5 between the baseline and the first follow-up”. Without a more widespread recognition of these terms and validation, the generalizability of these concepts or study validity might be limited. Finally in the validation cohort, the predication model’s performance was at best, reasonable.
The high mortality of patients with SE after discharge has remained a conundrum for a long time. This study uncovers the association of age, consciousness level at admission, and duration of SE with the post-SE long-term outcomes. More importantly, the duration of the SE, which is a modifiable risk factor, can be considered a potential treatment target to improve the prognosis in SE patients. Additional studies are needed to explore interventional strategies that could improve mortality in SE survivors.