Markov modelling of treatment response in a 30-year cohort study of newly diagnosed epilepsy

Economic Evaluation of the Community Benefit of the Australian Pregnancy Register of Antiseizure Medications

BACKGROUND AND OBJECTIVE

The Raoul Wallenberg Australian Pregnancy Register (APR) was established to collect, analyze, and publish data on the risks to babies exposed to antiseizure medications (ASMs) and to facilitate quality improvements in management care over time. It is one of the seveal prospective observational pregnancy registers of ASMs that has been established around the world. Although the APR and other registries have contributed to knowledge gain that has been applied to decrease adverse pregnancy outcomes, their cost-effectiveness remains unknown. Here, we aimed to evaluate the economic impacts of the APR from both societal and health care system perspectives.

METHODS

Using decision analytic modeling, we estimated the effectiveness (prevention of adverse pregnancy outcomes) and costs (costs of adverse pregnancy outcomes and the register itself) of the APR over a 20-year time horizon (2000-2019). The comparator was set as the adverse pregnancy outcomes collected by the APR between 1998 and 2002 (i.e., no APR derived improvements in care). In the scenario analysis, we conservatively assumed a 2.5% and 5% contribution of the APR to the savings in health care and societal costs. Adverse pregnancy outcomes included stillbirth, birth defects, and induced abortion. All cost data were derived from published sources. Health and economic outcomes were extrapolated to the total target Australian epilepsy population. The primary outcomes of interest were the return of investment (ROI) for the APR and incremental cost-effectiveness ratio (ICER) for cost per adverse outcome avoided.

RESULTS

Over the 20-year time horizon, the ROI from the APR from a societal perspective was Australian dollars (AUD) 2,250 (i.e., every dollar spent on the program resulted in a return of AUD2,250). Over this time, it was estimated that 9,609 adverse pregnancy outcomes were avoided, and health care and societal costs were reduced by AUD 191 million and AUD 9.0 billion, respectively. Hence, from a health economic point of view, the APR was dominant, providing cost saving ICERs from both perspectives.

DISCUSSION

Following its inception 20+ years ago, the APR has represented excellent value for investment for Australia, being also health-saving and cost saving from a societal and a health care perspective. With the growing number of marketed ASMs, the APR is expected to continue to have a major impact in the foreseeable future.

Treatment Outcomes of Newly Diagnosed Epilepsy: A Systematic Review and Meta-analysis

BACKGROUND AND OBJECTIVES

Understanding the multi-faceted treatment outcomes of newly diagnosed epilepsy is critical for developing rational therapeutic strategies. A meta-analysis was conducted to derive pooled estimates of a range of seizure outcomes in children and adults with newly diagnosed epilepsy commenced on antiseizure medication treatment, and to identify factors associated with different outcomes.

METHODS

PubMed/EMBASE were screened for eligible articles between 1 January, 1995 and 1 May, 2021 to include unselected cohort studies with a ≥ 12-month follow-up of seizure outcomes. Proportions of patients seizure free at different follow-up timepoints and their characteristics at the study population level were extracted. The patients were group-wise aggregated using a random-effects model. Primary outcomes were proportions of patients with cumulative 1-year seizure freedom (C1YSF), and 1-year and 5-year terminal seizure freedom (T1YSF and T5YSF). Secondary outcomes included the proportions of patients with early sustained seizure freedom, drug-resistant epilepsy and seizure-free off antiseizure medication at the last follow-up (off antiseizure medications). A separate random-effects meta-analysis was performed for nine predictors of importance.

RESULTS

In total, 39 cohorts (total n = 21,139) met eligibility criteria. They included 15 predominantly adult cohorts (n = 12,024), 19 children (n = 6569), and 5 of mixed-age groups (n = 2546). The pooled C1YSF was 79% (95% confidence interval [CI] 74-83). T1YSF was 68% (95% CI 63-72) and T5YSF was 69% (95% CI 62-75). Children had higher C1YSF (85% vs 68%, p < 0.001) and T1YSF than adult cohorts (74% vs 61%, p = 0.007). For secondary outcomes, 33% (95% CI 27-39) of patients achieved early sustained seizure freedom, 17% (95% CI 13-21) developed drug resistance, and 39% (95% CI 30-50) were off antiseizure medications at the last follow-up. Studies with a longer follow-up duration correlated with higher C1YSF (p < 0.001) and being off antiseizure medications (p = 0.045). Outcomes were not associated with study design (prospective vs retrospective), cohort size, publication year, or the earliest date of recruitment. Predictors of importance in newly diagnosed epilepsy include etiology, epilepsy type, abnormal diagnostics (neuroimaging, examination, and electroencephalogram findings), number of seizure types, and pre-treatment seizure burden.

CONCLUSIONS

Seizure freedom is achieved with currently available antiseizure medications in most patients with newly diagnosed epilepsy, yet this is often not immediate, may not be sustainable, and has not improved over recent decades. Symptomatic etiology, abnormal neuro-diagnostics, and increased pre-treatment seizure burden and seizure types are important predictors for unfavorable outcomes in newly diagnosed epilepsy. The study findings may be used as a quantitative benchmark on the efficacy of future antiseizure medication therapy for this patient population.

Development and Validation of a Deep Learning Model for Predicting Treatment Response in Patients With Newly Diagnosed Epilepsy

Importance

Selection of antiseizure medications (ASMs) for epilepsy remains largely a trial-and-error approach. Under this approach, many patients have to endure sequential trials of ineffective treatments until the "right drugs" are prescribed.

Objective

To develop and validate a deep learning model using readily available clinical information to predict treatment success with the first ASM for individual patients.

Design, Setting, and Participants

This cohort study developed and validated a prognostic model. Patients were treated between 1982 and 2020. All patients were followed up for a minimum of 1 year or until failure of the first ASM. A total of 2404 adults with epilepsy newly treated at specialist clinics in Scotland, Malaysia, Australia, and China between 1982 and 2020 were considered for inclusion, of whom 606 (25.2%) were excluded from the final cohort because of missing information in 1 or more variables.

Exposures

One of 7 antiseizure medications.

Main Outcomes and Measures

With the use of the transformer model architecture on 16 clinical factors and ASM information, this cohort study first pooled all cohorts for model training and testing. The model was trained again using the largest cohort and externally validated on the other 4 cohorts. The area under the receiver operating characteristic curve (AUROC), weighted balanced accuracy, sensitivity, and specificity of the model were all assessed for predicting treatment success based on the optimal probability cutoff. Treatment success was defined as complete seizure freedom for the first year of treatment while taking the first ASM. Performance of the transformer model was compared with other machine learning models.

Results

The final pooled cohort included 1798 adults (54.5% female; median age, 34 years [IQR, 24-50 years]). The transformer model that was trained using the pooled cohort had an AUROC of 0.65 (95% CI, 0.63-0.67) and a weighted balanced accuracy of 0.62 (95% CI, 0.60-0.64) on the test set. The model that was trained using the largest cohort only had AUROCs ranging from 0.52 to 0.60 and a weighted balanced accuracy ranging from 0.51 to 0.62 in the external validation cohorts. Number of pretreatment seizures, presence of psychiatric disorders, electroencephalography, and brain imaging findings were the most important clinical variables for predicted outcomes in both models. The transformer model that was developed using the pooled cohort outperformed 2 of the 5 other models tested in terms of AUROC.

Conclusions and Relevance

In this cohort study, a deep learning model showed the feasibility of personalized prediction of response to ASMs based on clinical information. With improvement of performance, such as by incorporating genetic and imaging data, this model may potentially assist clinicians in selecting the right drug at the first trial.