To develop models to predict the probability of new or worsening depressive symptoms in adults following temporal lobe epilepsy (TLE) surgery.

Neurosurgery for pharmacoresistant TLE is a potentially curative treatment. Between 10 - 40% of patients can develop new or worsening depression post-operatively. Despite its prevalence and significant impact on patient quality of life, less than 3% of all research on TLE surgery has examined psychiatric morbidity.

Candidate clinical predictors were retrospectively abstracted from the Electronic Medical Records for an initial cohort of 235 adults (54% F, M_age = 36). All participants had undergone TLE surgery and completed the Beck Depression Inventory-II (BDI-II) pre- and post-operatively. Stepwise backward selection with p > 0.10 as the criteria for removal was utilized to identify the predictors to include in the final multivariable logistic regression model.

Median time from surgery to follow-up testing was 6.2 months. Eleven percent of patients experienced clinically significant increases in depressive symptoms, defined as a 15% increase in BDI-II score and a post-operative score > 11. The final model included history of depression OR 5.05 (CI 95%: 2.06 – 12.38; p < 0.001), history of other psychiatric comorbidity OR 3.74 (CI 95%: 1.37 – 10.22; p = 0.01), and left-sided resection OR 2.47 (CI 95%: 0.949 – 6.421; p = 0.064). The model had a sensitivity of 12%, specificity of 99%, negative predictive value of 90%, and positive predictive value of 60%. 

Pre-operative psychiatric history and left-sided resection are important predictors of post-operative depressive symptoms in adults with TLE. With its high specificity, when our predictive model classifies a subject as high risk there is a strong indication for increased depressive symptoms following TLE surgery. However, our model’s accuracy needs to be refined as we look to improve our sensitivity and validate our final model in an independent dataset.