Background. The model of neuroinflammation has been proposed as a possible explanation of depression. Investigations of serum levels of tumor necrosis factor-α (TNF-α) in depressed patients have previously shown contradictory results of increased and decreased levels of TNF-α during the treatment of depression. Methods. We compared the serum levels of TNF-α in two cohorts of patients suffering from depression (ICD-10 criteria): one cohort from a psychotherapeutic unit ( n = 18 ), where patients were treated with Cognitive Behavioral Analysis System of Psychotherapy (CBASP), and the other cohort from a psychiatric day care unit ( n = 16 ). Both cohorts were investigated at the beginning and at the end of treatment. The intensity of depression was measured by means of the Beck Depression Inventory, 2nd edition (BDI-II) at both time points. Results. We observed a statistically significant increase of TNF-α in the psychotherapeutic unit at time point 2 compared to time point 1 ( T = − 14.71 , p < 0.001 ), but not in the psychiatric day care unit. In both cohorts, BDI-II scores at time point 2 were significantly decreased compared to time point 1 (psychiatric day care unit: T = 3.32 , p = 0.005 ; psychotherapeutic unit: T = 6.22 , p < 0.001 ). There was a significant correlation in the psychotherapeutic unit at time point 2 ( r = − 0.682 , p = 0.02 ). Conclusion. As TNF-α was increased at time point 2 in the psychotherapeutic unit but not in patients of the psychiatric day care unit, we propose the different durations of pretreatments in both cohorts and the associated processes of neuroinflammation as a possible explanation for our results. The lack of information about the time course of TNF-α in depression could in general explain the huge variety of TNF-α levels in different cohorts of depressed patients reported in the literature.