Though the majority of the patients were of the same age group (i.e. middle-aged), considerable diversity in the characteristics of their disease was observed (i.e., duration, severity, number, type of episodes). In contrast to randomized trials, the composition of the therapy cohorts in this study is diverse: the baseline data suggest that the regimens chosen are the result of differential considerations of the treating psychiatrists according to their patients’ history and needs [22
The observed discontinuation rate of 28.4% in our study was low compared to rates observed in controlled clinical trials [13
]. Furthermore, we found comparably high retention rates in all treatment cohorts, even though the patients in these cohorts differed perceptibly in clinical preconditions. In general, patients on lithium tended to retain their medication longer, and discontinued less often than patients in any of the other cohorts. As lithium is one of the oldest and most established medications for the treatment of bipolar disease, a high percentage of patients in this group has been taking this medication for a long period of time (some >10 years). As patients would hardly retain a medication this long if they were not comfortable with it, the implication is that in the lithium group we observe a pre-selection of rather satisfied and therefore compliant patients. Kessing et al [23
] in 2007 published data from a Danish medical register study with the finding for lithium that the mean time to discontinuation was 181 days. The discrepancy in time on medication to our observation might again be due to the patient population on a long standing lithium medication, who had passed the time point of early discontinuation found in the same publication to be at 45.2 days for 25% of patients.
In contrast, the OC and OMS cohorts comprised more difficult-to-treat patients: patients with mixed episodes, rapid cycling, and low adherence. Especially, OMS patients tended to suffer more often from additional psychiatric illnesses, experience more hospitalizations, less clinical improvement, reach remission status less frequently, switch to new medications earlier, and have a less positive attitude towards their medication.
The various treatment cohorts were largely comparable with regard to time on mood stabilizing medication. Though the longest time on mood stabilizing therapy was seen in LM, the difference was only notably large compared to OM and OMS. No further notable differences were observed between the cohorts. In addition, there was no relevant difference at all in the retention rates. The hazard for discontinuation was also comparable in most cohorts. Only in LM the risk was lower than in OM. In AM and OMS it was higher than in OM. This observation, as well as other factors found associated with increased risk of discontinuation (non-stable social environment, comorbid psychiatric diseases, and rapid cycling) matches well with the notion that OMS comprised patients who were rather difficult to treat. A similar but less pronounced tendency could be observed for the OC cohort. It should be noted that in this cohort with combinations of less frequently used mood stabilizing agents, the highest proportion of patients with rapid cycling was found.
The drug attitude measured as DAI was positive and high from the start in all cohorts, indicating that the majority of these outpatients had a positive attitude regarding their medication at the start of the study. It improved only slightly over the course of the study. This might result from several factors: 1) the selection criteria of stabilization, 2) the natural treatment setting where physicians are familiar with their patients through a long-lasting therapeutic relationship, and 3) increased adherence through the patients’ awareness of taking part in a study. Quite in line with this, good baseline adherence was found to be one predictor of adherence throughout the study, while a higher number of manic episodes and a high baseline CGI were predictors of non-adherence: OMS had the highest number of manic episodes, the highest percentage of patients with a baseline CGI-BP >3, and the lowest DAI.
Interestingly, alcohol consumption was positively associated with adherence. The variable in the model was “any consumption”, and 35.1% of the patients reported consumption, whereas only 4.1% stated alcohol abuse or addiction. One could speculate, that this observation possibly could reflect the occasional social drink, which might be an indicator of social integration presumably associated with general well-being and stability.
The majority of patients were without relapse at Visit 7 (day 540), rates between the cohorts were largely similar. The only notable difference was seen in OC compared to OM and LM which corresponds with the predictors found in the logistic regression model. OC had indeed the highest number of depressive episodes and also the highest number of patients with a CGI-BP >3.
As a CGI-BP of ≤3 was an inclusion criterion, there was little room for improvement, and thus the CGI-BP scores decreased only slightly over time in all cohorts. Correspondingly, the rates of patients with a CGI-BP ≤3 were around 80% to 90% in all cohorts at both visits, with only minor variations in both directions. The aim of the maintenance treatment, i.e. stabilization, could thus be regarded as achieved.
Due to the observational design of the study and the determination of the mood stabilizing medication that was chosen for the individual patient according to clinical reasoning, a comparison between the different cohorts of medications and combinations is not intended and not possible. We should stress the point that patient enrollment to the study was at the discretion of the treating physician. Therefore, results have to be interpreted in the context that medication and patients were selected on the basis of their individual preferences and history. In summary, the different treatments we observed achieved similar effects regarding maintenance treatment of bipolar disease. This is surprising regarding the diversity of the patients’ disease and social characteristics at the start of documentation, hinting again at the capability of physicians and their patients to optimize individual treatment using the spectrum of medications available. Thus, this trial design mimics clinical treatment reality better than a randomized study taking into account physicians’ skills and patient diversity. Besides the classic treatments with lithium and anticonvulsants, olanzapine was found to be a relevant mood stabilizing treatment option for a considerable number of patients in this German study sample. Other atypical antipsychotics were also used in clinical practice, but as these had not yet been approved for the treatment of bipolar disease at the start of the study, they were applied in few patients only (OMS-cohort).
Only small numbers of patients were reported with TEAEs or adverse events as assessed by the CGI-tolerability.
On the other hand, 87.0% of the overall sample reported at least one adverse event at least once over the course of the study on the predefined solicited checklist of side effects specific for the various mood stabilizing medications. Obviously, these effects were observed, but not judged by the patients to be relevant enough to be reported spontaneously as AEs. Solicited adverse events also were no reason for physicians or patients to stop or switch medication.
Within the cohorts, OLC patients reported the highest overall number and variety of solicited adverse events, closely followed by the OC groups (which also included patients receiving lithium plus one or several other substances for mood stabilization). Compared to LM it appears that these patients experienced adverse events of lithium as well as typical ones for the other mood stabilizers. In all cohorts, the most common AE was weight gain. This underlines the potential risk of weight gain in the maintenance therapy of bipolar disorder, the need to monitor and if necessary to treat bipolar patients for metabolic-related adverse events.