Selection of studies

The PuMed database was used for a computer assisted search up to March 2012. The search strategy used the key words: ‘orofacial pain’, ‘oral pain’, ‘facial pain’, ‘neuralgia’, ‘trigeminal neuralgia’, ‘tendomandibular dysfunction’, ‘burning mouth syndrome’, ‘odontalgia’, ‘(post) herpetic neuralgia’, ‘stomatodynia’, ‘atypical facial pain’, ‘cancer facial pain’ in combination with ‘antidepres*’ and ‘(facial) pain’. The following limits were applied: the type of article was ‘randomized controlled trial’ and ‘controlled-clinical-trial’, and the language ‘Dutch or English’. Figure 1 shows the search strategy to identify relevant studies. Additional reports were identified from the reference lists of retrieved reports and from review articles.

Search strategy used to identify studies on orofacial pain and the use of antidepressants.

Two investigators reviewed all identified trials independently to determine whether a study should be included. Inclusion criteria were single or double blinded randomized controlled trials (RCTs), dealing with patients suffering from orofacial pain disorders, with pain intensity as main outcome measure and antidepressants as treatment modality. A flow-chart of the retrieved and included studies is presented in Fig. 2 .

Flow chart of the studies retrieved and included in this review.

Methodological quality of the studies

Trials concerning treatment effectiveness were scored using a 15 item criteria list ( Table 1 ). Different criteria lists measuring the methodological quality of a RCT have been described. It is not clear which list is most valid. For this systematic review the most extended item list was chosen. This list includes selection and restriction of the study group, treatment allocation, study size, prognostic comparability, drop outs, interventions, extra treatments, blinding procedure, outcome measurements, follow-up period, side effects and analysis and presentation of data. A maximum score of 100 for each study can be achieved, when each criterion was weighted. Comparability of the different treatment groups is the essence of a good clinical trial. As allocation procedure and drop-out rates are key elements in (randomized) controlled trials these criteria received the highest possible scores in the check list. Two independent investigators (W.J.J.M. Martin and T. Forouzanfar) assessed the methodological quality of the trials. In case of disagreement, consensus between the two investigators had to be found. If no agreement could be reached a third investigator was consulted. The assessment resulted in a hierarchical list in which higher scores indicate studies with a higher methodological quality.

Statistics

Only trials with pain intensity as the main outcome measure were considered as relevant. As mentioned in the study by van Tulder et al., a cut-off point of 50 was used for the methodological quality score. A trial was qualified as high quality if the methodological score was 50 points or more. If the score was less than 50 points, the study was considered to be of low quality. The level of evidence for therapeutic intervention effectiveness was categorized according to four levels of scientific evidence, based on the quality, outcome and relevance of the studies. This includes: strong evidence, moderate evidence, limited evidence and no evidence. For strong evidence, multiple relevant, high quality trials were identified; for moderate evidence, one relevant, high quality trial and one or more relevant low quality trials were identified; for limited evidence, one relevant high quality trial or multiple relevant, low quality trials were identified; and for no evidence, one relevant, low quality trial or no relevant trials or trials with contradictory outcomes were identified.