"Functional somatic syndromes, one or many?" An answer by cluster analysis.

Objective: The aim of the present study was to address the lumper-splitter discussion on functional somatic syndromes by applying k-means cluster analyses on a heterogeneous sample of persons with unexplained somatic complaints. In favor of the lumper-side of the debate, clusters should differ only on the overall severity of the somatic complaints that were assessed. According to the splitters view, clusters should differ in symptom-specific patterns.

Methods: Three-hundred ninety four subjects with functional somatic symptoms were clustered based on their scores on 47 somatic symptoms. Three cluster solutions (k=2,3, and 4 clusters) were compared on overall symptom severity, symptom patterns, and psychological distress.

Results: Results showed that in all three solutions the clusters were defined by increasing total symptom scores and increasing psychological distress. Cluster-specific symptom patterns where evident only when more clusters (three or four) were allowed. The best fit index was found for a 2-cluster solution.

Conclusion: The finding of symptom specific patterns in clusters which could not be differentiated on overall symptom severity is in favor of the splitters' view. The finding that all other clusters could be discriminated on overall symptom severity and that the 2-cluster solution had the best fit is in favor of the lumpers' view.