Article Text

PDF
OP0067 Cluster analysis of clinical data identifies fibromyalgia subgroups
  1. E. Docampo1,
  2. G. Escaramis1,
  3. R. Rabionet1,
  4. J. Carbonell2,
  5. J. Rivera3,
  6. J. Alegre4,
  7. J. Vidal5,
  8. X. Estivill1,
  9. A. Collado6
  1. 1Genes and Disease, CRG, UPF and CIBERESP
  2. 2Fibromyalgia Unit. Rheumatology service, Parc de Salut Mar, Barcelona
  3. 3Rheumatology Unit, Instituto Provincial de Rehabilitaciόn, Hospital Universitario Gregorio Marañόn, Madrid
  4. 4Chronic fatigue syndrome Unit, Hospital Vall d’ Hebrόn, Barcelona
  5. 5Rheumatology Unit, Hospital de Guadalajara, Guadalajara
  6. 6Fibromyalgia Unit. Rheumatology service, Hospital Clínic, Barcelona, Spain

Abstract

Background Fibromyalgia (FM) is characterized by widespread pain and other symptoms, such as psychiatric and physical comorbidities. The heterogeneity of FM hinders its assessment and management.

Objectives The aim of this work is to identify FM subgroups by classifying clinical data into simplified dimensions.

Methods 44 variables were evaluated in 560 unrelated Spanish FM cases of Caucasian origin. All participants fulfilled 1990 ACR FM criteria and were evaluated at one of five Rheumatology Units. A partitioning analysis was performed to find groups of variables similar to each other, thus describing underlying FM dimensions. Given the mixed nature type of the variables, these were transformed into binary types (0=mild; 1=severe), and a generalization of the Gower method was applied to find similarities between variables. A score was constructed per sample and dimension based on the weights of the variables depicting the specific dimension. Kmeans clustering procedure was then applied into resulting scores to create FM subgroups. This analysis was also performed in a replication set of 950 cases.

Results Variables clustered into three independent dimensions: pain and other symptoms, family and personal comorbidities and clinical scales (fig. 1). Only the two most reliable dimensions (pain and other symptoms and comorbidities) were considered for FM subgroups construction. Resulting scores classified FM samples into three subgroups: high symptomatology and comorbidities, high symptomatology but low comorbidities and low symptomatology and comorbidities. Both the variable clustering and the sample subgrouping were replicated in the second cohort.

Figure 1. Clustering of variables. The variables included in each dimension are listed and sorted by their weighted contribution.

Conclusions We have identified three subgroups of FM samples in a large cohort of FM by clustering clinical data. This partitioning method could be used as a useful tool in FM severity assessment and personalized treatment.

Disclosure of Interest None Declared

Statistics from Altmetric.com

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.