Objective: Studies have provided evidence of heterogeneity within chronic fatigue syndrome (CFS), but few have used data from large cohorts of CFS patients or replication samples.Methods: 29 UK secondary-care CFS services recorded the presence/absence of 12 CFS-related symptoms; 8 of these symptoms were recorded by a Dutch tertiary service. Latent Class Analysis (LCA) was used to assign symptom profiles (phenotypes). Regression models were fitted with phenotype as outcome (in relation to age, sex, BMI, duration of illness) and exposure (in relation to comorbidities and patient-reported measures).Results: Data were available for 7,041 UK and 1,392 Dutch patients. Almost all patients in both cohorts presented with post-exertional malaise, cognitive dysfunction and disturbed/unrefreshing sleep, and these 3 symptoms were excluded from LCA. In UK patients, six phenotypes emerged:'full' polysymptomatic (median 8, IQR 7-9 symptoms) 32.8%; 'pain-only' (muscle/joint) 20.3%; 'sore throat/painful lymph node' 4.5%; 'oligosymptomatic' (median 1, IQR 0-2 symptoms) 4.7%.Two 'partial' polysymptomatic phenotypes were similar to the 'full' phenotype, bar absence of dizziness/nausea/palpitations (21.4%) or sore throat/painful lymph nodes (16.3%). Women and patients with longer duration of illness were more likely to be polysymptomatic. Polysymptomatic patients had more severe illness and more comorbidities. LCA restricted to 5 symptoms recorded in both cohorts indicated 3 classes (polysymptomatic, oligosymptomatic, pain-only), which were replicated in Dutch data.
Conclusions:Adults with CFS may have one of 6 symptom-based phenotypes associated with sex, duration and severity of illness, and comorbidity. Future research needs to determine whether phenotypes predict treatment outcomes, and require different treatments.