Data collection
Data were collected from the National Health (NHS) Service Cerner® database and a private healthcare database (Carebit, London UK). Patient demographic data collected, included age and ethnicity. Clinical data was extracted from patient notes and referral letters, including menstrual frequency; hyperandrogenism and hypoestrogenism symptoms; exercise history; weight loss preceding amenorrhoea; psychological stress; and occupation. Occupations were stratified using International Standard Classification of Occupations 2008 (ISCO-08).15 Data was collected on patient medical history, including fractures, eating disorders (ED), psychological comorbidities and family history. Patients BMI was collected and used to categorise patients as: underweight (<18.5kg/m2); healthy (18.5-24.9kg/m2); overweight (25-29.9kg/m2); obese (30-39.9kg/m2); or morbidly obese (≥40kg/m2). Ultrasonography reports provided endometrial thickness and polycystic ovarian morphology (PCOM) data. Reports provided BMD and Z-score of L2-L4 segment and femora from dual energy X-ray absorptiometry (DEXA) scans. Blood results either from the first clinic visit or GP referral letter were examined for LH [reference value >3 IU/L] and FSH [ reference value >6 IU/L] (used to calculate LH:FSH ratio [reference value <2]), estradiol [reference value >100pmol/L], thyroid-stimulating hormone (TSH) [reference range 0.4-4.0mU/L], free T4 (fT4) [reference range 9.0-25.0pmol/l], and prolactin [reference range 45-375IU/mL]. Given the retrospective nature of this study, it was expected that a significant proportion of patients would have missing data; to overcome this comparison between the FHA and PCOS groups were only performed when the majority (≥50%) of patients had data available.