This paper utilises baseline data from a population-based prospective observational cohort study, the Clinical Assessment Study of the Foot (CASF) . Adults aged 50 years and over registered with four general practices in North Staffordshire, United Kingdom, were invited to take part in the study, irrespective of foot-related consultation. Ethical approval was obtained from Coventry Research Ethics Committee (REC reference number: 10/H1210/5) and all participants gave their written consent to participate.
At baseline, eligible participants were mailed a health survey that included general health (Short Form-12 (SF-12)) , Hospital Anxiety and Depression Scale (HADS) , comorbidities, anthropometric characteristics (self-reported height and weight), foot pain, footwear, healthcare consultation, pain medication use, and demographic and socio-economic characteristics (age, gender, marital status, education, current employment status, and occupation). Foot pain questions included: pain in the foot in the last 12 months; pain, aching or stiffness in the foot in the last month , number of days with foot pain in the last 12 months; and the Manchester Foot Pain and Disability Index (MFPDI) . The location of foot pain in the last 4 weeks was ascertained from shading a foot manikin (© The University of Manchester 2000. All rights reserved) . Participants who reported foot pain in the last 12 months and provided written consent to further contact were invited to attend a research assessment clinic where weight-bearing dorso-plantar and lateral radiographs of each foot, clinical interview and physical examination were undertaken, in accordance with defined standardised protocols [12, 18]. Health survey responders were also invited to consent to medical record review.
Scoring of radiographs and case definitions
Plain radiographs were scored by a single reader (MM) blinded to all other participant information. Osteophytes and joint space narrowing at the 1st and 2nd CMJs, NCJ and TNJ were scored (0–3) according to a validated atlas and classification system . Eight weeks later a random selection of radiographs from 60 participants were re-scored by MM and independently by HBM. Intra-rater and inter-rater reliability for the presence of OA in a midfoot joint were excellent (mean κ = 0.95, mean % agreement = 99 %) and moderate (mean κ = 0.40, mean % agreement = 78 %) respectively.
Midfoot pain was defined as self-reported pain in the last 4 weeks by shading the midfoot region on a foot manikin, designated using a pre-defined regional marking template [17, 19].
Symptomatic midfoot OA was defined as a radiographic score of 2 or more for osteophytes or joint space narrowing on either weight-bearing dorso-plantar or lateral views, in one or more midfoot joints (1st CMJ, 2nd CMJ, NCJ or TNJ), and midfoot pain in the last 4 weeks in the same foot (as defined above).
Disabling symptomatic midfoot OA was defined as symptomatic radiographic OA together with at least one of the ten items within the MFPDI function construct scored at the level of ‘on most/every day(s)’ . If all items were scored at the level of ‘none of the time’ or ‘on some days’, symptomatic midfoot OA was classed as non-disabling. Individuals were defined as having any of the above case definitions if either or both feet were affected.
Individuals identified as having non-specific inflammatory arthritis, rheumatoid arthritis, or psoriatic arthritis were excluded from the analyses based on medical record review (primary care and local hospital) or clinical X-ray report by a consultant musculoskeletal radiologist .
Estimating population prevalence
Using baseline health survey and radiographic data, the population prevalence of midfoot pain, symptomatic midfoot OA and disabling symptomatic midfoot OA in the individual were estimated using multiple imputation and weighted logistic regression modelling. Multiple imputation was used to account for missing item-level data from the health survey and estimates were then weighted to take into account selective non-response to the health survey . Missing data were inspected to ensure that the missing at random assumption was reasonable.
Imputation involved all baseline responders and utilised the following variables: age, gender, general practice, social class, marital status, number of days with foot pain in the last 12 months, Rasch-transformed MFPDI interval level scores for the pain and function constructs , individual MFPDI function items to estimate disabling symptoms , self-reported foot pain, aching, or stiffness in the last month, SF-12 score, HADS score, and radiographic foot OA and pain regions. Fifteen imputed datasets were generated and combined using Rubin’s combination rules . Prevalence estimates (and 95 % confidence intervals (CI)) were calculated using the mim: proportion command and applied to the total baseline responder population. Selective non-response to the health survey was handled by generating weighted estimates that likely reflect the profile of non-responders using information available for the whole eligible baseline population (age, gender and general practice). Weighted logistic regression was combined with the imputed datasets to generate prevalence estimates (and 95 % CI) in the whole eligible baseline population. Crude population prevalence estimates for midfoot pain, symptomatic radiographic midfoot OA and disabling radiographic symptomatic midfoot OA were then stratified by gender, age, and socio-economic class, based on occupation. This approach to estimating population prevalence mirrors our procedures adopted for estimating overall symptomatic radiographic foot OA .
The analyses described below were conducted using data from the CASF clinic cohort. A complete case approach was taken due to the generally very low levels of missing data within clinic participants.
Potential aetiological determinants of interest
Binary logistic regression estimated the crude, and as appropriate, age-gender-body mass index (BMI)-adjusted associations between symptomatic midfoot OA and selected variables. Obesity (≥30 kg/m2) at time of baseline assessment was calculated from clinic-measured height and weight. Lifetime recall of previous foot and ankle injury was ascertained on standardised personal interview (yes/no response to the question ‘Have you ever injured your feet or ankles?’, with follow-up questions to ascertain type, anatomical location and duration of time since injury, analysed in right feet only). Lifetime recall of frequent use of high-heeled footwear among females was ascertained from a health survey item on footwear  (high frequency use was defined as reported use of high-heeled shoes on most days for at least one 10-year period between the age of 20 and 49 years). Nodal interphalangeal joint (IPJ) OA was defined as a Kellgren and Lawrence  score of two or more, in two or more IPJs (digits 2–5) and the presence of two or more Heberden or Bouchard nodes (digits 2–5) across either hand .
Associated impairment and comorbidities
The following self-reported impairment and comorbidities were ascertained from the health survey and their associations with symptomatic midfoot OA examined using binary logistic regression: general health and function (SF-12 physical and mental components, with each variable dichotomised around the median of the data distribution), HADS score (categorised as normal, mild, moderate or severe), chest problems, heart problems, deafness, eyesight problems, hypertension, diabetes, stroke, cancer, circulation problems in the legs, intermittent claudication (defined by the Edinburgh Claudication Questionnaire ) and co-occurring joint pain in the last month at other weight-loaded sites (low back, hip, knee, hindfoot/ankle and forefoot). Pain location was defined using recognised body and foot manikins and templates (low back ; hip ; knee ; foot and ankle ) that have demonstrated excellent inter- and intra-rater reliability [19, 29]. Crude odds ratios (OR) with 95 % CI were reported together with estimates adjusted for age and gender.
Frequency of primary healthcare use
Among adults with symptomatic midfoot OA, the frequency of foot pain-related consultation with a general practitioner (GP) or allied health professional (physiotherapist or podiatrist/chiropodist) was summarised as the 12-month period prevalence, and the proportion of consultations that were with the National Health Service (NHS) or private practice was described. The frequency of medication use for foot pain among adults with symptomatic midfoot OA was summarised as 1-month period prevalence.
All analyses were conducted using STATA V.12.0 (StataCorp, College Station, TX, USA).