General practitioners directly referred 252 patients, of whom we randomised 201 (80%) (fig 2); 13 (5%) were not eligible, and 38 (15%) did not consent. The age and sex of these last 51 patients was similar to those randomised to the trial (mean age 67 years, 65% female). In addition, we sent letters of invitation to 439 patients after the review of records: 156 responded, of whom 124 (79%) were randomised, 17 (11%) were not eligible, and 15 (10%) did not consent. Treatment allocation and baseline characteristics were similar between participants recruited directly and those recruited by record review, although the second group were less likely to report knee pain of less than three months' duration.
Fig 2Trial profile. GP=general practitioner; TKR=total knee replacement
We randomised 325 participants to the trial: 108 to pharmacy, 109 to physiotherapy, and 108 to control. Numbers of randomised participants ranged from 1 to 88 patients per general practice. Table 1 shows the baseline characteristics of randomised participants (mean age 68 (range 55-92) years, 64% female). Recruitment characteristics were similar between treatment groups. Treatment allocation and recruitment characteristics were similar between the highest (31 or more patients recruited) and lowest (up to 30 patients recruited) recruiting practices, although patients in the lower recruiting practices were more likely to have less than three months' duration of symptoms and had higher average WOMAC pain and physical function scores. Response to follow-up questionnaires at six months was 91% (n=98) for the control group, 95% (103) for the pharmacy group, and 88% (96) for the physiotherapy group. Those lost to follow-up at six months were more likely to be male (39% (11/28) v 35% (105/297)), be older (mean age 69.9 v 67.9 years), and have higher baseline WOMAC pain and function scores (pain 9.75 v 9.08; function 32.8 v 29.7) than those who completed follow-up. Concealment of treatment from the study nurse was effective: treatment allocation was revealed to the nurse by 15 of 325 participants (seven in the control intervention and four in each treatment arm).
Baseline characteristics of participants, by treatment group. Values are numbers (percentages) unless stated otherwise
Treatments were in line with the study protocols, and no serious adverse events were reported. In the control arm, 103 participants were contacted by telephone and one protocol violation occurred. In the pharmacy arm, 105 attended for their intervention, of whom 101 (96%) had three or more intervention sessions (median 3, range 1-5), and one protocol violation occurred. A 70% increase in the prescribing of simple and compound analgesics occurred and a 52% reduction in the prescribing of non-steroidal anti-inflammatory drugs; 104 (99%) participants received advice reinforcing the advice leaflet. Ninety nine of 109 participants randomised to physiotherapy attended for at least one session; 83 (84%) had three or more sessions (median 4, range 1-6), 97 received a home exercise programme, and 92 received advice and information. Smaller numbers had at least one session of pain reducing modalities.
At three months, significant improvements in WOMAC pain and function scores occurred in the physiotherapy group, and in pain scores in the pharmacy group, when we compared each intervention separately with control. The significant differences persisted after adjustment for sex, age, and baseline WOMAC scores and duration of pain.
No statistically significant differences existed in mean WOMAC change scores between the control group and the pharmacy or physiotherapy groups at six and 12 months (table 2). Figure 3 illustrates the changes in the WOMAC pain and function scores during the whole follow-up period. We found similar results when we did the sensitivity analysis.
WOMAC pain and function scores at 3, 6, and 12 months' follow-up. Values are mean (SD) unless stated otherwise
Fig 3Mean Western Ontario and McMaster Universities osteoarthritis index (WOMAC) scores at recruitment and at 3, 6, and 12 months' follow-up. Top: WOMAC pain scores. Bottom: WOMAC function (more ...)
Table 3 shows participants' perceived global assessment of change in their knee problem compared with baseline. More of the pharmacy and physiotherapy groups, compared with the control group, were classified as responders according to the OMERACT-OARSI criteria at each of the three follow-up points, but the difference was statistically significant only at three months. Table 4 shows data on other secondary outcome measures.
Global assessment of overall change and OMERACT-OARSI response*. Values are numbers (percentages)
Secondary outcome measures at 3, 6, and 12 months' follow-up, by treatment
A higher proportion of participants in the control group than in the physiotherapy group reported consulting their general practitioner for knee pain during the six month follow-up (table 5). Self reported use of non-steroidal anti-inflammatory drugs and simple analgesia in the six month post-randomisation period was significantly lower in the physiotherapy group than in the control group (−15%, 95% confidence interval −2% to −28%; and −16%, −3% to −29%). In the pharmacy group, use of non-steroidal anti-inflammatory drugs was significantly lower than for controls (−16%, −3% to −29%), but use of simple analgesia was significantly higher (15%, 0% to 28%).
Self reported consultations with healthcare practitioners and drug use at six month follow-up. Values are numbers (percentages) unless stated otherwise