summarises the participation of subjects from baseline to follow-up. A total of 77% (n = 3000) of subjects who were free of CWP at baseline and eligible for follow-up participated at the second stage of the study. However, 12% (n = 350) of these subjects did not provide SF-12 and/or pain information in their follow-up questionnaire. This analysis is therefore concerned with the 2650 participants (88%) who provided full data.
Fig. 1 Flowchart of study participation. Questionnaire sent to subjects who were free of CWP at baseline. †Incomplete participants include subjects who completed only a short or telephone questionnaire at follow-up, who did not provide complete (more ...)
3.1. Characteristics of study subjects
The prevalence of CWP at follow-up was 9.4% (n = 248). The median age of subjects was 46.8 years (95% CI 46.0–47.5) and 55.7% (n = 1476) were female. The SF12-MCS scores ranged from 7.4 to 69.0, the median score was 53.0 (95% CI 52.7–53.3). This was categorised into thirds for analysis purposes: good = 55.7–69.0, moderate = 47.5–55.6, and poor = 7.4–47.4. The range of scores for SF12-PCS was 14.6–66.1, and median score 53.8 (95% CI 53.6–54.2). The thirds of the SF12-PCS were good = 55.4–66.1, moderate = 48.6–55.3, and poor = 14.6–48.5.
3.2. Association between age and gender and HRQoL
shows the associations between age, gender, CWP status and psychosocial measures having poor, moderate and good SF12-MCS and SF12-PCS scores. Age was significantly different between each of the three levels of both SF12-MCS and SF12-PCS scales (p < 0.01). Older subjects had poorer SF12-PCS scores, whereas younger subjects had the poorest SF12-MCS scores. A greater proportion of females reported poor SF12-MCS and SF12-PCS scores.
Baseline measures and SF12-MCS and SF12-PCS scores at follow-up.
3.3. Association between pain status and HRQoL
SF12-MCS and SF12-PCS scores were significantly poorer in subjects with new onset of CWP compared to those who remained free of the disorder. Median scores were: SF12-MCS = 49.5 (95% CI 47.3–51.7) and 53.1 (95% CI 53.0–53.5), respectively; SF12-PCS = 45.5 (95% CI 43.2–47.2) and 54.2 (95% CI 53.9–54.4), respectively. shows the breakdown of scores for the three categories of SF12-MCS and SF12-PCS.
Subjects with some pain at baseline were significantly more likely to have poor SF12-MCS and SF12-PCS scores compared to those subjects who reported having no pain ().
3.4. Association between psychosocial risk markers and HRQoL
High levels, representing a poor psychosocial state, of illness behaviour, health anxiety, psychological distress (GHQ), anxiety, depression, life events, sleep problems and reporting one or more somatic symptoms were associated with both poor SF12-MCS and SF12-PCS scores (p < 0.01) (). There was no significant statistical interaction between poor scores on “mood” variables (depression, anxiety, illness behaviour, health anxiety and psychological distress) and sleep or somatic symptoms, in predicting poor SF12-MCS or SF12-PCS scores.
3.5. Relationship with SF12-MCS
Subjects with new onset of CWP were twice as likely (RRR = 2.3; 95% CI 1.6–3.2) to have the poorest SF12-MCS scores compared to subjects free of CWP at follow-up (). This relationship was not observed for the moderate SF12-MCS category. Scoring in the moderate and poor category of all the psychosocial measures put subjects at an increased risk of having a moderate or poor SF12-MCS score at follow-up.
Multinomial logistic regression analysis of the association between baseline psychosocial measures and CWP onset with SF12-MCS scores.
All significant associations were included in a multivariate model, and backward elimination was used to achieve the most parsimonious model. Variables that remained in this model are reported in . After adjustment for baseline psychosocial status CWP onset was no longer associated with SF12-MCS score (RRR = 1.2; 95% CI 0.8–1.8). However, subjects scoring in the moderate or poorest third of the GHQ, life events inventory, illness behaviour, health anxiety, HAD depression and HAD anxiety scales were all significantly more likely to report having the poorest SF12-MCS score compared to subjects scoring in the best third of these scales. The strongest of these relationships were observed with anxiety and depression, with subjects scoring in the poorest category of these scales being 4 (RRR = 4.1; 95% CI 2.8–6.1) and 5 (RRR = 5.4; 95% CI 3.5–8.4) times more likely to have the poorest SF12-MCS score, respectively.
3.6. Relationship with SF12-PCS
There was a significant increased risk of scoring in the moderate or poor categories of the SF12-PCS at follow-up in subjects who had new onset of CWP compared to those free of the disorder: RRR = 2.1; 1.4–3.2 and RRR = 8.0; 95% CI 5.4–11.8, respectively (). Subjects in the poorest two categories of sleep problems, anxiety, depression and life events were at an increased risk of having the poorest SF12-PCS scores. However, poor scores of health anxiety and illness behaviour, and reporting one or more somatic symptoms, put subjects at an increased risk of having either moderate or poor SF12-PCS scores.
Multinomial logistic regression analysis of the association between baseline psychosocial measures and CWP onset with SF12-PCS scores.
The relationship between CWP onset and being in the poorest category of the SF12-PCS was attenuated in the parsimonious multivariate model (as shown in ); however, there was still a 2-fold (RRR = 1.8; 95% CI 1.2–2.9) and 5-fold (RRR = 4.8; 95% CI 3.1–7.4) increased risk of having a moderate or poor SF12-PCS score, respectively. High levels (scoring in the poorest third) of baseline illness behaviour (RRR = 7.0; 95% CI 4.9–9.85), sleep problems (RRR = 1.9; 95% CI 1.4–2.6) and reporting one or more somatic symptoms (RRR = 1.7; 95% CI 1.4–2.2) were also significant independent predictors of poor SF12-PCS scores. High levels of depression were moderately associated with poor PCS scores (RRR = 1.4; 95% CI 0.998–1.9), and, although this relationship was not statistically significant, a log likelihood test indicated that it was an important predictor and should be retained in the model.
3.7. Baseline pain status
Having some pain at baseline was associated with an increased risk of having CWP at follow-up (RRR = 6.0; 95% CI 4.1–8.8), and having the poorest SF12-MCS (RRR = 2.1; 95% CI 1.7–2.6) and SF12-PCS (RRR = 4.8; 95% CI 3.9–6.1) scores at follow-up. The final multivariate models were then adjusted for the presence of some pain at baseline. The relationship between new onset of CWP and SF12-MCS did not change (RRR = 1.2 95% CI 0.8–1.8). However, the risk of subjects with new onset of CWP being in the poorest third of the SF12-PCS remained, although slightly attenuated (RRR = 4.0; 95% CI 2.6–6.2).
3.8. Comparison of participants and non-participants at follow-up
Baseline measures were compared for participants (n = 2650) and non-participants, including subjects who did not respond or who did not provide complete data at follow-up (n = 1225), these results are presented in . Non-participants were more likely to be younger than participants (p = 0.02) and to report having some pain at baseline (p = 0.02). Scores on the GHQ, HAD anxiety and depression, and illness behaviour scales were poorer in non-participants than participants (p < 0.01). There was also a difference in somatic symptom scores between groups; however, the median and 95% CI was 0 and 0–0 for each.
Comparison of baseline measures between non-participantsa and participants at follow-up.