Table S1. Change in primary and secondary outcomes after 12 weeks treatment. All values are baseline-adjusted mean change (CI) obtained from ANCOVA unless stated otherwise.
Placebo group / Etanercept group / F statistic / SignificanceAS-WIS / -0.68 (-2.56 to 1.21) / -2.75 (-4.63 to -0.87) / 2.47 / p=0.125
BASDAI / -0.10 (-1.11 to 0.90) / -1.97 (-2.98 to -0.97) / 6.96 / p=0.012
BASFI / 0.21 (-0.64 to 1.06) / -1.35 (-2.20 to -0.50) / 6.92 / p=0.012
ASQoL / -0.11 (-2.01 to 1.78) / -3.26 (-5.16 to -1.37) / 5.53 / p=0.024
HAQ-DI / 0.020 (-0.142 to 0.182) / -0.232 (-0.395 to -0.070) / 4.92 / p=0.033
Gait velocity† cm/s / 2.11 (-5.50 to 9.72) / 14.52 (6.91 to 22.13) / 5.54 / p=0.026
Step Length† cm / 0.19 (-2.33 to 2.71) / 4.14 (1.62 to 6.66) / 5.12 / p=0.031
Risk of job loss: / MW-U=153 / p=0.160
Increased total n (%)
low to med
low to high
med to high / 3 (15%)
1 (5%)
0 (0%)
2 (10%) / 1 (5%)
-
-
1 (5%)
Unchanged total n (%)
low
medium
high / 10 (50.0%)
0 (0%)
7 (35%)
3 (15%) / 8 (40%)
-
4 (20%)
4 (20%)
Decreased total n (%)
med to low
high to low
high to med / 7 (35%)
4 (20%)
1 (5%)
2 (10%) / 11 (55%)
8 (40%)
0 (0%)
3 (15%)
† both groups n=16
Table S2. Change in primary and secondary outcomes after 12 weeks treatment for patients whose risk of job loss (defined by AS-WIS score category) stayed the same or increased, compared to those whose risk decreased. All values are baseline-adjusted mean change (CI) obtained from ANCOVA.
Risk of job loss at 12 weekssame or higher
n=22 / lower
n=18 / F statistic / Significance
BASDAI / 0.11 (-0.78 to 0.99) / -2.43 (-3.41 to -1.45) / 14.90 / <0.001
BASFI / 0.24 (-0.54 to 1.02) / -1.56 (-22.42 to -0.70) / 9.88 / 0.003
ASQoL / 0.81 (-0.67 to 2.29) / -4.74 (-6.38 to -3.11) / 25.68 / <0.001
HAQ-DI / 0.03 (-0.12 to 0.19) / -0.28 (-0.45 to -0.11) / 7.39 / 0.010
Gait velocity† cm/s / 2.79 (-4.71 to 10.29) / 14.57 (6.58 to 22.56) / 4.77 / 0.037
Step Length† cm / 0.30 (-2.44 to 3.03) / 4.85 (1.93 to 7.77) / 5.18 / 0.030
†n=17, n=15
Online Supplementary Results
Results by job type and working hours lost
Seven of the twenty patients in the placebo group (35.0%) had manual jobs compared to 9/20 (45.0%) in the etanercept group. Patients with manual jobs (treatment groups combined) showed relatively little change in AS-WIS over 12 weeks (adjusted mean [CI] = 0.42 [-1.62 to 2.47], N=24) compared to those with non-manual jobs, the majority of whose AS-WIS scores decreased (-3.14 [-4.79 to -1.9], N=16; F=7.16, P=0.011). We did not obtain full work diary data during the course of the 12 week treatment period; subjects were included if they had completed at least 8 weekly diaries during that time. Combining patients across treatment groups with AS-WIS and work diary data available (N=25), the proportion of hours lost over the study period was substantively correlated with AS-WIS score at baseline (rho=0.502, P=0.010), and at 12 weeks (rho=0.366, P=0.072). Similar numbers of patients lost at least some hours due to arthritis (placebo=3/10 [30.0%], etan=6/15 [40.0%]; χ2=0.26, P=0.610).
Association between disease activity, functional impairment, work instability, and quality of life
Following the biopsychosocial model of health, we expect there to be a logical progression from disease activity (measured by BASDAI) to functional impairment (BASFI), leading to work instability (AS-WIS), all of which may directly or indirectly affect quality of life (ASQoL). Looking simply at bivariate associations at baseline, AS-WIS was found to be substantively associated with BASDAI (r=0.64, P<0.001), BASFI (r=0.41, P=0.009) and ASQoL (r=0.79, P<0.001). Changes in AS-WIS over 12 weeks were also associated with changes in BASDAI (r=0.62, P<0.001), BASFI (r=0.63, P<0.001) and ASQoL (r=0.76, P<0.001).
It was possible to calculate the ASAS 40 response(1) in 11 etanercept and 12 placebo patients. The 4 patients (all in the etanercept group) who achieved an ASAS40 response showed more improvement in ASWIS (mean [SD]=-6.70 [3.68]; SRM=1.82) than those who did not (-0.45 [3.67]; SRM=0.12). 7/20 etanercept patients (35%) achieved a BASDAI50 response compared to 1/20 placebo patients (5%; χ2=5.63, P=0.018). Mean improvement in AS-WIS was also greater for BASDAI50 responders (mean [SD]=-6.02 [3.65]; SRM=1.65) than non-responders (mean [SD]=-0.64 [3.64]; SRM=0.17). Regarding the change in risk category, 3 ASAS40 responders went from medium to low risk of job loss, 1 remained at high risk. Of the BASDAI50 responders, 6 went from medium to low risk, 1 remained at medium risk, 1 at high risk.
There were substantive correlations between objective and patient reported measures. There were moderate associations between gait velocity and BASFI (r=-0.398, P=0.024) and BASDAI (r=-0.426, P=0.015) at baseline, and also between step length and BASFI (r=-0.474, P=0.006) and BASDAI (r=-0.386, P=0.029) at baseline. Association between variables measuring change over 12 weeks was also explored. There was a strong relationship between change in velocity and change in BASFI (r=-0.486, P=0.005), and between change in velocity and change in BASDAI (r=-0.652, P<0.001). Change in step length correlated well with change in BASFI (r=-0.513, P=0.003), and with change in BASDAI (r=-0.689, P<0.001). The negative correlations denote velocity and step length increasing where BASDAI/BASFI scores decrease indicating less impairment. At baseline, there were weaker but substantive correlations between AS-WIS score and gait velocity (r=-0.387, P=0.029) and step length (r=-0.389, P=0.028). Change in AS-WIS was correlated with change in velocity (r=-0.348, P=0.051), and change in step length (r=-0.373, P=0.035) to a similar degree. At baseline, the correlations between absolute ASQoL scores and gait velocity (r=-0.270, P=0.136) and step length (r=-0.304, P=0.091) were not substantive, but change in ASQoL was correlated with change in velocity (rho=-0.427, P=0.015), and change in step length (rho=-0.613, P<0.001).
Adverse events reported
Sixteen patients who received placebo (80.0%) experienced at least one adverse event, compared to 19 (95.0%) patients in the etanercept group. There were a total of 115 adverse events (4.15 per patient-year), 51 (3.68 pp-y) in the placebo group (1 ‘definitely’ treatment-related [bruise around injection site], 22 ‘possibly/probably’ treatment-related) and 64 (4.62 pp-y) in the etanercept group (0 definitely, 26 possibly/probably). There were a total of 14 infections (1.01 pp-y) in 9 patients in the placebo group compared to 19 (1.37 pp-y) in 12 patients in the etanercept group.
References
1. Brandt J, Listing J, Sieper J, Rudwaleit M, van der Heijde D, Braun J. Development and preselection of criteria for short term improvement after anti-TNF alpha treatment in ankylosing spondylitis. Ann Rheum Dis 2004;63(11):1438-44.