The main findings of this study were that high fear avoidance beliefs about work at baseline were significantly associated with still being on sick leave, having no improvement in disability and no improvement in pain after 1 year in patients with CLBP. To our knowledge, this is the first study to investigate the association between fear avoidance beliefs, measured on the FABQ work and the FABQ physical activity separately at baseline and the outcomes sick leave, disability and pain after 1 year in a large sample of patients with entirely CLBP.
The findings of this study are supported by the existing literature regarding sick leave in a systematic review by Wertli et al. [6], in which two studies reported that higher levels of fear avoidance beliefs were related to lower chances returning to work in patients with CLBP [23, 24]. A direct comparison of these two studies to the present study is, however, not possible, since the former studies used the FABQ total score, whereas the present study analysed on the FABQ work and the FABQ physical activity subscales separately. The associations between fear avoidance beliefs about work and the outcomes disability and pain in patients with CLBP have been found in a few previous studies [25, 26]. Overall, the results suggest that fear avoidance beliefs about work are more strongly associated with the outcome sick leave than with the outcomes disability and pain in patients with CLBP. Furthermore, the association between fear avoidance beliefs about work and disability found in our study may be uncertain in as much as the lower limit of the 95% confidence interval is close to 1.00. This might not be surprising, since the FABQ work subscale was developed specifically to measure fear avoidance beliefs about work in relation to work loss [8]. This study included patients at sick leave as well as patients at risk for transitioning to sick leave. The 50% patients at sick leave are expected to have had relatively higher scores on the FABQ at baseline and might have contributed the most to the predictive value of fear avoidance beliefs about work found in this study. The finding that fear avoidance beliefs about physical activity were not associated with any of the examined outcomes is in accordance with those of previous studies [27, 28]. It has been suggested that the two subscales of the FABQ measure the same construct, namely pain-related fear [8], in which case both subscales would have been expected to be associated with the examined outcomes. However, the non-significant associations between fear avoidance beliefs about physical activity and the outcomes in our studies might indicate that fear avoidance beliefs about physical activity reflect other domains, i.e. lack of motivation or poor expectations regarding recovery, as previously stated [29].
In the present study, the levels of pain intensity and disability at baseline were associated with unsuccessful outcome with respect to disability and pain. Higher pain intensity was related to unsuccessful outcome in disability scores, whereas lower pain intensity was related to unsuccessful outcome in pain scores. A possible explanation for the opposing impact of pain intensity may be that patients with a high baseline pain level might have a better chance of reducing pain compared to patients with lower baseline pain levels. These findings regarding disability and pain were in line with the results of previous studies [18, 30, 31], but not in concurrence with a systematic review reporting no association with disability [32].
Although the variables in the final models were significantly associated with the three outcomes, the relatively low R2s indicate that neither of the models appear to offer a full explanation of the outcomes examined.
Limitations and strengths
The main limitation in the present study is the risk of selection bias due to missing values on the outcome variables. The proportions of missing values of 41, 34 and 35% in the outcome variables sick leave, disability and pain, respectively, might have caused misleading results inasmuch as the patients included in the analyses differed significantly from the dropouts in several characteristics. In our opinion, the relatively large amount of dropouts on the outcome variables were not likely to cause an over- or underestimation of the associations, since none of these variables were significantly associated with the outcomes in the univariate analyses (Table 3), and the differences between the patients included in the analyses and the dropouts were minimal (Additional files 1, 2 and 3: Tables S1–S3).
Another limitation in the present study is missing values on the outcome variables. In this study, although none of the variables included in the multiple logistic regression analyses had more than 10% missing values (Table 3), the final number of observations included in the analyses of the outcomes sick leave, disability and pain were reduced from a sample size of 161 to 113, 302 to 286, and 363 to 284, respectively. However, the results of the simpler multiple logistic regression analyses did not change the ORs for the association between fear avoidance beliefs about work and the outcomes disability and pain, and the OR for the outcome sick leave decreased only slightly (Additional file 4: Table S4). This might indicate that the results are relatively robust.
This study was conducted as a secondary analysis of a randomised controlled trial. Consequently, information on factors considered important for the outcomes may have been missed, i.e. catastrophizing and job satisfaction [33,34,35]. Furthermore, using data from an intervention study holds the risk of the intervention confounding the associations. However, inasmuch as the variable “group” was not significantly associated with the outcomes in any of the adjusted analyses, this aspect is not likely to be a serious risk in the present study. We did not include treatment group interactions in the model because no difference was found between groups in the original randomised controlled trial [11]. It would have been of interest to report the number of patients that transitioned to sick leave during the 12-month follow-up. Unfortunately, data for estimating this number and performing separate analysis on how these patients fared are not available.
It is a strength in the present study that both à priori selected variables and variables with a p-value below 0.1 were included in the adjusted analyses. Including à priori variables that are known or presumed to be risk factors between fear avoidance beliefs and the outcomes of interest can increase the comparability to results from previous studies and might prevent the results from being too closely fitted to the data set [36]. Had the multiple logistic regression analyses been conducted solely based on the variables with a p-value below 0.1, the backward stepwise elimination may have resulted in an over-optimistic model and random chance associations with the outcomes [37].
Clinical implications and further research
In summary, the findings of this study indicate that higher fear avoidance beliefs about work at baseline are associated with unsuccessful outcome with respect to sick leave, disability and pain in patients with CLBP after 1 year. Given the inconsistency in the existing literature, more studies are needed prior to making any firm recommendations for the use of the FABQ in clinical practice.
It is unlikely that fear avoidance beliefs are a stand-alone predictor of long-term sick leave, disability and pain in patients with CLBP. Therefore, findings on the FABQ may be included as a part of a more comprehensive composite classification in combination with other questionnaires known to be of value in a treatment oriented subgrouping of patients with CLBP, e.g. the STarT Back Screening Tool [38]. This questionnaire has been validated as a prognostic screening method to allocate patients with mixed duration of LBP into low, medium or high risk subgroups [38]. The use of the FABQ to provide further specific information on patients in the high risk subgroup might help clinicians to better understand the clinical course of patients with CLBP and to identify the individual predictors that need to be targeted in the treatment strategy.