Quebec Task Force Classification of Back-Related Leg Pain
Quebec Task Force Classification of Back-Related Leg Pain
This study investigated whether subgrouping of LBP patients based on leg pain patterns had any prognostic implications. Patients with LBP + NRI improved more than other subgroups on change in activity limitation but had a poorer outcome as measured by absolute RMDQ scores after one year. Patients with Local LBP, LBP + pain above knee, and LBP + pain below knee all had similar trajectories of activity limitation. This resulted in similar absolute RMDQ scores for LBP + pain above knee and LBP + pain below knee patients. In contrast, the Local LBP was the subgroup least affected by activity limitation both at baseline and after one year.
There was no significant association between subgroups and global perceived effect above that which could be explained by differences in duration. For the outcome of sick leave, patients in the LBP + NRI subgroup had a larger risk of long-lasting sick leave at 3 months compared with patients in the Local LBP and LBP + pain above knee subgroups.
The larger improvement in activity limitation within the LBP + NRI group was not explained by other measured baseline factors. Duration, age, gender, previous LBP episodes, pain irritability, LBP intensity, leg pain intensity, depression, and general health were all taken into account but differences on these factors between subgroups were not shown to be the reason for the different trajectories. Thus, the presence of neurological signs was associated with larger improvement, but at the same time a poorer outcome, and this is likely to be a direct effect of nerve root involvement. The finding that patients with neurological signs report better global perceived effect and poorer absolute outcome has been observed in previous studies that used unadjusted analyses. Those studies included patients from surgical departments and a workplace setting. In a primary care cohort, which predictably had a shorter LBP duration than our secondary care cohort, Hill et al. found that prognostic differences between subgroups with local LBP, LBP + pain above knee, and LBP + pain below knee were explained by other baseline characteristics. It may be that such baseline characteristics are important covariates early in the clinical course but our results highlight that the inclusion of neurological signs is more prognostically important than only distinguishing between pain above and below the knee, and we believe presence of the LBP + NRI subgroup is likely to be a central explanation for why our results differ from those of Hill et al.
A strength of the current study is that data were collected prospectively from a near-complete cohort of people in routine care. We believe this strengthens the generalisability of our results to other chronic LBP populations. Furthermore, the sample size was adequate for the conducted analyses, and data were available that made possible analyses of outcomes across different domains of health. Lastly, the response rates of 76% and 70% at the two follow-up time points, that were very similar in all the studied subgroups, we consider to be acceptable for a clinical registry.
The study also had limitations. The most important limitation from our perspective was the definition of nerve root involvement. Classification into the group with LBP + NRI required the presence of just one positive finding in the neurological examination, and the reliability of these findings in our clinical department, notwithstanding an ongoing quality assurance program, is unknown. A lack of such knowledge and less stringent procedures for data collection than are possible in clinical trials are inherent limitations of data from large clinical databases that were not collected for a specific research project. Moreover, for unknown reasons, answers to sick leave questions at follow-up were more often incomplete than other outcome measures. However, this did appear to affect subgroup differences.
Overall, this simple QTF classification of LBP displayed an association with the outcome of activity limitation that was above what could be explained by other measured baseline characteristics, and the QTF subgroups were also associated with sick leave after 3 months when only duration was included as a covariate. Subgroup differences were most marked between Local LBP and LBP + NRI and sometimes these groups also differed from other groups. However, whether leg pain location was above or below the knee was not an important distinction for the outcome measures investigated.
Despite the QTF classification displaying statistically significant associations at a subgroup level, it explained very little of the variance (2%) in the outcome activity limitation at an individual patient level and the predictive ability relating to sick leave was also low when measured by the AUC statistic. It is not uncommon in LBP that prognostic factors show statistically significant associations with outcome at a group level but little predictive value at an individual level and there is no evidence for a single factor that substantially affects LBP prognosis on its own for all individuals. Also, investigating separate prognostic factors is a necessary step to inform more sophisticated modelling of multiple factors that may be more accurate for individuals. Hayden et al. classified prognostic research as a 3-step sequential process. Initially, factors that are associated with outcome are identified, then tested for their independent effect on outcome, and lastly prognostic pathways are investigated by mapping how prognostic factors, mediators and moderators interact and influence outcome. The current study would be classified as a second step investigation and suggests that leg pain and presence of neurological signs should be included in studying prognostic pathways.
However, another potentially important result from prognostic research is the treatment implications of subgroup-targeted treatment. Classification tools such as the STarT Back Tool have shown that appropriate matching of treatment pathways to prognostic subgroups can result in better patient outcomes that are also cost-effective. It may be that the cost-effectiveness of LBP care can be improved by subgroup-focused treatment of patients, even if those groups, such as QTF subgroups, are not completely homogenous. Put another way, useful improvements in outcomes may result at a clinical population level, even if the predictability of outcome in individuals remains limited. The current study was unable to explore this as treatment was not targeted to the QTF subgroups, but the principle of subgroup-targeted treatment is a promising direction for research.
Discussion
This study investigated whether subgrouping of LBP patients based on leg pain patterns had any prognostic implications. Patients with LBP + NRI improved more than other subgroups on change in activity limitation but had a poorer outcome as measured by absolute RMDQ scores after one year. Patients with Local LBP, LBP + pain above knee, and LBP + pain below knee all had similar trajectories of activity limitation. This resulted in similar absolute RMDQ scores for LBP + pain above knee and LBP + pain below knee patients. In contrast, the Local LBP was the subgroup least affected by activity limitation both at baseline and after one year.
There was no significant association between subgroups and global perceived effect above that which could be explained by differences in duration. For the outcome of sick leave, patients in the LBP + NRI subgroup had a larger risk of long-lasting sick leave at 3 months compared with patients in the Local LBP and LBP + pain above knee subgroups.
The larger improvement in activity limitation within the LBP + NRI group was not explained by other measured baseline factors. Duration, age, gender, previous LBP episodes, pain irritability, LBP intensity, leg pain intensity, depression, and general health were all taken into account but differences on these factors between subgroups were not shown to be the reason for the different trajectories. Thus, the presence of neurological signs was associated with larger improvement, but at the same time a poorer outcome, and this is likely to be a direct effect of nerve root involvement. The finding that patients with neurological signs report better global perceived effect and poorer absolute outcome has been observed in previous studies that used unadjusted analyses. Those studies included patients from surgical departments and a workplace setting. In a primary care cohort, which predictably had a shorter LBP duration than our secondary care cohort, Hill et al. found that prognostic differences between subgroups with local LBP, LBP + pain above knee, and LBP + pain below knee were explained by other baseline characteristics. It may be that such baseline characteristics are important covariates early in the clinical course but our results highlight that the inclusion of neurological signs is more prognostically important than only distinguishing between pain above and below the knee, and we believe presence of the LBP + NRI subgroup is likely to be a central explanation for why our results differ from those of Hill et al.
A strength of the current study is that data were collected prospectively from a near-complete cohort of people in routine care. We believe this strengthens the generalisability of our results to other chronic LBP populations. Furthermore, the sample size was adequate for the conducted analyses, and data were available that made possible analyses of outcomes across different domains of health. Lastly, the response rates of 76% and 70% at the two follow-up time points, that were very similar in all the studied subgroups, we consider to be acceptable for a clinical registry.
The study also had limitations. The most important limitation from our perspective was the definition of nerve root involvement. Classification into the group with LBP + NRI required the presence of just one positive finding in the neurological examination, and the reliability of these findings in our clinical department, notwithstanding an ongoing quality assurance program, is unknown. A lack of such knowledge and less stringent procedures for data collection than are possible in clinical trials are inherent limitations of data from large clinical databases that were not collected for a specific research project. Moreover, for unknown reasons, answers to sick leave questions at follow-up were more often incomplete than other outcome measures. However, this did appear to affect subgroup differences.
Overall, this simple QTF classification of LBP displayed an association with the outcome of activity limitation that was above what could be explained by other measured baseline characteristics, and the QTF subgroups were also associated with sick leave after 3 months when only duration was included as a covariate. Subgroup differences were most marked between Local LBP and LBP + NRI and sometimes these groups also differed from other groups. However, whether leg pain location was above or below the knee was not an important distinction for the outcome measures investigated.
Despite the QTF classification displaying statistically significant associations at a subgroup level, it explained very little of the variance (2%) in the outcome activity limitation at an individual patient level and the predictive ability relating to sick leave was also low when measured by the AUC statistic. It is not uncommon in LBP that prognostic factors show statistically significant associations with outcome at a group level but little predictive value at an individual level and there is no evidence for a single factor that substantially affects LBP prognosis on its own for all individuals. Also, investigating separate prognostic factors is a necessary step to inform more sophisticated modelling of multiple factors that may be more accurate for individuals. Hayden et al. classified prognostic research as a 3-step sequential process. Initially, factors that are associated with outcome are identified, then tested for their independent effect on outcome, and lastly prognostic pathways are investigated by mapping how prognostic factors, mediators and moderators interact and influence outcome. The current study would be classified as a second step investigation and suggests that leg pain and presence of neurological signs should be included in studying prognostic pathways.
However, another potentially important result from prognostic research is the treatment implications of subgroup-targeted treatment. Classification tools such as the STarT Back Tool have shown that appropriate matching of treatment pathways to prognostic subgroups can result in better patient outcomes that are also cost-effective. It may be that the cost-effectiveness of LBP care can be improved by subgroup-focused treatment of patients, even if those groups, such as QTF subgroups, are not completely homogenous. Put another way, useful improvements in outcomes may result at a clinical population level, even if the predictability of outcome in individuals remains limited. The current study was unable to explore this as treatment was not targeted to the QTF subgroups, but the principle of subgroup-targeted treatment is a promising direction for research.