This study described patients with LBP seen in a secondary care outpatient department using four predefined subgroups based on the Quebec Task Force classification
. Overall, the differences observed between these groups were small. However, people who had local LBP only differed from those in the leg pain subgroups on almost all the described characteristics. Furthermore, there was a consistent trend in all health domains of increasing severity from the subgroup with local pain, across the subgroups with pain referral above and below the knee, to the subgroup with signs of nerve root involvement.
Different patient profiles were observed among patients with leg pain depending on whether they had signs of nerve root involvement or not. Patients with signs of nerve root involvement had a more severe profile on measures from all health domains. Similar findings were observed in a previous study using the Quebec Task force classification
, and in studies that demonstrated that the presence of a positive straight leg raise identified patients with a characteristic profile
Comparing people with pain above the knee and those with pain below the knee revealed higher leg pain intensity and more activity limitation in the group with more distal pain. There were little differences in other parameters according to location above and below the knee, although generally patients with LBP + pain below knee in this secondary care setting tended to be more severely affected than those with LBP + pain above knee as previously observed in other settings
LBP with any leg pain tended to be more severe than local LBP only, regardless of how severity was measured, and was also associated with higher frequencies of psychological risk factors. This does not clarify any causal direction and it may simply be that more distal pain actually is experienced as more severe and results in heavier psychological distress, and therefore these are mutually interdependent covariates. The relationship between pain and depression was demonstrated to be reciprocal in a study by Kroenke
, but the potential causative influence of leg pain on psychological risk factors or the reverse needs to be investigated, including potential differences in this relationship between leg pain subgroups.
The main strength of this study was the large sample size which is likely to produce more trustworthy estimates. Moreover, we expect the results to be generalisable to other secondary care settings since data were collected as part of the everyday routine of the department and not constrained by the strict inclusion criteria that are often necessary in clinical trials. The downside of using registry data is less control of data quality and so, in order to describe subgroups with very clearly defined profiles, we took the approach of excluding patients who reported any ambiguous data. By so doing, we chose to maximise the ‘signal’ in data (clear subgroups) by reducing the ‘noise’ (ambiguous data).
A limitation of this study was the definition of nerve root involvement. Classification into the group with signs of nerve root involvement required the presence of just one positive finding in the neurological examination, and the reliability of these findings in this setting is unknown. The unquantified reliability of our examination procedures is an inherent limitation of data from large clinical databases that were not collected for a specific research project. However, the data were also inherently more representative of routine care and our results demonstrate that even this ‘loose’ definition of signs of nerve root involvement defines a distinctive subgroup. It is possible that our definition of ‘nerve root involvement’ led to underestimation of subgroup differences, but the findings indicated that even very simple clinical examination data can add valuable information.
Another potential weakness of this study is that the inclusion criteria and the gradual implementation of the electronic database over the study period meant that only about 25% of the patients seen during this time period were described in this study. All known selection biases have been reported. There were no clinically important differences on gender or age between those who filled in the electronic questionnaires and those who did not. People who could be unambiguously allocated to the defined subgroups had slightly more severe symptoms and more disability than those who could not be allocated. This bias towards those with a more severe profile means that the proportions allocated to each group may not truly reflect the whole patient population.
A further weakness is that some constructs, notably the psychosocial constructs, were measured using screening questions only, though such an approach has been used in other contexts including in the Bournemouth Questionnaire
. The screening questions used had high overall accuracy (86.8% to 88.0%) for predicting the scores on the reference standard questionnaires for depression and fear avoidance [unpublished data]. The main reason for using screening questions is that it is simply not possible in our setting to collect comprehensive psychological data on all patients as part of a multi-domain assessment procedure. Our experience suggests that we are at the limit of responder burden that is acceptable to our patients.
Lastly, it should be recognised that some differences between subgroups may have been a result of referral patterns. For example, patients in the cohort with local LBP only had longer pain duration when seen at the department which was probably because leg pain prompts referral at an earlier stage than local LBP.