The Nordic back pain subpopulation program: Can low back pain patterns be predicted from the first consultation with a chiropractor? A longitudinal pilot study

Much has been written on non-specific low back pain (LBP) in the scientific literature. Presently, however, there are no easy answers to the clinicians' questions on how best to treat this condition; it seems that a number of different treatments have an effect, but only to a very limited degree [1‐3]. In an attempt to break the stalemate, a number of researchers have shown an interest in the study of subpopulations of LBP [4‐8] and preliminary results suggest that classification-based interventions are more effective than treatments directed towards mixed populations with non-specific LBP [9].

Different approaches exist to identify specific profiles of patients within the amorphous definition of non-specific LBP. Clinicians typically attempt to detect the pain generating structure and classify their patients accordingly into diagnostic subgroups. This information is then used both to determine the most relevant type of treatment and to predict outcome of treatment (prognosis). Such a pathoanatomical classification has also been suggested by some researchers [10] whereas others have focused on single clinical features [11, 12] or clusters of characteristics that are predictive of response to treatment [7, 13‐16]. However, it is a challenging task to validate any classification system, as it would be necessary to test whether the base-line features actually make a difference to the outcome and if this difference is related to specific treatments. The latter would have to be done in randomized trials designed specifically for the purpose of subgroup identification [5].

In randomized trials, the outcome is typically calculated as the difference between the patients' status before and the status after treatment. However, LBP is a fluctuating or episodic condition for many [17‐20]. These fluctuations occur even within a few months and have been shown to have varying patterns [19, 20]. Therefore, it may not be relevant to measure outcome solely at one specific point in time, such as after 3 or 6 months, as there is no obvious end-point for LBP. However, presently this is how outcome of treatment for LBP is measured in clinical studies. A better outcome measure would rather be one that takes into account the course of pain over the post-treatment surveillance period.

Presently, very little is known of what happens between the time, when a patient seeks care, and when the final outcome is measured. However, with the advent of a new method to collect data using mobile phones, study subjects can be surveyed at frequent and regular intervals with the help of automatically generated text messages. This makes it possible to identify course patterns rather than end-point outcome thus approaching this problem from a different angle. Both diagnostic subgroups and other clinical characteristics could be held up against the clinical course, in order to see if they represent clinically relevant subgroups. After all, what matters for the patient is probably rather the every-day events than the arbitrarily selected point of outcome 3, 6 or 12 months after treatment took place.

For these reasons, a practice-based pilot study was performed, in which clinical data were collected at base-line and over a period of 18 weeks as continuous follow-ups by means of weekly text-messages. The rationale for the study was that clinical observations at the first consultation for an event of LBP would predict the ensuing course pattern. We have previously reported that improvement occurred early in the course [21] and that different course patterns existed within this study population of patients who were treated by chiropractors for a new event of LBP [20]. The objectives of the present report are 1) to get a feel for the inter-observer reliability of a diagnostic classification system [10], and 2) to investigate whether patients with different clinical profiles have different course patterns or different prognoses in terms of number of LBP days over a period of 18 weeks.

This appears to be the first study to compare baseline characteristics of LBP patients to pain patterns generated by very frequent follow-ups over a period of time. Moreover, it was the first attempt to study whether the prognosis of primary care patients with LBP is related to diagnostic classes as defined by the classification system described by Petersen [10]. Although we had a relatively small study sample, and a large number of subgroups, it was still possible to obtain some useful information.

First, it appears that at least some of the diagnostic classes relate to the prognosis. Patients classified as having disc-related pain reported more pain days and were less likely to experience the pain course 'mainly recovered' than others. Patients with disc pain had on average between 13 and 19 more days with pain than patients with muscle pain, mechanical dysfunctions, or SI-joint pain. It would be relevant to investigate such differences in more depth including whether diagnostic classes differ not only regarding pain, but also in relation to activity of daily living or disability. If similar associations between diagnosis and prognosis are confirmed by other studies, the differences are large enough to be important to patients and indicate that this classification system makes a distinction between relevant subgroups of patients.

In accordance with previous studies [26] men had a better prognosis than women. They had fewer days with LBP in total, were more likely to undergo the course pattern 'mainly recovered', and seemed to have less fluctuating patterns than women. The present results suggest that this could be, at least partly, explained by the difference in diagnostic classes between men and women, since men were less often classified with disc pain than women were. In addition, age was related to outcome patterns in the way that young patients had a milder course than older. The present cohort was not large enough to explore in more detail whether certain pain patterns relate to each gender or certain age groups and this should be explored in larger studies. In accordance with previous cohort studies on chiropractor patients [15], but maybe surprising to many clinicians, the duration of the present LBP episode was not associated to any of the outcome measures.

Because of the small numbers within each diagnostic class, statistical testing in relation to agreement was unworkable and the agreement was therefore only evaluated in percentages that do not take into account agreement by chance. The agreement concerning the diagnostic classes was high when based on the most manifest class, and markedly lower if absolute agreement was demanded. However, we consider the obtained agreement sufficient for the classification to be meaningful. The reliability of the classification system was tested in a set up with two chiropractors being present at the same consultation. This could have introduced bias toward higher agreement, but was chosen to avoid an altered symptom response at the second examination. The same decision was made in earlier studies [25, 27]. The agreement on all classes was high (54%) as compared to a previous study on this classification system [25] with 34% agreement. The main difference, between the methods of the previous study and our, was that we allowed the use of more than one of the diagnostic classes. In the original study the classes were described as mutually exclusive. Therefore, in our study, it was possible to use more classes instead of making a compulsory final choice between two seemingly relevant classes. This approach seems reasonable because pain can be generated from more than one structure. We are aware of studies concluding that disc pain very seldom coexists with facet joint or SI-joint pain [28, 29] and that pain is not likely to originate from both facet- and SI-joints at the same time [29]. However, these studies included only few patients who were not recruited from primary care, and in our analyses only one class was included in the analyses, consistent with the intention of the classification system.

The main limitation of this pilot study was the relatively large drop out from follow-up. As discussed in previous papers [20, 21] this was in line with other primary care studies in which patients were followed up less frequently [16, 19]. Fortunately, baseline characteristics in those who dropped out resembled those of the compliant patients. We suppose that a more enthusiastic information strategy directed to the participating patients could have helped maintaining the interest of the patients.

As a consequence of the quite small cohort we chose to pool the three disc classes from the original classification system into one. This may limit the prognostic value of the classification since we did not distinguish between mechanically reducible and irreducible discs, i.e. pain that can be centralized and pain that cannot, which is known to be of predictive value [30‐32].

In conclusion, our results suggest that different diagnostic classes have different pain courses and indicate that patients with different low back conditions can be identified through the physical examination. The next step will be to perform a large-scale practice based study with a sufficient number of patients to make it possible to include more of the diagnostic classes and evaluate prognosis within each of these.