Participants

Selected chiropractors in private clinics in one Danish region were invited to participate in the recruitment of patients. Inclusion criteria were: a new episode of LBP with or without sciatica as main complaint (i.e. the patients had not seen the chiropractor for this specific pain episode previous to inclusion), 18 - 65 years, and having a mobile phone. The non-inclusion criteria were: previous back surgery, pregnancy, other significant musculoskeletal problems in addition to the LBP, and inability to read or speak Danish. Prior to inclusion, patients received written and verbal information about the study. The project was presented for the local ethical committee who stated that it did not need approval.

Clinical procedures

Patients who agreed to participate had a standardised clinical examination. Based on the examination they were classified according to a diagnostic system [9]. According to this system, the possible diagnoses were disc pain, nerve root compression, spinal stenosis, postural syndrome, mechanical dysfunction, sacroiliac joint pain, facet joint pain, abnormal nerve tension, muscle pain and abnormal pain syndrome. Information regarding symptoms duration of the present episode and localisation was collected during the patient history at the first consultation. Further data on aspects of the pain course in relation to mechanical diagnosis and other baseline characteristics will be presented elsewhere. Chiropractors were free to choose whichever treatment they found appropriate.

LBP registration

Follow-up was conducted by text messages that were automatically sent by a system marketed as "SMS-track"[10]. One SMS (short message service) was sent for each question and the participants replied to the questions by returning a text message. The replies were incorporated into a data file on a server at the SMS-track supplier's office. Follow-up was initiated on the first Sunday following inclusion and thereafter automatically repeated every Sunday for 18 weeks. An automatic reminder was sent if the text message had not been answered on the first coming Thursday. Every week the patients were asked the following questions:

Question 1: 'Please answer how much your lower back hurts today? Choose a number: 0 = no pain at all/1 = some pain/2 = severe pain'. (Referred to as LBP-intensity)

Question 2: 'Using a number from 0 to 7, please answer how many days you have been bothered from your lower back this week'. (Referred to as LBP-days)

Question 3: 'Using a number from 0 to 7, please answer how many days you have been off work because of your lower back this week. (Answer with X if you are not working)'

Data analysis

The information from the text messages was automatically incorporated into a spread sheet and afterwards transmitted to STATA 10.1. When answers other than a number were given, data were manually recoded as a number when possible, e.g. "I have no pain" was recoded as 0, and "2 days last week" as 2. Answers that could not be transformed directly into a number were coded as missing values. Data were only included in the analysis from those who had participated at least until the 12^th week. Participants were also excluded if they did not participate for three or more weeks in a row during the trial. Since days off work due to LBP (question 3) were infrequent no analysis was made for that item.

Frequency tables were constructed for each of the 18 weeks in relation to LBP-days (answers from 0 to 7). These data were then reduced to three categories: 0 days, 1 - 5 days, and 6 - 7 days, which were transformed into bar graphs and used to visualize the point in time when changes in the number of reported LBP days took place. The reduction of LBP-days into the three categories was done in order to isolate those fully recovered and those with a definite problem, and was supported by the raw plots that indicated that this breakdown would form three distinct groups of patients. The LBP-intensity variable was handled in the same manner, though not reduced into fewer categories.

Participants

Low Back Pain

What is the general development of LBP? - A comparison between week one and eighteen

Fig. 1 shows that the most frequent answer was seven days with LBP the preceding week at the first follow-up, and at the end of follow-up the most frequent response was no days with LBP the preceding week. At the first visit, "some pain" was the LBP-intensity most frequently reported, whereas no pain was the most common response at the end of week 18 (Fig. 2). Three participants reported to have had no days with LBP during the first week.

What are the proportions of patients who are recovered, mildly or severely affected throughout the 18 weeks course?

When grouping participants into those reporting no days of LBP, those with 1 - 5 days and those with 6 - 7 days of LBP, it was observed that the number of participants classified as reporting no days of LBP increased until week 10 and tended to decrease again after the 12^th week (Fig. 3). The highest frequencies of pain free participants were in weeks 10 and 12, when 53% reported zero LBP-days. The number of participants with LBP for 6-7 days a week decreased most rapidly until week four, was reduced a little further until the 7^th week, and remained relatively stable after that with the lowest frequency from week 12 to week 14. In relation to pain intensity there was a very similar pattern with the proportion of patients reporting no pain increasing until week seven and remaining relatively steady after that; again with a small decline at the end of the period. In the first week, 16% reported severe pain, and that proportion declined during the first 4 weeks, after which it remained small (Fig. 3).

Results of the study

This appears to be the first study in which weekly follow-ups were performed over a prolonged period of time in patients seeking care for LBP, and thus the first attempt to make a detailed description of the course of LBP following treatment.

We found that the general development of LBP during 18 weeks was improvement both in relation to the number of LBP days in the past week and pain intensity on the day of the follow-up, which resembled each other closely. When interpreting these results it should be noted that "week 1" was not a baseline score, but the reporting of symptoms the first Sunday following the first consultation, i.e. usually after treatment had been initiated. At the beginning, daily LBP was most frequently reported with a gradual shift to no pain days at the end of the 18 weeks. In relation to pain intensity, "some" present pain was by far the most frequent answer in week one, whereas no pain was most frequently reported at the last follow-up.

It was hypothesized that patients seek care when their symptoms are at a peak, and that they therefore will improve in the subsequent period. This hypothesis was supported by the fact that a very quick improvement was observed until week four, which was in line with previous cohort studies on chiropractor patients [4, 5]. Whether LBP patients seeking chiropractic care improve rapidly because of or regardless of treatment can of course not be determined with this type of study design.

Our data also indicate that on a group level no further improvement should be expected later than week seven after treatment was initiated. Further, our results tend to support that LBP is a recurrent condition since a slight increase in pain days and pain intensity was observed again after the 12^th week. However, a longer follow-up period would be necessary to determine an exact point of time when a possible worsening should be expected to occur.

The highest frequency of being pain free was reached in week ten, when 54% reported no LBP-days, but about half of the patients then keep on experiencing some LBP on and off, and hence do not report complete recovery within a course of 18 weeks.

Methodological considerations

It was a limitation of the study that we were only able to achieve a 69% response rate at the end of the follow up, but as compared to other primary care studies, we considered this acceptable [2, 11, 12]. Compared to patients in the secondary sector it may be difficult to motivate primary care patients to spend the time to participate in prospective studies since they, generally, are less troubled by their LBP. Those who dropped out from the study were more frequently men and had a longer duration of symptoms prior to seeing chiropractic care. Age, pain location, LBP-days the first week and LBP-intensity the first week did not differ between the study population completing the study and drop outs. Nonetheless, the longer duration of the current episode in those who dropped out may have affected results although the association between this factor and the prognosis is uncertain [13]. It is possible that a more vigorous information strategy would have helped maintain the interest of the participants throughout the entire study period.

Unfortunately, we did not register how many patients declined to participate or if some potential participants were not invited, and hence we do not know to what extent our results can be generalized to all chiropractic patients. In retrospect, the participating chiropractors estimated that no more than ten patients refused to participate and that only two persons were excluded because they could not use the SMS function. It is also not possible to perform a comparison between the baseline status of this population and other populations treated by chiropractors since we did not collect any pain scores prior to the first treatment. This is a shortcoming in relation to describing the profile of the populations, but did not weaken the answering of our objectives.

The main limitation of the SMS-track method is that only few and simple questions can be presented to the participants at follow-up. In the present study we chose to ask about number of pain days, present pain intensity and number of days sick-listed. We found that sick-listing was not a suitable measure in this population since only few patients had any days with sick-listing. This question could therefore be exchanged for a question on disability, which would provide a more comprehensive picture of the LBP status.

We did not try to fit the curves with any statistical model; instead the course of pain was described by the authors simply from what was visualized in the presented figures. These curves could perhaps be interpreted somehow differently by others, but we have previously shown that it is possible to agree well on visual analyses of individual LBP patterns [14]. Further, statistical methods to identify the shifts in the LBP course were not considered useful since this would be subject to large uncertainties as well with such few observations. Also, this was an initial study intended to be a first step in developing a method for investigating LBP as a fluctuating condition. In that context we find this pragmatic approach relevant, but future full-scale studies should evaluate LBP patterns by established statistical methods for this purpose.

Recommendations

In relation to follow-up studies concerning primary sector LBP care in which traditional questionnaires are used, we would recommend that the first follow-up takes place in week seven after treatment was initiated to ascertain the short-term level of improvement and around week 12 to observe for early recurrence. Obviously, further knowledge is needed in relation to the need for further follow-up after 12 weeks.

In clinical practice we recommend that patients' LBP status is systematically followed for the first four weeks since fast improvement is expected during that period. Further, the absence of early improvement was previously observed to be associated with a poor long-term outcome [4, 12] and clinicians should be aware that no further changes in LBP days or intensity happen later than week seven on a population level. Concerning the timing of secondary prevention, we cannot make any recommendations. Our results indicated that some patients have a recurrence of symptoms around week 12, and it should be explored further whether there is a certain time following a LBP episode when patients are at risk of recurrence and whether any preventive efforts can hinder this.

One should note that recommendations based upon this study apply to a group level. It is necessary to study individual pain patterns in order to identify potentially relevant sub-groups within LBP with different responses to treatment and different pain courses. Such individual patterns within the population reported on here are presented elsewhere [14].