Acute low back pain is marked by variability: An internet-based pilot study

The terms 'acute' and 'chronic' are commonly used in both research and clinical practice to characterize low back pain (LBP) and its prognosis. These terms as used in back pain research have been criticized, however, on the grounds that they overemphasize symptom duration and fail to capture important information about the experience of LBP and back-related disability[1–3]. Prospective studies have demonstrated that the experience of back pain is characterized by change and variation- through recurrences and remissions across pain episodes, and through flares of pain within pain episodes[3, 4]. The fluctuating nature of back pain is often observed by clinicians caring for back pain patients, who are well familiar with individuals who return to clinic intermittently with flare-ups of pain during recent-onset, chronic, or recurrent LBP.

Flare-ups (or 'flares') of back pain have been defined as 'a period when back pain is markedly more severe than is usual for the patient'[1]. Flares of back pain in individuals with 'chronic' symptoms (>3 months) are associated with greater disability and work absenteeism, when adjusting for important factors, including pain intensity[5, 6]. Pain variability therefore may be an important component of the pain experience that is related to disability, yet is distinct from pain intensity alone. Pain that is highly variable would be expected to have periods of relatively higher pain intensity (flares) and periods of relatively lower pain intensity; such fluctuations in pain intensity may increase the unpredictability of the pain experience, and may make pain more bothersome than pain that is stable and predictable. Nevertheless, pain variability in the acute LBP period (<3 months) has received limited study, with the majority of prior cohort studies not reevaluating patients until 1-3 months after the time of seeking care[7]. It is unknown whether discrete flares of pain from the course of acute LBP can even be discerned by patients in the context of the rapid improvement in pain intensity typical for acute LBP, which is distinct from the more stable baseline pain intensity expected in chronic LBP. A major obstacle to the study of flares and fluctuations of acute LBP is the logistical problem of how to go about documenting them. Even in the context of chronic LBP, flares of pain often last less than three days, and may recur frequently[6]. This creates a need for frequent assessments in order to 'capture' fluctuations of pain that may be recalled inaccurately- or not at all- when assessed retrospectively weeks to months later. However, frequent serial assessments using mailed questionnaires or telephone interviews has the potential to be both burdensome for participants to complete, and infeasible for researchers to administer.

To our knowledge, only a single study of acute LBP has employed frequent serial assessments permitting a characterization of pain fluctuations during the early course of acute LBP[8]. Sieben et al. followed 44 patients with acute LBP for the first 14 days after seeking care, in order to study changes in fear-avoidance beliefs during this period. Although this study demonstrated that flares of pain do exist in acute LBP, the authors did not characterize the frequency of flares, their occurrence over a more extended course after the first 14 days, or factors associated with flare occurrence. Furthermore, the use of written pain diaries to assess pain fluctuations in this earlier study posed a problem, since participants were not blinded to prior assessments. This had the potential risk of 'smoothing out' of variability due to the participant's ability to access and repeat the values used for earlier pain ratings. In addition, Sieben et al. did not examine associations between pain fluctuations and disability outcomes. This is important, because although much prior research has examined the prognostic utility of factors in predicting disability outcomes after acute LBP[9], no prior study has examined the role of acute pain variability on the transition to chronic back-related disability.

We proposed a pilot study to examine the feasibility of measuring pain variability and 'flares' of pain during an acute LBP episode, using a cohort design with frequent serial assessments. We utilized Internet-based data collection specifically to improve the feasibility of the sampling scheme. The objectives of this pilot study were to characterize the fluctuations in pain intensity that occur during the course of an acute LBP episode, and to examine whether self-reported 'flares' of pain during acute LBP represented discrete periods of increased pain intensity which could be distinguished from periods when not in flare. Furthermore, we examined whether the frequency of flares was associated with back-related disability six weeks after seeking care for LBP, after accounting for other factors related to disability. In the course of this study, we evaluated the feasibility of maintaining high levels of subject compliance with frequent serial Internet-based assessments, during an acute pain episode.

Pain flares were reported by 42 of 47 participants (89%) over six weeks of follow-up. 148 discrete pain flare periods were reported. Among those participants who reported pain flares, the average number of discrete flare periods was 3.5 (range 1-11) over the course of the study. Figure 1 depicts the frequency of flares of different duration. More than half of reported flares were less than 4 hours in duration, and about 75% of flares were less than one day in duration. Of the 23% of flares that were more than 24 hours in duration, the mean ± SD length of a flare was 8.1 ± 8.9 days.

At six-week follow-up, mean ± SD back pain (1.4 ± 2.0) and disability scores (15 ± 16) demonstrated dramatic improvements over the severity of the clinical presentation at time of seeking care. Linear spline models using knots at 1-week intervals suggested that the majority of back pain improvement occurred over the 1^st week after seeking care, suggesting that a model assuming a linear time effect would not be appropriate. However, a model including a quadratic time trend (i.e., including terms for both time [days since seeking care] and time² since study entry) showed improved model fit (AIC = 2364) over a linear trend model (AIC = 2445), and over linear spline models with knots at one-week intervals (AIC = 2521), and with a single knot at one week (AIC = 2518). This quadratic trend over time indicated that pain intensity decreased rapidly during the first 14 days after seeking care (49% improvement from baseline), decreased at a lesser rate during the next 14 days (71% improvement from baseline), and leveled off after about 28 days; the predicted course of acute LBP intensity produced by this model is depicted in Figure 2. A model with a quadratic trend for time also produced slightly improved model fit over the linear model when reported flare periods were excluded from the analysis (data not shown). A model with a quadratic trend for time was therefore used in subsequent analyses to examine longitudinal relationships between patient-reported flares and pain intensity

Next, we used linear mixed-effects regression to model pain intensity over the six-weeks of follow-up as a function of time (days since seeking care), time², and current flare status. Time (ß [SE] -0.091[0.014]; p < 0.0001) and time² (0.0013 [0.0003]; p < 0.0001) were highly significantly associated with longitudinal pain intensity. A self-report of a concurrent pain flare was also associated with a large marginal increase in current pain intensity (ß [SE] 2.70 [0.11]; p < 0.0001). That is, at any time during follow-up, patients who reported a flare experienced an approximate 3-point increase in pain intensity (on a 0-10 scale) as compared to those who did not experience a flare. This flare 'effect' was independent of the curvilinear trajectory of improvement imposed by the quadratic trend for time. Longitudinal changes in ODI score closely paralleled the course of pain intensity in acute LBP (Additional File 2).

Multiple, discrete flares of increased pain were experienced by nearly all individuals recovering from acute LBP. Although the majority of reported flares lasted less than 24 hours, almost one quarter of flares lasted about one week. The time course of back pain improvement was curvilinear, with the majority of improvement occurring in the first two weeks after seeking care, and a leveling off of improvement after 4 weeks. However, despite the fact that pain levels were constantly changing during acute LBP, individuals were able to identify pain flares that corresponded to statistically significant increases in pain intensity as compared to other individuals not reporting flares, with a flare 'effect' of almost 3 points on the NPRS at any time during follow-up. This flare effect exceeds the threshold of 2 NPRS points commonly accepted as a minimal important change in NPRS[20]. The high degree of variability typical during acute LBP is difficult to accurately conceptualize using descriptions of flare frequency and magnitude of common pain intensity changes; Figures 3 and 4 present two examples of individual patient data that better convey the degree of variability seen during acute LBP.

Our study findings fit well into the context of prior cohort studies of acute LBP, which have shown dramatic improvements in pain at 1-2 weeks after seeking care[18, 19]. Sieben et al. demonstrated a rather gradual course of improvement in pain intensity over 2 weeks after seeking care for many individuals, although fluctuations in pain were noted, and individuals with increasing levels of pain-related fear did not have decreases in pain[8]. Our study confirms that most patients have a rapid improvement in pain intensity, but not to complete resolution. Furthermore, flares of pain occur even after pain has largely improved, and for many individuals, multiple discrete episodes of transient pain worsening occur. Since early improvement is accompanied by the potential for brief returns to higher-intensity pain levels, individuals with recently improved pain intensity may be hesitant to advance function immediately after pain becomes more tolerable. A possible connection between flares and disability is supported by the results of our secondary, cross-sectional analysis, which found that individuals who had a higher flare frequency had greater disability at study end, irrespective of other factors including baseline disability or duration of symptoms.

A recent study of chronic LBP flares found that the presence of flares was associated with higher levels of disability even when adjusting for demographic factors and pain intensity[6]. Independent of other factors, individuals with flares were found to have high levels of maladaptive coping, which is an important predictor of long-term disability in LBP[9]. Although coping was not assessed in the current study, our finding of an association between higher flare frequency and disability raises the question of whether pain variability and maladaptive coping are related in the acute LBP period. Such a relationship could have a number of possible explanations. Flares- which occur when pain is variable- may be a pain experience that reinforces maladaptive coping or fear-avoidance. Indeed, it seems logical that pain variability would lead to less predictable pain, posing a greater problem for those with passive coping rather than adaptive coping skills. Alternatively, individuals with poor coping skills may be predisposed to experiencing greater pain variability for other reasons. Despite much research into factors that may predict the acute-to-chronic transition in LBP, no prior research has investigated pain variability as a prognostic factor. Future prognostic research should examine pain variability, as well as interrelationships with other known risk factors for chronicity, including maladaptive coping.

This pilot study demonstrates the feasibility of maintaining subject compliance with repeated data collection during an acute pain episode. Compliant subjects were more likely to be female and more likely to be working, but were otherwise not materially different from non-compliant subjects. However, missed questionnaires and late-completion questionnaires did occur, creating a potential for recall bias. Future studies may benefit from measures to improve compliance with questionnaire reminders utilizing modes of communication other than the Internet, such as phone calls or text message reminders. In addition, more sophisticated measures may be employed to limit recall bias, such as application of stringent, automated time constraints for questionnaire completion.

There are several limitations of this study that deserve mention. First, our use of frequent serial assessments and the fact that this was a pilot study required that this be a relatively small sample of patients. Second, some concerns apply regarding generalizability. Although these participants were recruited from a specialty spine clinic, some assurance of generalizability to primary care settings is provided by similarities between our study findings and those of the Sieben, which examined acute LBP with frequent serial assessments in a primary care setting. In addition, we included only nonspecific LBP, excluding cases of LBP with more specific causes such as radiculopathy, symptomatic lumbar spinal stenosis, trauma, infection, malignancy, fracture, and pregnancy; our findings may not be generalizable to acute LBP related to these causes. The use of the Internet raises a separate generalizability concern, but a growing literature suggests comparable reliability, internal validity, and external validity with use of the Internet as compared to conventional modes of data collection [21]. Although Internet-based data collection is not yet widely utilized in epidemiologic research, it offered several major advantages for the conduct of our study, given the frequent sampling employed, and the short time intervals between questionnaires. Internet-based questionnaires are returned more rapidly than postal questionnaires[22], which allowed us immediate notice of questionnaire completion, and the ability to send reminders for late questionnaires. Although telephone interviews might also have permitted this, Internet-based questionnaires take less time to complete[23], and minimize social acceptability bias. As previously mentioned, we wished to avoid the lack of blinding to earlier responses inherent in the use of written pain diaries. The use of electronic hand-held pain diaries may have provided additional benefits over our Internet-based approach, and future replication of our findings using these devices would be recommended. Third, as we mentioned, participants were allowed late completion of questionnaires more than 24 hours after notification was sent to complete the questionnaire. We made a pragmatic decision to allow this when planning the study, because we had no preliminary data to inform as to the feasibility of conducting frequent serial assessments of subjects in acute pain over the Internet. A prior electronic diary study reported that missed data points were not significantly associated with fluctuations in pain, mood, or stress[24], but were best explained by participants missing reminders to complete entries. This notion is supported by our finding that participants with fewer missed questionnaires were more likely to be working, and therefore to have access to email reminders. However, our study provides preliminary evidence of the feasibility of utilizing the Internet to closely follow patients with acute musculoskeletal pain. A final limitation of this study is that psychosocial factors including mood and coping, as well as objective functional measures that may have contributed to flares, were not assessed.

Despite these limitations, this study demonstrates that the course of acute LBP is characterized by variability. Patients with acute LBP often have multiple distinct flares of pain in the six weeks after seeking care, and these flares correspond to discrete increases in pain intensity relative to periods not in flare. Future research should examine possible relationships between pain variability and long-term outcomes, as well as identify those demographic and psychosocial factors that are related to pain variability.