Intended for healthcare professionals

General Practice

Clinical course and prognostic factors in acute low back pain: an inception cohort study in primary care practice

BMJ 1994; 308 doi: (Published 26 February 1994) Cite this as: BMJ 1994;308:577
  1. J Costea,
  2. G Delecoeuillerie,
  3. A Cohen de Lara,
  4. J M LeParc,
  5. J B Paolaggi
  1. a INSERM Unite U 292, Hospital de Bicetre, 94275 Le Kremlin-Bicetre Cedex, France
  2. Laboratories CASSENNE 1,92800 Puteaux, France Service de Rhumatologie, Hospital Ambroise Pare, 92104 Boulogne Cedex, France
  1. Correspondence to: Dr
  • Accepted 11 January 1994


Objective : To describe the natural course of recent acute low back pain in terms of both morbidity (pain, disability) and absenteeism from work and to evaluate the prognostic factors for these outcomes.

Design : Inception cohort study.

Setting : Primary care.

Patients : 103 patients with acute localised non-specific back pain lasting less than 72 hours.

Main outcome measures : Complete recovery (disappearance of both pain and disability) and return to work.

Results : 90% of patients recovered within two weeks and only two developed chronic low back pain. Only 49 of 100 patients for whom data were available had bed rest and 40% of 75 employed patients lost no time from work. Proportional hazards regression analysis showed that previous chronic episodes of low back pain, initial disability level,initial pain worse when standing, initial pain worse when lying, and compensation status were significantly associated with delayed episode recovery.These factors were also related to abseteeism from work. Absenteeism from work was also influenced by job satisfaction and gender.

Conclusions : The recovery rate from acute low back pain was much higher than reported in other studies. Those studies, however, did not investigate groups of patients enrolled shortly after the onset of symptoms and often mixed acute low back pain patients with patients with exacerbations of chronic pain or sciatica. Several sociodemographic and clinical factors were of prognostic value in acute low back pain. Factors which incluenced the outcome in terms of episode recovery (mainly physical severity factors) were only partly predictive of absenteeism from work. Time off work and return to work depended more on sociodemographic and job related incluences.

Clinical implications

  • Clinical implications

  • The prognosis of acute low back pain is generally thought to be good, but chronic low back pain is common

  • Data on the natural course and prognostic factors in acute low back pain are fragmentary

  • In this inception cohort study recovery from episodes of acute low back pain was more rapid than previously described: 90% of patients recovered within two weeks and fewer than 2% developed chronic low back pain

  • Previous episodes of chronic low back pain and factors related to severity seem to strongly influence the recovery from episodes; conversely, absenteeism from work seems to depend more on sociodemographic and job related factors

  • These results will be useful in identifying patients at risk of a poor outcome and as an aid to more appropriate randomisation in controlled trials.


Low back pain is a common disorder with major consequences for health care resources.1,2 Though the prognosis of acute low back pain is considered to be good,3 chronic low back pain is very frequent. Identifying factors for chronicity and disability is required to allow adequate care 4 and may also be useful for assessing treatments.5,6 No satisfactory indicators for prognosis have been identified in studies in occupational settings which used lost work time to evaluate outcome.4 Moreover, 16-40% of patients lose no time from work after a back injury, even if work related.7,8 Studies from general practice are few.*RF 9-11* Definitions of acute attacks of pain in these studies lacked precision: patients with pain lasting 7-30 days or associated with sciatica were not excluded. Outcome indicators were also crude (for example, lost work time) and psychosocial factors were not addressed.

We studied acute low back pain patients in primary care (a) to investigate the natural course in terms of morbidity and absenteeism from work, and (b) to identify the clinical, psychological, and sociodemographic factors with prognostic value.

Patients and methods

All consecutive patients aged 18 and over, self referring to participating doctors (n=39) for a primary complaint of back pain between 1 June and 7 November 1991 were eligible. Only patients with pain lasting less than 72 hours and without radiation below the gluteal fold were included. Patients with malignancies, infections, spondylarthropathies, vertebral fractures, neurological signs, and low back pain during the previous three months were excluded, as were non-French speaking and illiterate patients. The resulting study group was 103 subjects.

Doctors received training in clinical and psychiatric evaluation before the study. Clinical data collected at the time of the first visit included sociodemographic and occupational characteristics, compensation status (which is temporarily but invariably awarded in France for any pain episode occurring in the workplace), medical and surgical histories, pain intensity (on a visual analogue scale), type of onset and duration, aggravating and relieving factors, 12 assessment of lumbar movements, and the straight leg raising test.13 Current psychiatric symptoms were investigated by using a structured psychiatric interview based on DSM-III-R (Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised) classification flowsheets.14,15 Patients who could stand (ambulant patients; n=85) filled in a validated French translation 16 of the Roland and Morris disability questionnaire. 17,18

We did not include any radiological investigations because of the unreliability of interpretation19,20 (radiographs helped only in diagnosing “specific” low back pain).

To optimise description of the natural course of the low back pain episode doctors prescribed oral analgesics containing only paracetamol. Prescription of bed rest and sick leave was left to the discretion of the doctors.

Patients completed a diary every evening from day 1 to day 7. It included a visual analogue score for mean back pain for the day, the disability questionnaire, and the time spent in bed. Follow up visits were scheduled on day 8 and, while back pain or disability persisted, on days 15,30,60, and 90. The data collected at each visit included the patient's evaluation of pain and disability. The dates of recovery, defined as the disappearance of both pain and disability, and return to work (in cases with sick leave) were recorded.

Evolution of pain and functional disability were described by using means and standard errors. The two main outcomes of the study - recovery and return to work - were assessed by life table analysis according to Kaplan-Meier. The prognostic values of factors on these outcomes were tested by log rank methods. Proportional hazards models were fitted to study factors simultaneously and to adjust for the potential confounding effect of pain duration at entry by using a forward stepwise procedure (enter P value=0.05, remove P value=0.10).21 For each factor in the final model the hazard ratio and 95% confidence interval were calculated. (The hazard ratio may be interpreted as the relative risk of recovery at any measurement within three months.) Assumptions of proportional hazards were checked from plots of log minus log (survival) functions against time.


Baseline characteristics of the 103 patients are shown in table I. There were 11 drop outs, who were similar as a group to the 92 patients remaining for follow up with regard to all characteristics (data not shown).


Baseline characteristics of subjects (n=103) at entry to study. Except where stated otherwise, values are numbers (percentages) of subjects

View this table:

The evolution of pain and disability during the first week is shown in figure 1. There was a large decrease in pain every day until day 4 and smaller decreases thereafter. The proportion of subjects unable to stand and the disability score followed similar patterns.


Evolution of pain (visual analogue scale scores) and disability (proportion of patients unable to stand, Roland and Morris disability scale scores) during first week of follow up. Plots shown as means and 95% confidence intervals

Figure 2 shows the cumulative rates of recovery and attendance at work among employed patients. The median duration of episodes was 7 days, and 90% of patients recovered within the first two weeks (95% confidence interval 84% to 96%). Only two patients (1.9%; 0 to 4.7%) did not recover during the three month period and developed chronic low back pain.One other patient presented with sciatica at day 15 (1.0%; 0 to 2.7%). Only 49 of 100 patients for whom data were available had bed rest (table II). The distribution of numbers of days in bed was trimodal, with maxima at no days, 2-3 days, and 7-8 days. Forty per cent of employed patients lost no time from work. Return to work was slower than recovery (fig 2).


Life table analysis showing cumulative rates of recovery and attendance at work (for employed patients) over three month follow up (Kaplan -Meier method)


Distribution of reported numbers of days of bed rest Reported days of bed rest

View this table:

Associations between various factors and recovery were tested (table III) and a prognostic recovery model constructed (table IV). Previous chronic low back pain was associated with a fourfold lower probability of recovery; pain worse when standing or lying, disability at entry, compensation status, and employment status were also predictive.


Prognostic factors for recovery among subjects with acute low back pain (n=103)

View this table:

Final prognostic model for epidose recovery: hazard ratios(dagger) by proportional hazards model(double dagger)

View this table:

Factors were tested for association with lost work time (table V). The same set of variables (previous chronic low back pain, pain worse when standing or lying, disability at entry, and compensation status) plus male sex and low job satisfaction were predictive (table VI).


Prognostic factors for attendance at work among employed subjects with acute low back pain (n=75)

View this table:

Final prognostic model for attendance at work (for those patients with occupation; n=75): hazard ratios(dagger) by proportional hazards mode(double dagger)

View this table:


Data on the natural course of low back pain are fragmentary.22,23 Other studies did not investigate patients enrolled shortly after the onset of symptoms and often mixed acute low back pain patients with patients with recent exacerbations of chronic low back pain or sciatica. In our inception cohort avoiding these selection and left truncation biases, 24 90% of patients recovered within two weeks. This is a much higher rate of recovery than the 60-80% observed in otherwise similar studies.9,11 Most patients had no bed rest. Prolonged bed rest thus seems unnecessary for most patients with acute low back pain.22 We also examined the relation between pain, disability, and physical impairment and social consequences. Consistent with a previous study,8 40% of employed patients did not stop working during the pain episode. Moreover, curves of recovery and return to work, and factors associated with these outcomes, were not identical. This implies that these outcomes should be differentiated.

We identified several factors that may be of prognostic value for recovery from acute low back pain. In particular, previous chronic low back pain was a strong predictor of poor recovery, as previously suggested.10,11,25,26 This is consistent with certain people being highly prone to develop chronic pain. Initial disability rather than initial pain intensity seemed predictive of poor recovery, as reported.11 Pain worse on standing or lying was also predictive of poor recovery; these variables may identify different aetiological subgroups. There was no association between physical examination findings and recovery, in contrast with studies which included sciatica patients.11,25 The only psychosocial variable predicting recovery was compensation status, consistent with many studies.4,10,27,28 Compensation status seemed to correlate with pain and disability among patients with clear signs of organic disease who were not psychologically disturbed. This observation has major implications for public health and work legislation. We found no significant association between psychiatric diagnoses and recovery; whether psychiatric disorders are primary or secondary remains unclear.15,29

Factors previously described as influencing recovery were also associated with attendance at work among employed patients. However, two supplementary variables associated with lost work time were male sex and poor job satisfaction. This role of sex on absence from work has been described only once before.11 Manual work and job difficulty were not related to absenteeism, in contrast with studies*RF 7,27,30-32* in which analyses were not adjusted for job satisfaction or compensation status. Return to work seemed mainly dependent on sociodemographic and job related factors and only partly dependent on physical severity factors. This has implications for the interpretation of studies in occupational settings and those analysed in terms of work absenteeism alone.

This study has limitations. Firstly, the population studied cannot be considered representative of the general population of acute low back pain patients, despite being unselected primary care patients. All the subjects sought medical care, which may bias various socioeconomic factors.33 Moreover, the exclusion criteria may have led to an underrepresentation of poorly educated and foreign origin patients. Secondly, data were mainly obtained from interview and physical examination and their quality may therefore be questioned; to minimise this we used DSM-III-R criteria for psychiatric assessment and standardised techniques for pain and physical assessment. Finally, our sample size was small when the low prevalence of some exposure factors is considered.

This study suggests that recovery from acute low back pain is more rapid than previously described and identifies several prognostic factors for poor outcome of interest for medical care and clinical research.