Epidural anaesthesia and low back pain after delivery: a prospective cohort study

Introduction

Over the past 35 years epidural anaesthesia has been used to provide effective pain relief for women during labour and delivery. Two retrospective surveys in the United Kingdom have suggested an association between epidural anaesthesia and long term low back pain of new onset after delivery.1 2 In both studies women receiving epidural anaesthesia during labour and delivery were almost twice as likely to develop back pain after delivery as women not given epidural anaesthesia. Both surveys, however, were compromised by their retrospective design and low response rate. The objective of our prospective cohort study was to determine whether epidural anaesthesia during labour and delivery was a determinant of postpartum low back pain.

Methods

All women attending the (tertiary level) labour and delivery unit of the Royal Victoria Hospital in Montreal for delivery of a live infant(s) were eligible for study. Women who before pregnancy had a history of back pain requiring medical attention were excluded as there is evidence that such pain may be an independent risk factor for postpartum back pain.3 Also excluded were women undergoing elective caesarean section as spinal anaesthesia rather than epidural anaesthesia was the standard for this procedure.

A prospective cohort study design was used, with the exposure of interest being epidural anaesthesia during labour and delivery (yes/no). Epidural anaesthesia was performed on request, with six obstetric anaesthetists participating in the study. The anaesthetic regimen was not standardised, but similar protocols were used--that is, epidural analgesia was established with 10-12 ml bupivacaine 0.25% followed by boluses of 8 ml bupivacaine 0.25% as required and the use of lignocaine or chlorprocaine for second stage analgesia. Difficulty in establishing epidural anaesthesia was defined a priori as more than three needle insertions or two or more intervertebral spaces tried before success.

The primary outcome variable was development of postpartum low back pain (yes/no). Back pain was also quantified by using a self report numeric rating scale ranging from 0-10, with 0 representing no pain and 10 representing the worst pain imaginable. This pain scale has been shown to be both reliable and valid in adults with the chronic pain syndrome.4 Lastly, the degree of functional impairment associated with low back pain was also assessed. Functional impairment was categorised as interference with daily activities sufficient to require analgesia or physician consultation (yes/no), or both. All subjects were interviewed by a trained research nurse at one day, seven days, and six weeks after delivery. The research nurse was blind to the study hypothesis, with blinding enhanced by the inclusion of questions on postpartum gastrointestinal and genitourinary symptoms.

Information on potential confounding variables--that is, age, parity, method of delivery (vaginal versus emergency caesarean section), ethnicity, socioeconomic status, weight gain during pregnancy, weight at delivery, time in stirrups, and history of low back pain during pregnancy--was also collected on both groups. Socioeconomic status was quantified with the two factor score developed by Green.5

Two retrospective surveys have shown that the incidence of postpartum low back pain in women not receiving epidural anaesthesia during labour and delivery is around 12%.1 2 Therefore, on the basis of an (alpha) error of 5% and an estimated prevalence of postpartum back pain in the non-epidural group of 12%, to provide 80% power to detect a “clinically important” relative risk of 2.0 (that is, a twofold increased risk of low back pain among women receiving epidural anaesthesia), about 160 women in each group were required.6 Subjects were recruited over 12 months (May 1992 to April 1993).

The frequency of low back pain in each group (epidural and non-epidural) at one day, seven days, and six weeks postpartum was estimated. These data were then used to calculate a crude or unadjusted relative risk--that is, the risk of postpartum low back pain among women receiving epidural anaesthesia compared with women not given epidural anaesthesia for each of the three time points. Confidence intervals around the point estimates were derived with Miettinen's test based method.7

In addition, to take into account differences in the distribution of potential confounding variables across the two groups, multivariate logistic regression analysis was used to provide adjusted relative risk estimates for the association between epidural anaesthesia and low back pain. Stepwise logistic regression analysis was used to build the multivariate model, with (alpha) to enter and (alpha) to exit levels at 0.15 and 0.20, respectively.8 (Logistic regression provides odds ratio estimates; given the infrequency of low back pain, however, odds ratios were taken to approximate relative risks.) Differences in the pain scores on the numeric rating scale between the two groups were tested with the non-parametric Mann-Whitney U test whereas differences in the proportion of women in each group with functional impairment because of low back pain were tested with the χ² test (P values <0.05 were considered significant).

Data analyses were performed both on the entire sample and on the subgroup of women with no history of low back pain during their pregnancy. For the latter group, the development of postpartum low back pain was defined as “new onset” back pain.

Results

In total, 329 women enrolled in the study (only one eligible subject declined to participate). Of these, 164 women received epidural anaesthesia during labour and delivery, and 165 did not. Table I shows the baseline characteristics of the two groups. There were no differences between the two groups on age, socioeconomic status, weight at delivery, weight gain during pregnancy, or history of low back pain during pregnancy. Women in the epidural group, however, were significantly more likely to be primiparous, to be white, to have delivered by emergency caesarean section, and to have spent longer in stirrups during the second stage of labour.

Based on the a priori entry and exit criteria stepwise model building resulted in a final logistic regression model that included five variables--that is, the exposure of interest (epidural anaesthesia, yes/no) as well as parity, method of delivery, ethnicity, and weight at delivery. Therefore, logistic regression analysis quantified the relation between epidural anaesthesia and the risk of postpartum low back pain while taking into account differences between the two groups (epidural and non-epidural) on important confounders. The final logistic regression model was considered biologically plausible for two reasons. Firstly, previously published studies had shown that the variables included in the model were potential confounders of the association between epidural anaesthesia and postpartum low back pain,1 2 3 and, secondly, there were clear differences in the distribution of these variables between the two groups.

The frequency of low back pain was highest in both groups on day one after delivery with around half the women in both groups complaining of low back pain. Table II gives further details and the crude and adjusted relative risks over time. Follow up at one day and seven days was complete; three patients (two from the epidural group and one from the non-epidural group), however, were lost to follow up at six weeks.

The pain scores on the numeric rating scale were clearly skewed to the right given that most women in both groups reported no back pain (that is, a score of 0) at one day, seven days, and six weeks after delivery. Non-parametric testing with the Mann-Whitney U test showed no significant differences in pain scores between the two groups at one day, seven days, or six weeks. Table III provides the median pain scores for the two groups. Likewise, there were no differences between the two groups in the proportion of women reporting interference with daily activities because of low back pain. On the first day after delivery 30 (18%) women in the epidural group compared with 35 (21%) women in the non-epidural group complained of interference with activities (P=0.490). At seven days these figures were 11 (7%) and 10 (6%). At six weeks only three (2%) women in each group complained of interference with daily activities because of low back pain (P=0.704).

Establishment of epidural anaesthesia was considered difficult in 11 women (7%). The risk of low back pain in this subgroup, however, was not significantly different from the risk in those women in whom establishment of epidural anaesthesia was not considered difficult.

In total, 108 women in each group had no history of low back pain during their pregnancy. Table IV gives details of the analyses on this subgroup of women and shows the results to be similar to those for the entire sample. There were also no differences between the two groups on pain scores at one day, seven days, or six weeks after delivery (median pain scores are shown in table III). Table IV also shows that there were no differences between the two groups in the proportion of women with functional impairment because of low back pain of new onset.

Discussion

This prospective cohort study found that the incidence of low back pain on the first day after delivery was significantly higher in women who had received epidural anaesthesia during labour and delivery than in women not given epidural anaesthesia. Over time, the incidence of low back pain declined in both groups (from around 50% on day one to around 10% at six weeks) with no differences in incidence of back pain noted between the two groups at seven days or six weeks postpartum. Although not significant, however, the risk of low back pain at six weeks was about twice as high among women in the epidural group as in the non-epidural group. This trend in relative risk of low back pain (epidural v non-epidural)--that is, significant on day one, diminution by day seven, yet increasing again by week six postpartum--was also noted in the subgroup of women with back pain of new onset--that is, women with no history of back pain during their pregnancy.

The research nurse recruited patients the morning after delivery during the regular working week only (Monday to Friday). Therefore, women delivering on Friday or Saturday and discharged home over the weekend were missed. Sample distortion bias was considered unlikely, however, for several reasons. Firstly, women were missed whether or not they had received epidural anaesthesia; secondly, patient demographics for the labour unit were similar by day; and, lastly, both the obstetricians and anaesthetists were on a rotating work schedule. Information bias was minimised by using a trained research nurse who was blind to the study hypothesis. Confounding bias was dealt with by using multivariate logistic regression analysis to provide adjusted relative risks--that is, risk estimates that took into account the effects of differences in the distribution of potential confounding variables, such as parity and method of delivery, across the two groups.