Prognostic factors for non-success in patients with sciatica and disc herniation

A prospective observational multicenter cohort study was conducted. The Regional Committee for Medical Research Ethics and The Ombudsmann for Privacy in Research at the Norwegian Social Science Data Services approved the study protocol.

Patients were recruited from specialty back clinics at four public hospitals in southeast Norway (Sykehuset Østfold, Sørlandet Sykehus, Oslo Universitetssykehus Ullevål and Sykehuset Innlandet). The inclusion period was 2 years, throughout 2005 and 2006. The patients were referred to the specialty back clinics by primary health care services. The participating specialty back clinics were all in public hospitals, and the Norwegian Social Insurance pays for patients referred to public hospitals. The majority (60%) of the patients (n = 280) were recruited from Sykehuset Østfold, which serves a district with 260,000 inhabitants. About 200 patients with sciatica were referred yearly to this specialty back clinic at the start of the study. This suggests that about 3/4 of the patients referred to this hospital for sciatica were included in the study. The next largest population was from Sørlandet Sykehus (n = 89), located in a district with 162,000 inhabitants. This back clinic treated about 120 patients with sciatica yearly, so some more than 1/3 of eligible patients were included in the study. We do not have data on the proportion of eligible patients recruited to the study from the other 2 hospitals, or the number of eligible patients who either were not invited or declined to participate in the study. In the centre that enrolled the majority of the patients, less than 5% of eligible patients declined to participate during the first year of the study.

In each back clinic, consecutive eligible patients were invited to participate in the study by a physician or a physiotherapist.

Inclusion criteria were age ≥18 years, radiating pain and/or paresis below knee level and a disc herniation at the corresponding level and side that had been verified by magnetic resonance imaging (MRI) or computed tomography (CT). Exclusion criteria were prior surgery at the same disc level, fracture, infection, malignancy, pregnancy and lack of fluency in Norwegian. All participants received oral and written information about the study and gave their informed consent to their participation.

Study participation did not involve any specific type of intervention or alter treatment considerations for the patients, nor did it involve any specific advantages or benefits for the patients in the clinics. Hence, patients received the usual consultations including a clinical examination, information about sciatica and disc herniation, back exercises, physical therapy, and pain medication. Information and treatment were given on an individual basis to the patients at each center.

Surgery was performed for patients with severe symptoms. The patient was referred from the back clinic to an orthopedic surgeon, who made the final decision about surgical treatment. The decision about surgery was made for individual patients at each center and no standardized criteria were established for surgical treatment.

At the day of inclusion patients completed a comprehensive questionnaire. Baseline data were collected at the first visit to the department. Clinical examination was conducted by a physician or physiotherapist. A follow-up questionnaire and a prepaid envelope were sent to the patients after 3, 6, 12 and 24 months. A reminder was sent after 2 weeks if no reply was obtained.

At inclusion, the following sociodemographic factors were recorded: age, sex, education (years of schooling), smoking status and work status. Patients also reported their history of back pain (dichotomised to <; 1 year or ≥ 1 year), previous episodes of sciatica (0 or ≥ 1), and the duration of the current episode (dichotomised to <; 3 months or ≥ 3 months).

The following patient-reported variables were also recorded.

Pain intensity in the back and in the leg (sciatica) during the previous week was recorded on a horizontal visual analogue scale (VAS) ranging from 0 (no pain) to 100 (worst pain ever).

The comorbid subjective health complaints inventory included 29 common somatic and psychological complaints, such as muscular pain, headaches, stomach pain and discomfort, hot flushes, extra heart beats, sleep problems, tiredness, dizziness, anxiety and depressed thoughts during the previous month [21, 22]. Patients graded the intensity of each complaint during the previous month as not at all (0), a little (1), some (2) or severe (3). Each item was dichotomised to absent (0) or present (1, 2, 3). Because most patients with sciatica have low back pain and leg pain during exercise, the 2 items referring to these symptoms were excluded, resulting in a score in the range of 0–27.

The patients also completed the Tampa Scale for Kinesiophobia (TSK) [23, 24], a 13-item, four-point questionnaire. Scores range from 13 to 52, where higher scores indicate increased kinesiophobia.

Emotional distress was assessed using the Hopkins Symptom Check List-25 (HSCL-25) [25]. The questionnaire includes 25 items on depression, anxiety and somatisation during the previous week and ranges from 1 (not at all) to 4 (extremely). The score is calculated as the mean of the completed items. An average item score of ≥ 1.75 was found to be a good predictor of current help-seeking behaviour in a Norwegian epidemiological study and is commonly used to define cases with emotional distress [26].

The sciatica-specific clinical assessment included motor function (deemed abnormal if the extension or flexion of the knee or ankle or the extension of the big toe was reduced; if, when standing on one leg, the pelvis tilted to the side opposite the stance leg [positive Trendelenburg test]; or if abnormal tiptoe or heel walking was present), sensibility (deemed abnormal if tactile sensibility was reduced), reflexes of the Achilles tendon or patella (deemed abnormal if reduced or not elicited) and the straight-leg-raising test (deemed abnormal if pain provocation in the leg emerged at <; 60°).

The results of the clinical examination were dichotomised as normal or abnormal. The testing procedures and a description of each clinical test were discussed and standardized in meetings with the participating centres before the patients were enrolled.

In order to assess the association between the surgery and the outcome of surgery for the herniated disc, patients who underwent surgery during the observational period reported the date of surgery at follow-up questionnaires. In each questionnaire, the participants were asked whether they had undergone surgery for disc herniation in the period since the last follow-up period, and if so, the patient reported the date of surgery.

The Maine–Seattle Back Questionnaire (MSBQ) was the main outcome measure [27]. The MSBQ is a shortened version of the Roland–Morris Disability Questionnaire that was modified for patients with sciatica and spinal stenosis [28]. The scale is composed of 12 items, each with the answer yes (1) or no (0), achieving a score range of 0–12. The MSBQ assesses disability and functional limits due to sciatic and back pain, and higher scores indicate worse limitations on activity. We have previously reported that the MSBQ is the best measure for distinguishing between success and non-success in sciatica at 1 year of follow-up [29]. Non-success was defined as a MSBQ score ≥ 5.

The secondary outcome measure was the Sciatica Bothersomeness Index (SBI), which is a patient-reported questionnaire that assesses sciatic symptoms [27]. The SBI is a composite of the scores for four symptoms: leg pain (sciatica); numbness or tingling in the leg, foot or groin; weakness in the leg or foot; and back or leg pain while sitting. The scores are in the range of 0–6 for each item, summing to a total score in the range of 0–24, where higher scores indicate worse symptoms. Non-success was defined as a SBI score of ≥ 7 [29].

It has been suggested that for prognostic studies, at least 10 outcome events are required for each factor studied [30]. In the present study, a total of 20 prognostic factors were included. With a sample of 400 patients and poor outcomes for about 50% of the sample, we should have enough power to assess about 20 prognostic factors.

Baseline prognostic variables were analysed using logistic regression with MSBQ ≥ 5 as the primary dependent variable and SBI ≥ 7 as the secondary dependent variable. To compensate for missing items in a questionnaire, the missing item was substituted with the arithmetic mean of the actual respondent’s available item values.

A chi-square test and a Studen’s t-test were used to compare responders and non-responders. The significance level was set at 0.05. The variance inflation factor (VIF) was used to check for multicollinearity and variables with a VIF <; 5 were accepted for inclusion in the model [31].

The results are presented as odds ratios (ORs) with 95% confidence intervals (CIs). Baseline variables from the univariate analysis with a p-value <; 0.10 were included in a multivariate model. The final model controlled for age, sex and baseline variable for the dependent outcome, regardless of the significance in the univariate analysis. The same statistical models were used for 1 and 2 years of follow-ups.

In a backward approach, the non-significant variable with the highest p-value was removed in a stepwise manner until all variables had p <; 0.05. The variable surgery (yes/no) was added to the model in the last step, to assess the association between surgery for the herniated disc and the outcome. Variables remaining in the final models were tested for possible interaction effects where the level of interaction was set to p <; 0.01. Nagelkerke R² was used to describe the proportion of variance in the outcome. Data were analysed using the SPSS package (version 18.0) for Windows (SPSS Inc., Chicago, IL).

In total, 466 patients were included. Four hundred nine patients (88%) responded to the 1-year follow-up questionnaire and 380 (82%) responded to the 2-year follow-up questionnaire. Among the responders at 1 year, 120 (29%) had received surgical treatment. At 2 years, 120 (32%) of the responders were recorded as surgically treated. For patients who were operated, surgery was performed within 3 months of follow-up for 81% of the patients.

The participants who did not respond to the questionnaire at 1 year were significantly younger, had higher back pain scores and were more frequently positive for the straight-leg-raising test at baseline. The non-responders at 2 years were younger, more often current smokers, had higher back pain scores, more emotional distress and were more frequently positive for the straight-leg-raising test at baseline.

The numbers of missing values replaced at baseline were: SBI, 3 (0.6%); TSK, 26 (5.6%); and MSBQ, 25 (5.4%). The numbers of missing values replaced for the MSBQ were 28 (6.9%) at 1 year and 29 (7.6%) at 2 years. The item in the MSBQ regarding sexual activity had the most missing data: 15 (3.2%) at baseline, 19 (4.6%) at 1 year, and 13 (3.4%) at 2 years.

The baseline characteristics are presented in Table1. The mean age at baseline was 43.6 years and 58% were males. Most of the patients were employed although only approximately 20% were in work at inclusion. Furthermore, 45% of the sample reported a sciatica episode for the first time. In general, the patients reported high scores on the self-reported questionnaires, in particular for leg pain, disability and emotional distress.

Patients with non-success (MSBQ ≥ 5) numbered 178 patients (44%) at 1 year and 145 (39%) at 2 years. Among the surgically treated patients, 42 (35%) had non-success at the 1-year follow-up, and 47 (39%) had non-success at the 2-year follow-up. In the non-surgical group, 136 (47%) and 98 (39%) patients had non-success at 1 and 2 years, respectively.

One hundred ninety-four patients (47%) reported non-success, defined as SBI ≥ 7, after 1 year and 159 (42%) after 2 years. Among the surgically treated patients, 36 (30%) reported non-success at the 1-year follow-up and 40 (33%) at the 2-year follow-up. In the non-surgical group, 157 (54%) and 118 (47%) patients reported non-success at 1 and 2 years, respectively.

Table2 presents the associations between non-success and all baseline factors both in the univariate analysis and in the final multivariate analyses for the main outcome variable. Higher scores for all self-reported health status variables, shorter education length and longer duration of back problems were significantly associated with non-success at both 1 and 2 years according to the univariate analyses. To assess whether the associations with non-success in the univariate analysis differed when adjustments were made for surgical treatment during follow-up, each baseline variable was adjusted for surgical treatment. This adjustment did not change results. Data are not shown.

According to the final multivariable regression models, comorbid subjective health complaints was the only prognostic factor significantly associated with non-success at both the 1- and 2-year follow-ups. Other variables independently and significantly associated with non-success at 1 year were male sex, higher intensity of back pain and abnormal reflexes on the clinical examination. Prognostic factors for non-success at 2 years included higher kinesiophobia and longer duration of back pain and sciatica. Adding surgical status to the final models showed non-surgical treatment to be significantly associated with non-success at 1 year (OR = 2.97 [1.75 − 5.04]), but not at 2 years (OR = 1.32 [0.78 − 2.23]). The adjustment variable, the baseline scores of the MSBQ, remained significant in the multivariate models both at 1 and 2 year follow-ups. For the final models, the explained variance assessed by Nagelkerke R² was 26% at both 1 and 2 years. No multicollinearity or interactions were detected for the selected baseline variables and testing for collinearity after the surgery variable was introduced did not change the results. At baseline the correlations assessed by Spearman’s Rho were: Subjective health complaints and kinesiophobia 0.19; subjective health complains and emotional distress 0.59; and kinesiophobia and emotional distress 0.41.

Table3 presents the associations between non-success and all baseline factors both in univariate analysis and in the final multivariate analyses for the secondary outcome variable. When adjusting all univariate analyses for surgical treatment during follow-up, only leg pain was significantly associated with non-success at 1 year (p = 0.003). Other data are not shown. In multivariate analyses back pain, comorbid subjective health complaints, muscular weakness at clinical examination, and non-surgical treatment were independently prognostic factors both for the 1 and 2 year results. For the final models, the explained variance assessed by Nagelkerke R² was 26% at 1 year and 20% at 2 years. An interaction effect was detected between surgery and smoking in the 2 year results, p = 0.004.