Are degenerative spondylolisthesis and further slippage postoperatively really issues in spinal stenosis surgery?

2.1 Symptoms

The vast majority of the patients had a typical clinical history of both back and leg pain, which increased while walking and was relieved by sitting and bending forward. Leg pain was dominant in 49% of the patients, back pain was dominant in 40%, while 11% reported equal pain in the back and legs. Back pain was most often described as a 10–15 cm band of dull, aching pain in the lumbosacral area. In some patients the pain was predominant on one side. In a great majority of patients the leg pain was bilateral, diffuse, and aching, extending down the thighs and often to the ankles. Only a few patients described radicular pain.

Inclusion criteria: Patients were included in the study if they had a typical clinical history of spinal stenosis and radiological findings showing stenosis at one or more levels on an MRI.

Exclusion criteria: Patients were excluded if discectomy was part of the decompression, if they only suffered from lateral recess stenosis or if they had a history of previous spine fusion. Patients describing aching and stabbing back pain strictly in the midline, suggesting segmental pain [22], were also excluded. These patients were operated on by decompression combined with fusion (see Section 2.3 below).

2.2 Pre-operative examinations

All patients had an MRI of the lumbar spine, which revealed significant lumbar spinal stenosis at one or more levels. Some patients also had a plain X-ray. Flexion-extension radiographs were not obtained. Ordinary clinical examination did not reveal specific neurological abnormalities except in a few patients describing radicular pain radiation.

2.3 Surgery

We have previously described the microsurgical technique used which also enables full decompression of the lateral recess [21]. Our technique is the same as that later termed the interspinous microdecompression technique [23]. Decompression is performed by extirpation of the yellow ligament and chiselling a few mm of bone from the lower part of the upper lamina and upper part of the lower lamina as well as the most medial parts of the facets if bulging into the spinal canal, thus preserving almost the entire laminae, the facets and the spinal processes, being a bilateral laminotomy. The midline ligaments are, however, not preserved.

Surgery was performed from one to five levels as indicated by the MRI. Patients with and without degenerative spondylolisthesis were treated quite similarly. The number of operations performed by the eight participating surgeons was 5 (surgeon A), 11 (B), 62 (C), 2 (D), 12 (E), 79 (F), 1 (G) and 28 (H), respectively. At their first visit to the clinic the patients were consecutively assigned to one of the surgeons by the booking nurse.

During the same period (2000–2003) 10 patients with lumbar spinal stenosis were operated on by decompression combined with fusion. In these patients the decision was not based on the presence of spondylolisthesis but on symptoms indicating possible segmental pain [22]. These patients are not included in the present study.

2.4 Assessment of clinical outcome

Before treatment and on the follow-up occasion all patients answered the SF-36 questionnaire and assessed their back and leg pain on a visual analogue scale (VAS). VAS ratings are preferably given as median values as they are merely an order of assessment by the patients, being sequences on an ordinal scale. Therefore the values are not arithmetic and mean values should not be calculated, despite the fact that it is done frequently. The minimal clinical important difference (MCID) is given as a 30% decrease in the VAS scores [24] for leg pain.

At follow-up the patients were asked for their opinion about possible changes in back and leg pain (completely free, much better, somewhat better, unchanged, somewhat worse, much worse) and whether they were; satisfied with the outcome, in doubt or not satisfied and about the presence of other possible reasons for back or leg pain, e.g. hip or knee osteoarthritis etc. At the start of the study and on the follow-up occasion the patients were also asked about their walking distance as a measure of their clinical situation. Preoperative data were obtained from each patient’s records. The follow-up questionnaires were sent to the patients by post and a plain X-ray arranged at their nearest hospital.

2.5 Radiological determination of spondylolisthesis

The radiological examinations were evaluated by two specialists in neuroradiology (UM, PL) who performed the evaluations separately, not as a double check. Before treatment and at follow-up the presence or not of degenerative spondylolisthesis was determined in the lateral view on a plain X-ray or MRI central sagittal sections. The magnifying factor in plain X-rays was taken into account. Slippage was measured between the posterior lower corner of the upper vertebra and the posterior upper corner of the lower vertebra. Anterolisthesis as well as retrolisthesis was noted. The plain X-ray investigations were performed with the patients in a recumbent position. If slippage occurred at more than one level the patients were grouped according to the level showing the largest slippage. Due to possible methodological uncertainties a slippage of less than 3 mm was not noted. In each of the lumbar levels the patients were categorized as having no slippage, slippage of 3–6 mm or a slippage of more than 6 mm. At follow-up the same measurement procedure was performed with an X-ray of the lumbar spine and any change from the pre-operative situation noted, e.g. the appearance or increase of a slip in any of the operated levels and even at a non-operated level.

2.6 Statistical analysis

A brief description of the statistical methods used in this study is presented in this subsection. The covariates in the study are of three types: continuous, ordinal and nominal. In order to avoid groups containing too few observations, ordinal and nominal covariates were grouped into a smaller number of categories. For example, the variable Surgeon, which has eight levels, was regrouped into three levels, i.e. C, F, and the combined surgeon group (surgeons A, B, D, E, G, and H). See the online Appendix (https://osf.io/jnbu9) for a complete list of covariates and the regrouped variables.

An analysis of a subset of the dataset has been conducted in a master thesis project [25] using conventional methods in the spinal stenosis literature, namely F test-based stepwise regression for variable selection. No significant difference was found between patients with and without degenerative spondylolisthesis, or between patients with and without further slippage. However, the conventional approach has been considered out-dated [26].

In the present study we adopted the more sophisticated Lasso method [27], in order to obtain the best predictive power. The Lasso method is advantageous because model selection and estimation are conducted simultaneously and it tends to produce more accurate predictions than traditional methods. As the dataset contained missing values (no more than 3.5%), multiple imputation [28] was therefore used by repeatedly replacing them with plausible values, thus allow us to use more information. See the online Appendix (https://osf.io/jnbu9) for more discussions on the statistical approach adopted. The Lasso was performed in the R package glmnet [29] version 2.0-16. Multiple imputation was performed in the R package mice [30] version 3.0.0.

It is worth mentioning that the present study is not based on an experimental design. The primary goal is to possibly identify factors that can be used to predict surgical outcome, rather than estimating causal effects. For this reason, a power analysis was not conducted and the statistical analysis is not intended as the basis for affirmative causal statements.

2.7 Basic patient data

The basic patient characteristics are presented in Table 1.

2.8 Ethics

The study was approved by the Regional Ethics Committee, Stockholm, 2008/1252-31/2.

3.1 Radiological outcome

A slip arose in 36 (30%) of the 119 patients without a preoperative slip, Table 2. Among the 61 patients with an existing 3–6 mm slip preoperatively it increased in 23, while three of the 20 patients with a >6 mm slip before surgery were also found to have increased post-operative slippage. Thus out of 81 patients with an existing slip preoperatively it increased in 26 post-operatively (32%).

There was no obvious difference in the frequency of post-operative slippage between patients operated on one, two or several levels (Table 3). It should be noted that during the follow-up period slippage at non-operated levels also occurred in 9% of the patients (Table 3).

3.2 Statistical analysis of clinical outcome

We first conducted a univariate analysis of the variables, one at a time. Then, four different multivariate analyses were conducted, starting with the same set of possible explanatory factors but with different outcome measures, namely I/ change in back pain, II/ change in leg pain, III/ overall satisfaction and IV/ change in walking distance.

3.2.1 Univariate analysis

For descriptive purposes we present a so-called univariate analysis in accordance with medical statistical tradition.

3.2.1.1 Pain and function

Median VAS values concerning back and leg pain for the whole patient group, as well as for the subgroups with and without preoperative slippage are shown in Fig. 1. The statistical analysis of the VAS values for the two patient subgroups does not show any difference, Fig. 2. MCID was achieved for 70% and 68%, respectively, for the groups with and without a preoperative slip, and for 69% of all patients.

Fig. 1:

Pre- and post-operative VAS scores (median) for back and leg pain for the whole patient group and for the subgroups with and without preoperative slippage.

Fig. 2:

Box plot of pre-operative VAS scores minus post-operative VAS scores.

The patients’ assessments of change in back and leg pain are presented in Fig. 3.

Fig. 3:

Percentage of patients reporting change in back and leg pain at follow-up, patients with and without pre-operative slippage.

The SF-36 scores revealed statistically significant changes (p<0.001) post-operatively for all domains except general health (GH). The changes are illustrated in Fig. 4. No statistically significant difference was found between the two subgroups with and without slippage.

Fig. 4:

The patients’ assessments of their quality of life using the SF-36 questionnaire. The preoperative figures for patients with/without a preoperative slip, respectively, were PF 37.46/36.64, RP 11.56/11.23, BP 30.13/30.71, GH 64.09/67.83, VT 43.60/46.32, SF 62.70/61.43, RE 51.46/45.40, MH 66.51/68.61, and the post-operative figures PF 56.45/62.60, RP 48.33/53.42, BP 58.14/57.50, GH 62.57/65.64, VT 57.44/60.57, SF 78.09/78.36, RE 75.76/71.23, MH 78.88/79.18, respectively.

At follow-up 10 patients reported having had no back pain before surgery and 21 that they had no preoperative leg pain. However, at the start of the study all patients reported pain in both back and legs (Table 1). This discrepancy is probably due to recall bias as a result of the long follow-up period, being similar among patients with and without a pre-operative slip.

3.2.1.2 Walking distance

Only 28 patients were able to walk more than 1 km before surgery compared to 111 patients at follow-up. No statistically significant difference in improved walking ability was found between patients with and without a degenerative slip.

Among the 69 patients with a walking distance of less than 500 m postoperatively, 44 reported factors other than back problems that influenced their walking ability, e.g. knee and hip osteoarthritis (29 patients), foot problems (3), balance disturbance (4), and heart and lung diseases (3 patients).

3.2.1.3 Surgery parameters/complications/reoperations/additional operations

Mean bleeding during surgery was 304 mL and mean post-operative stay in hospital 5.7 days. There were no major complications affecting heart or lungs, nor new neurological impairments. During the 6.8 years (mean) of follow-up, the reoperation rate at the index level was 12% further decompressions and 2.5% fusions, thus a total of 14.5% reoperations at the index level, indicating possible “instability”. In three of these patients reoperation at the index level was combined with decompression at another level. In addition, there were 7.5% decompressions and 0.5% fusions on other levels, thus an overall reoperation rate of 22.5%. The reoperation rate was not found to differ between patients with and without a preoperative slip.

3.2.1.4 Satisfaction

At follow-up 157 of the 200 patients (78.5%) reported being completely satisfied with the clinical outcome, 35 patients were in some doubt and seven patients were not satisfied. One patient did not answer this question. In the statistical analysis the 42 patients reporting “In doubt” and “Not satisfied” were classified as Unsatisfactory, versus 157 reporting being Satisfied.

3.2.1.5 Individual surgeons

Various parameters for the different surgeon groups are presented in Table 4.

Table 4:

Differences between surgeons.

	Individual surgeons
	C 62 pat.	F 79 pat.	Combined group 59 pat.
Bleeding, op., mL. (mean)	206	341	358
Op. time, min (mean)	102	90	135
Patient satisfaction rate (%)	88.5	78.5	69.5
Decrease in VAS back pain (median)	−26	−21	−8
Decrease in VAS leg pain (median)	−48.5	−35	−32
Reoperations, index level
Decompression, no. and %	4 (6)	6 (7)	14 (24)
Fusion, no. and %	2 (3)	2 (3)	1 (2)
Total reop. rate index level (%)	10	10	25
Op. other level
Decompression, no. and %	3 (5)	5 (6)	7 (12)
Fusion, no. and %	1 (2)	0	0
Total reop. rate other level (%)	6	6	12
Total reop. rate (%)	16	16	37

The medians of the decrease in back pain on the VAS for patients operated by the respective surgeons are −26, −21, and −8. The differences in the decreases of the back pain VAS scores are weakly significant (p=0.0304). The medians of decrease in leg pain on the VAS are −48.5, −35 and −32. The difference in the VAS scores for decreased leg pain is not statistically significant. The differences between surgeons in reoperation rate at the index level are weakly statistically significant (p=0.0244). In particular, the most experienced surgeons C and F tend to have a lower reoperation rate than the combined group (p=0.0083) with a 95% confidence interval for the odds ratio (0, 0.7063).

It should be noted that the operations performed in this study were not planned in order to investigate possible differences between surgeons. The number of operations per surgeon differed and no randomization was applied. The differences found must therefore be interpreted with caution.

3.2.2 Multivariate analysis

For each post-operative outcome measure patients were grouped as having a Satisfactory or Unsatisfactory outcome, as follows:

1. Change in back pain: Satisfactory (Completely free, much better, somewhat better) and Unsatisfactory (unchanged, somewhat worse, worse), 153 and 37 patients, respectively. Ten patients reported no pre-operative back pain when asked at follow-up (not included in the analysis, see Section 3.2.1.1 above).

2. Change in leg pain: Satisfactory (Completely free, much better, somewhat better) and Unsatisfactory (unchanged, somewhat worse, worse), 139 and 39 patients, respectively. Twenty-one patients reported no pre-operative leg pain when asked at follow-up (not included in the analysis, see Section 3.2.1.1 above). One patient did not answer this question.

3. Overall Satisfaction: Grouping described above under Satisfaction.

4. Change in walking distance: Patients who reported longer post-operative walking distances than their pre-operative distances and those who reported walking more than 1 km both pre- and post-operatively were grouped as Satisfactory, the others as Unsatisfactory, 142 and 58 patients, respectively.

All outcome measures are thereby binary, which is why the logistic regression model was used. A description of the statistical method employed is presented in the online Appendix that can be retrieved from https://osf.io/jnbu9. We attempted to identify models that best predict the outcome measures. Estimates of the odds ratio and the corresponding confidence intervals are presented in Fig. 5. The confidence intervals were obtained from 20,000 bootstrap replications. For example, the odds of the re-operated patients being able to walk a longer distance are approximately 44% of those for the patients who did not undergo re-operation. However, such a difference is not statistically significant. The confidence intervals in Fig. 5 are mostly asymmetric around the estimated odds ratio. This means that the variation in the parameter estimators is large and the estimated coefficients are zero. For all analyses, most covariates show either a very small or no contribution to the prediction of post-operative outcome measures (Fig. 5), which also applies for age, gender, the various SF scores, bleeding during the operation, operation time, number of days in hospital, preoperative back and leg VAS scores and months since first surgery (not shown in the figure). Consequently, it predicts that a majority of the patients will have a satisfactory outcome, yielding around 25% incorrect prognostic decision.

Fig. 5:

Dumbbell plot of the estimates of the logarithm of the odds ratio (cross) and the corresponding 95% confidence intervals (thin line). An odds ratio c means that the odds of being satisfactory are c times as high as the reference group. A confidence interval covering 0 indicates non-significance at the 5% significance level. The reference groups in the analysis are “no reoperation (index level)”, “no slippage (another level)”, “no further slippage (index level)”, “no pre-operative slippage”, “pre-operative walking distance <500 m”, “<2 years leg pain”, “<2 years back pain”, “<2 levels operated”, “male”, and “combined surgeon group”. (A−D) Indicates Analysis I to Analysis IV, respectively.

3.2.2.1 Does slippage matter?

Figure 5 shows that neither a pre-operative slip nor further slippage at the index level is of importance for any of the outcome measures, as suggested by the 95% confidence interval of the odds ratios. This means that even though further slippage occurs on the index level, these patients are as likely to have a satisfactory outcome as patients without further slippage on the index level. Likewise, slippage on another level was not found to have a significant effect on any of the outcome measures at the 5% significance level.

Among the 81 patients with a preoperative slip 26 showed further slippage at follow-up. There was no difference between these patients and those not showing further slippage postoperatively in any of the outcome parameters, Table 5.

Table 5:

Outcome parameters for patients with preoperative slippage showing/not showing further slippage postoperatively.

	Slippage pre-op. 81 pat.
	No further slippage 55 pat.	Further slippage 26 pat.
VAS Back pain pre/post, median	55/13	66/32
VAS Leg pain pre/post, median	66/5	70/11
Percentage of patients with back pain improvement	73	81
Percentage of patients with leg pain improvement	65	73
Satisfaction rate (%)	74	73
Percentage of patients with improved walking distance	53	54
Percentage of patients achieving MICD of improvement in leg pain	69	73

3.2.2.2 Do different surgeons matter?

Despite differences between surgeons in the univariate analysis regarding the satisfaction rate of their patients, change in their patients’ back and leg pain VAS scores and also in the reoperation rate in favour of surgical experience, it can be seen in Fig. 5 that different surgeons do not have a significant effect on any of the outcome measures in the multivariate analysis at the 5% significance level.