Persistent or recurrent pain in the back or lower limbs is relatively common after an operation on the lumbar spine. The long-term rates of failure after primary operations have been reported to be as high as 30 per cent (fifty of 166 patients) after lumbar arthrodesis and 37 per cent (twenty-nine of seventy-eight patients) after lumbar discectomy 9. These patients constitute a substantial proportion of those who are referred to spine centers, and there is an enormous cost to society in terms of medical expenditures and lost productivity. Frymoyer et al., in a retrospective review of 312 patients who had had a lumbar disc operation, found that twenty-three (11 per cent) of 205 patients who had had an arthrodesis and fourteen (13 per cent) of 107 who had had a simple discectomy had needed at least one additional operation on the spine by the time of the ten-year follow-up evaluation 9,10. Most of the failures had occurred within five years postoperatively and were attributed to recurrent disc herniations or segmental instability. Other authors have reported similar results 4.
The medical literature is inconsistent with regard to the results of reoperations after failure of previous operations on the spine (Table I). Finnegan et al. reported a good result in eight (12 per cent) of sixty-seven patients who had had a repeat spinal operation, and Bernard reported a successful result in thirty-seven (82 per cent) of forty-five such patients. Other authors have reported a successful reoperation in 25 to 80 per cent of patients 2,12,13, but there has been little agreement as to which factors predict such an outcome. The variation among the findings is due in part to differing patient populations, operative indications 17, follow-up procedures, and criteria for success.
The purpose of the current study was to review a consecutive series of patients who had had a revision operation on the lumbar spine and to determine which factors contributed to a successful outcome, as perceived by both the patient and the surgeon.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. No funds were received in support of this study.
†Flynn Spine and Scoliosis Center, 100 West Gore Street, Suite 403, Orlando, Florida 32806. Please address requests for reprints to Dr. Stewart.
‡Department of Surgery (Orthopaedics), Albany Medical College, 47 New Scotland Avenue A-61-OR, Albany, New York 12208.
A consecutive series of forty-two patients had a repeat open operative procedure on the lumbar spine because of pain in the back or lower limbs that had been unresponsive to non-operative treatment. Of these patients, thirty-nine (93 per cent) had been followed for a minimum of two years and were available to complete a questionnaire and three had been lost to follow-up and were therefore excluded from the study. All of the revision procedures were performed by the senior one of us (B. L. S.) between September 1987 and February 1991.
The thirty-nine patients were followed for an average of forty-eight months (range, twenty-four to eighty-six months). There were twenty-seven men and twelve women; the average age at the time of the revision was forty-three and one-half years (range, twenty-nine to sixty-nine years). Twenty-four patients were receiving Workers' Compensation or had pending litigation. The patients had had one or two (average, 1.3) previous operative procedures on the lumbar spine, and an average of fifty-one months (range, six to 336 months) had elapsed since the most recent (index) operation. The average pain-free interval after the index procedure was twenty months (range, zero to 240 months). Twelve patients reported that they had had no pain-free period postoperatively, and five had had only a transient decrease in the symptoms. Twelve patients had had a previous attempt at arthrodesis, and the remainder had had a simple decompression. The average number of intervertebral disc levels involved in the index procedure was two (range, one to six). Thirty-seven patients had had the previous operation at the same vertebral level as the revision procedure. The index operation had been done by the senior one of us in six patients and by a different surgeon in thirty-three patients.
Four of the thirty-nine patients were employed and working at the time of the revision operation, and four were retired or were homemakers. The remaining thirty-one patients were employed but stated that they were unable to work because of the pain in the back or lower limbs; the average amount of time lost from work was 18.8 months (range, one to seventy-two months). Eight of the thirty-one disabled patients had a job that involved strenuous labor; nineteen, a job that required moderate activity; and four, a sedentary job.
On the analog pain-scale completed by the patient, the average rating for back pain before the revision was 7.2 points (range, 3 to 10 points) and the average rating for pain in the lower limbs was 6.7 points (range, 1 to 10 points). Back pain was the predominant symptom in ten patients, pain in the lower limbs was the predominant symptom in eight, and twenty-one rated the pain in the back and lower limbs as equal. Fifteen patients had an objective neurological deficit.
All thirty-nine patients used non-steroidal anti-inflammatory medications to help to control the pain in the back and lower limbs before the revision. Thirty-five patients also used narcotic analgesics. Thirty-four patients used a lumbar support: a soft corset was used exclusively by thirteen, and a rigid brace was used intermittently by twenty-one.
Before the revision, all patients had thorough clinical and diagnostic testing, and all completed analog pain-scale and pain-distribution diagrams. Radiographs were made with the spine in flexion and extension, and computed tomography, magnetic resonance imaging, or myelography was performed for all patients. Fifteen patients had discography. The preoperative diagnosis was residual or recurrent herniated nucleus pulposus in eight patients, spinal stenosis in eleven, spinal instability in twelve, pseudarthrosis in eight, internal disc derangement in eighteen, and painful internal fixation in one. Nineteen patients had more than one preoperative diagnosis.
The revision procedure consisted of posterior decompression in three patients; posterior decompression and arthrodesis in three; posterior decompression and arthrodesis with internal fixation in fifteen; anterior lumbar interbody arthrodesis in nine; anterior arthrodesis combined with either posterior decompression or arthrodesis, or both, in eight; and exploration and removal of the metal implant in one.
The postoperative assessment was conducted by the same one of us (G. S.), who was not the surgeon. The assessment consisted of a comprehensive written questionnaire, completed privately by the patient and returned to the reviewer by mail, and a review of office records, hospital records, and diagnostic studies. The questionnaire asked the patient to use a 10-point analog scale to rate the pain in the back and lower limbs (0 = none and 10 = extreme), work and functional limitation (0 = none and 10 = total), the over-all operative result (0 = much worse and 10 = perfect), and satisfaction (0 = very dissatisfied, 5 = neutral, and 10 = very satisfied). Use of any narcotic medication, including codeine, was classified according to frequency (never, occasionally, weekly, or daily); use of a brace and employment status were also recorded.
The operative results were classified according to three primary criteria: (1) return to work, (2) need for narcotic medication, and (3) satisfaction of the patient with the outcome of the operative procedure. The outcome was considered to be a success if the patient had returned to his or her pre-injury work capacity, did not need any narcotic medication, and had rated the outcome of the operative procedure and the over-all result highly (satisfied or perfect). If any of these three criteria was not met, the outcome was rated as a failure. With use of a Student t test or a chi-square analysis (level of significance, p < 0.05), several factors were evaluated to determine which were related to a successful outcome: (1) age of the patient, (2) gender of the patient, (3) a compensable injury, (4) number of previous operations on the spine, (5) type of index procedure, (6) pain-free interval after the index procedure, (7) employment status, (8) amount of time lost from work, (9) neurological deficit, (10) dermatomal pattern of pain on the pain-distribution diagram, (11) predominance of pain in the back or in the lower extremities, (12) preoperative diagnosis, (13) type of operative procedure performed at the revision, and (14) status of the fusion.
The success or failure of the revision procedure was assessed according to the patient's report about work status, use of narcotic medication, and satisfaction with the outcome. At the time of the latest follow-up evaluation, twenty-six patients had returned to full-time employment and two, to part-time employment (Fig. 1). Five of these patients had left a job involving strenuous labor for one that was less physically demanding. One patient was unable to work because of cardiac disease, and two patients had retired although they remained capable of physical labor. Four patients were unable to work because of residual pain in the back or lower limbs, and the outcome was rated a failure. Eleven patients reported occasional use of anti-inflammatory medication, and four used such medication on a daily basis. Thirty-three patients reported no need for narcotic medication to control the pain; two patients, occasional use; one patient, weekly use; and three patients, daily use. The outcome for the latter six patients was rated a failure. Thirty patients rated their satisfaction with the outcome as 7 to 10 points (satisfied to perfect) on the analog scale. Fourteen of these patients rated their level of satisfaction as 7 or 8 points and sixteen, as 9 or 10 points. All thirty also rated the operative result as 7 to 10 points (improved to perfect). Of the nine remaining patients, four were dissatisfied and five were neutral. The outcome for these nine patients was rated a failure.
With failure of treatment defined as failure of at least one of the three criteria, the outcome for eleven patients (28 per cent) was rated a failure. Five of these patients had failure according to more than one criterion. The remaining twenty-eight patients (72 per cent) returned to their pre-injury work status, did not need to use narcotic medications, and were satisfied with the outcome of the operation. To our knowledge, no patient in this study has had an additional operation on the spine.
Pain Ratings
The average postoperative rating of back pain was 3.2 points (range, 0 to 10 points) on the analog scale, compared with an average of 7.2 points (range, 3 to 10 points) preoperatively; this improvement was significant (p < 0.001; Fig. 2-A). Postoperatively, back pain decreased in thirty-five patients, increased in three, and was unchanged in one. The average postoperative rating of pain in the lower extremities was 2.4 points (range, 0 to 8 points), compared with an average of 6.7 points (range, 1 to 10 points) preoperatively (p < 0.001; Fig. 2-B). Postoperatively, pain in the lower extremities decreased in thirty-six patients, increased in one, and was unchanged in two. Patients categorized as having a successful outcome had significantly greater relief of pain in the back and lower extremities than those classified as having a failure (p < 0.001). The patients who had a failure had no significant decrease in back pain and only marginally significant relief of pain in the lower extremities (p < 0.10). The pain ratings were significantly related (p < 0.05) to use of narcotic medication, over-all satisfaction, and work status (Figs. 3-A, 3-B, and 3-C).
Status of the Fusion
A solid fusion was achieved in twenty-nine (83 per cent) of the thirty-five patients in whom an arthrodesis had been performed, including seven of the eight who had had a pseudarthrosis preoperatively and twenty-two of the twenty-seven who had had a primary arthrodesis. Six patients had a pseudarthrosis. With the numbers available, the association between pseudarthrosis and failure was not significant.
Use of a Brace
At the latest follow-up evaluation, fourteen of the thirty-nine patients had continued to use a back support intermittently. Of these fourteen patients, eight used a soft corset; five, a rigid brace; and one, both types of support. As might be anticipated, the use of a lumbar support was more common in the group for which the outcome was a failure (p < 0.05), as was the use of a rigid as opposed to a soft brace (p < 0.05).
Complications
Five patients had postoperative complications. One had a dural tear and a headache; one, reflex sympathetic dystrophy of the foot; one, a compartment syndrome of the calf; one, diverticulitis; and one, a myocardial infarction and a superficial wound infection. With the numbers available, the occurrence of a complication was not significantly associated with the subjective rating of satisfaction or the over-all outcome.
Statistical Analysis (Table II)
Statistical analysis revealed that the factors associated with a successful outcome were a younger age (p < 0.02), fewer spinal levels operated on previously (p < 0.05), a pain-free interval after the index operation (p < 0.01), working outside of the home (p < 0.05), and the type of procedure performed at the revision operation (anterior lumbar interbody arthrodesis alone or in combination with a posterior procedure) (p < 0.02).
With the numbers available, we could detect no significant relationship between the outcome and gender, a compensable injury, the number of previous operative procedures, the disability status, the amount of time lost from work, the type of index procedure, the interval between the index operation and the revision, the duration of the pain-free interval after the index operation, the presence of a preoperative neurological deficit, a dermatomal distribution of pain on the pain-drawing, or the predominance of pain in the back or pain in the lower limbs.
In the current study, the outcomes of revision procedures on the lumbar spine in a consecutive series of patients, managed by the same primary surgeon, were analyzed. The preoperative data were collected in a prospective manner, and all retrospective evaluation was done by the same independent reviewer. Our criteria for a successful outcome were stringent compared with those reported in the literature 1,7,12,13; they included measurements of function (as evidenced by return to work), comfort (as reflected by the need for narcotic analgesics), and over-all subjective satisfaction. These measurements seem to reflect adequately the goals of both the patient and the practitioner, and they appear to have internal consistency. Patients classified as having a successful outcome had significantly greater relief of pain than those classified as having a failure (p < 0.001), although the pain score was not used as a differentiating factor. The analog pain-ratings showed a significant association with each of our criteria for a successful outcome. The over-all outcome (a successful result in twenty-eight [72 per cent] of the thirty-nine patients) is comparable with rates that have been reported elsewhere 1,7,12,13.
Of the thirty-five patients who had been employed before the revision, thirty-one (89 per cent) were unable to work because of disability. After the revision, only four (10 per cent) of the thirty-nine patients remained disabled by pain. These four were among the twenty-four who were receiving Workers' Compensation or had pending litigation but, with the numbers available, no significant association was found between the over-all outcome and the status with regard to Workers' Compensation. Several studies have demonstrated a relationship between a compensable injury and a poor outcome 1,2,7,15; however, other investigators have not found a significant relationship 12. The lack of association in the current study may reflect our unique population of patients or our more rigorous evaluation and screening of operative candidates.
As have other investigators 2,7,15, we found that a pain-free interval after the previous operation was associated with a successful outcome of the revision (p < 0.01). The average pain-free interval was greater in the patients who had a successful outcome after the revision than in those who had a failure (22.5 compared with 13.8 months), but this difference was not significant. Younger patients were more likely to have a successful outcome (p < 0.02), as were patients who were in the workforce (p < 0.05). This may be due to more extensive degenerative changes in the spine with age or it may reflect the motivation of these patients and their ability to comply with postoperative rehabilitation. These findings are consistent with those of Hurme and Alaranta in patients who had primary procedures on the lumbar spine.
Waddell et al. found that the probability of a successful outcome of a revision operation on the lumbar spine decreases with each subsequent procedure. Although some authors have reported similar findings 7, most have found no significant relationship between the number of previous spinal procedures and the outcome of revision operations 1,2,12,15. In contrast to Waddell et al., we found no relationship between the outcome of revision procedures and the number of previous operative procedures, the nature of the previous procedures, or the time-interval since those procedures. However, a more extensive previous procedure, as evidenced by the number of operative levels, was associated with a poor outcome (p < 0.05). This may indicate more extensive disease in the spine or it may reflect difficulty in localizing the anatomical source of pain. To our knowledge, this factor has not been previously addressed in the literature.
Although neurological compromise is a common indication for operative intervention, we found no significant relationship between the preoperative presence of a neurological deficit or the radicular distribution of pain and the outcome. This is consistent with the findings of other authors 7,12,15. As has been noted, neurological signs may be absent in patients who have a decompressible lesion, such as spinal stenosis, and present in those who have a lesion that is more difficult to address operatively, such as epidural fibrosis. The predominance of pain in the lower limbs as compared with back pain was also not significant. These groups had similar over-all results and a similar degree of diminution of the pain, as measured on an analog scale. Other studies have yielded similar results 12,16.
Seventeen patients were managed with an anterior lumbar interbody arthrodesis, with or without a concomitant posterior procedure. These patients were more likely to have a successful outcome than those who were managed with a posterior procedure alone (p < 0.02). Fifteen of these patients had had positive findings on preoperative discography, with concordant reproduction of pain and a radiographically abnormal disc. The success of anterior lumbar interbody arthrodesis in the treatment of primary internal disc derangement has been demonstrated previously3,5,14 and may be due to decreased micromotion 8 at the interspace and removal of mediators of inflammation by removal of the damaged disc. In this series, all of the previous procedures had been performed with use of a posterior approach, and anterior revision through undisturbed tissue planes may have aided visualization, decompression, and stabilization. To our knowledge, the superiority of anterior lumbar interbody arthrodesis in revision procedures has not been demonstrated previously.
In conclusion, patients in whom treatment for back pain has failed present a diagnostic and therapeutic challenge. These patients are frequently severely disabled and represent an enormous cost in terms of lost productivity and medical expenditures. We have demonstrated that the operative correction of pathoanatomy can result in a dramatic decrease in symptoms and can restore patients to a functional lifestyle. The rate of positive outcomes in the current study has an even greater potential importance, as measurements of outcome are increasingly being used to determine not only the therapeutic but also the financial value of medical interventions.