SPINE Volume 30, Number 8, pp 927–935©2005, Lippincott Williams & Wilkins, Inc.

Long-Term Outcomes of Surgical and NonsurgicalManagement of Sciatica Secondary to a Lumbar DiscHerniation: 10 Year Results from the Maine LumbarSpine Study

Steven J. Atlas, MD, MPH,* Robert B. Keller, MD,† Yen A. Wu, MPH,*Richard A. Deyo, MD, MPH,‡ and Daniel E. Singer, MD*

Study Design. A prospective cohort study.Objective. To assess 10-year outcomes of patients

with sciatica resulting from a lumbar disc herniationtreated surgically or nonsurgically.

Summary of Background Data. There is little informa-tion comparing long-term outcomes of surgical and con-servative therapy of lumbar disc herniation in contempo-rary clinical practice. Prior studies suggest that theseoutcomes are similar.

Methods. Patients recruited from the practices of or-thopedic surgeons, neurosurgeons, and occupationalmedicine physicians throughout Maine had baseline in-terviews with follow-up questionnaires mailed at regularintervals over 10 years. Clinical data were obtained atbaseline from a physician questionnaire. Primary analy-ses were based on initial treatment received, either sur-gical or nonsurgical. Secondary analyses examined ac-tual treatments received by 10 years. Outcomes includedpatient-reported symptoms of leg and back pain, func-tional status, satisfaction, and work and disability com-pensation status.

Results. Of 507 eligible consenting patients initiallyenrolled, 10-year outcomes were available for 400 of 477(84%) surviving patients; 217 of 255 (85%) treated surgi-cally, and 183 of 222 (82%) treated nonsurgically. Patientsundergoing surgery had worse baseline symptoms andfunctional status than those initially treated nonsurgi-cally. By 10 years, 25% of surgical patients had undergoneat least one additional lumbar spine operation, and 25%of nonsurgical patients had at least one lumbar spineoperation. At 10-year follow-up, 69% of patients initiallytreated surgically reported improvement in their predom-inant symptom (back or leg pain) versus 61% of thoseinitially treated nonsurgically (P � 0.2). A larger propor-tion of surgical patients reported that their low back and

leg pain were much better or completely gone (56% vs.40%, P � 0.006) and were more satisfied with their currentstatus (71% vs. 56%, P � 0.002). Treatment group differ-ences persisted after adjustment for other determinantsof outcome in multivariate models. Change in the modi-fied Roland back-specific functional status scale favoredsurgical treatment, and the relative benefit persisted overthe follow-up period. Despite these differences, work anddisability status at 10 years were comparable amongthose treated surgically or nonsurgically.

Conclusions. Surgically treated patients with a herni-ated lumbar disc had more complete relief of leg pain andimproved function and satisfaction compared with non-surgically treated patients over 10 years. Nevertheless,improvement in the patient’s predominant symptom andwork and disability outcomes were similar regardless oftreatment received. For patients in whom elective discec-tomy is a treatment option, an individualized treatmentplan requires patients and their physicians to integrateclinical findings with patient preferences based on theirsymptoms and goals.

Key words: sciatica, herniated lumbar disc, prospec-tive cohort study, outcome research, lumbar disc surgery,natural history. Spine 2005;30:927–935

Sciatica resulting from a lumbar intervertebral disc her-niation is the most common cause of radicular leg pain inadult working populations.1 Such patients have a favor-able natural history associated with resorption of ex-truded disc material, but surgical treatment is frequentlyconsidered and performed in those with persistent orsevere symptoms.2 Indeed, discectomy is the most fre-quent lumbar spine operation in the United States.

The likelihood that a patient with an intervertebraldisc herniation will undergo surgery varies widely acrossgeographic regions.3–7 This may reflect different opin-ions among healthcare providers about the relative ben-efits of surgical and nonsurgical treatment.4 The primaryevidence for the relative benefit of surgical over nonsur-gical treatment comes from one randomized clinical trialperformed over 30 years ago. That study was conductedamong 126 inpatients at a single referral hospital in Nor-way.8 It reported that surgery was superior at 1 year andnonsignificantly better at 4 years compared with nonsur-gical treatment. However, after 10 years, treatment out-comes were similar.

No comparable data are available from patientstreated in the United States, but a large multicenter ran-domized trial is currently ongoing.9 For individuals

From the *General Medicine Division and the Clinical EpidemiologyUnit, Medical Services, Massachusetts General Hospital, HarvardMedical School, Boston, MA; †the Center for the Evaluative ClinicalSciences, Dartmouth Medical School, Hanover, NH; and ‡the Centerfor Cost and Outcomes Research and the Departments of Medicine andHealth Services, University of Washington, Seattle, WA.Acknowledgment date: March 8, 2004. First revision date: May 14,2004. Acceptance date: May 27, 2004.Supported by grants from the Agency for Healthcare Research andQuality (HS-06344, HS-08194, and HS-09804).The manuscript submitted does not contain information about medicaldevice(s)/drug(s).Federal funds were received in support of this work. No benefits in anyform have been or will be received from a commercial party relateddirectly or indirectly to the subject of this manuscript.Address correspondence and reprint requests to Steven J. Atlas, MD,General Medicine Division, Massachusetts General Hospital, 50 Stani-ford Street, Boston, MA 02114. E-mail satlas@partners.org

927

treated in contemporary clinical practice, the only pro-spective comparison of surgical and nonsurgical treat-ment is from an observational study.10 The Maine Lum-bar Spine Study has previously reported results after 1and 5 years11,12 that were similar to those from theWeber study.8 The goal of the current study was to assessthe relative benefits of surgical and nonsurgical treat-ment over a 10-year follow-up period using a broadrange of validated patient-reported outcome measures.

Materials and Methods

Details about the Maine Lumbar Spine Study design and meth-ods, and 1- and 5-year outcomes for patients with sciatica dueto an intervertebral disc herniation have been previously pub-lished.10–12 The study prospectively followed patients treatedby orthopedic surgeons, neurosurgeons, and occupationalmedicine physicians in community-based practices throughoutthe state of Maine. Treatment, either surgical or nonsurgicalcare, was determined in a routine clinical manner by the patientand the physician. Almost all patients undergoing surgery hadan open discectomy to remove an extruded disc fragment orherniation.11 Percutaneous discectomies were uncommon(4.3%), and no patient underwent fusion at the initial opera-tion. For nonsurgically treated patients, back exercises, physi-cal therapy, bedrest, spinal manipulation, narcotic analgesics,and epidural steroids were most frequently used, with mostpatients receiving several treatments.11

Study Population. The diagnosis of sciatica resulting from anintervertebral disc herniation was based on the physician’s as-sessment. Specific radiographic findings were not required forstudy entry. To restrict the study to patients for whom surgerywould be elective and acceptably safe, patients were excluded ifthey had prior lumbar spine surgery, cauda equina syndrome,developmental spine deformities, vertebral fractures, spine in-fection or tumor, inflammatory spondylopathy, pregnancy, orsevere comorbid conditions. A total of 507 patients were en-rolled from 1990 to 1992, with enrollment stratified to obtainroughly equal numbers of surgical and nonsurgical patients.10

Patients who initially chose nonsurgical treatment but under-went surgery before the first follow-up evaluation at 3 monthswere included in primary analyses as having been surgicallytreated (n � 38, 14%).11,12

Study Protocol. For eligible patients, written informed con-sent was obtained at study entry. Baseline interviews were con-ducted in person by trained research assistants. Follow-up wasobtained by mailed questionnaires at 3, 6, and 12 months, andthen yearly through 10 years. Physicians completed a detailedbaseline questionnaire including history, physical and neuro-logic findings, diagnostic procedure results, and planned treat-ment. Baseline imaging studies (CT, MRI, or myelogram) wereordered as directed by the treating physicians, and about halfwere available for independent review by a neuroradiologist.11

All study activities were approved by the Institutional ReviewBoards at the University of Washington, Seattle, ME MedicalCenter, Portland, ME, and Massachusetts General Hospital,Boston, MA.

Baseline Variables and Findings. Baseline findings, includ-ing demographic information, employment and disability sta-tus, comorbid conditions, past spine history, physical exami-

nation and imaging findings, symptoms, and functional status,have been reported previously.11,12 Patients initially receivingsurgical or nonsurgical treatment were similarly young,healthy, predominantly male, had not graduated from college,had been employed in the past month, and had symptoms forless than 6 months (Table 1). Patients treated surgically wereless likely than patients treated nonsurgically to be receivingWorkers’ Compensation but were more likely to have abnor-mal physical findings, moderate or severe findings on imagingstudies, severe back and leg pain, and greater back-related andoverall disability (Tables 1 and 2). Despite worse back-relatedsymptoms and findings, surgically treated patients had betterbaseline SF-36 general health perceptions. Patients not com-pleting 10-year follow-up (n � 107) had similar baseline char-acteristics as those not completing follow-up at 1 and 5years.11,12 Nonresponders tended to be younger and were sig-nificantly more likely to be male, smokers, not married orworking, and less educated (data not shown). Among otherbaseline characteristics, nonresponders also had significantlyworse low back pain severity and general health perceptions.

Table 1. Patient Characteristics and Features of BackDisorder at Baseline Evaluation

Characteristic*Surgical

(n � 217)Nonsurgical

(n � 183) P†

Age (yr) �mean (SD)� 42.2 (11.0) 41.8 (11.7) 0.77Gender, male 63.6 58.5 0.29Education, college graduate 34.1 26.8 0.11Smoking status, never or

quit � 6 mo ago59.3 63.9 0.34

Comorbid illnesses, yes‡ 26.7 21.9 0.26Employed in past 4 wk 59.9 50.3 0.05Receiving or applying for

Workers’ Compensation29.2 48.6 �0.001

Past episodes of back pain,none

20.7 23.0 0.59

Positive straight leg raise test 72.8 47.8 �0.001Abnormal examination findings

�mean (SD)� §1.1 (0.9) 1.4 (1.0) �0.001

Radiographic image reviewed(n)

115 75

Moderate or severe findings# 82.6 61.3 0.001Quebec classification, category

4 or 6**49.1 30.2 �0.001

Length of current episode,� 6 mo

70.5 70.9 0.94

Unilateral leg pain, yes 85.3 71.1 �0.001SF-36 score (0–100), mean (SD) ††

Physical function 31.4 (25.5) 46.7 (26.5) �0.001Bodily pain 19.2 (15.6) 35.5 (23.6) �0.001Role-emotional 48.2 (43.3) 57.4 (43.7) 0.04Mental health 61.4 (18.6) 66.6 (20.1) 0.009General health 79.1 (17.3) 72.8 (19.1) �0.001

*The data are expressed as the percentage, except as noted. Denominatorsdiffer slightly among variables because not all patients answered each ques-tion on the survey.†P values compare surgical and nonsurgical treatment groups using Fisher’sexact test or t test.‡Any self-reported chronic pulmonary disease, heart disease, stroke, cancer,or diabetes.§The mean no. of positive physical examination findings for a patient, includ-ing unilateral strength, sensation, or reflex abnormality (range, 0–3 findings).¶Any computerized tomography, magnetic resonance imaging, or myelogramavailable for independent review.#Global rating from normal to severe by study neuroradiologist blinded totreatment group and clinical information.**The Quebec classification system defines 11 categories including sciaticawith distal extremity radiation and neurologic findings (category 4), or theabove with radiographic findings of nerve root compression (category 6).††Higher scores indicate better function.

928 Spine • Volume 30 • Number 8 • 2005

Outcome Measures. On each follow-up questionnaire, pa-tients were asked to describe improvement in low back and legpain relative to baseline. Responses ranged from “muchworse” to “completely gone” on a 7-point scale. The primarysymptom outcome was improvement in the patient’s predom-inant symptom, either back or leg pain, as indicated at baseline.The outcome was categorized as improved if the response was“better,” “much better,” or “completely gone,” the same if theresponse was “about the same” or “a little better,” and worseif the response was “a little worse” or “much worse.” To assessfor a more definitive positive treatment outcome, symptom re-sponses were also classified as “much better” or “completelygone” compared with other responses.

Reports of symptoms in the past week were assessed atbaseline and follow-up, including the frequency (from 0 “not atall” to 6 “always”) and bothersomeness (from 0 “not bother-some” to 6 “extremely bothersome”) of low back pain, legpain, leg or foot weakness, leg numbness, and pain in the backor leg while sitting. Sciatica frequency and bothersome indexes,each with scores ranging from 0 to 24, were created by sum-ming the four leg-related questions.13 Back-specific functionalstatus was measured using the modified Roland disabilityscale.13 General health perceptions were assessed with a ques-tion from the Medical Outcomes Study Short Form 36-itemquestionnaire, “In general, would you say your health is” “ex-cellent” (1) to “poor” (5).14 For each variable, higher scoresindicate more severe symptoms or dysfunction.

Patients were considered to be satisfied with their currentstate if they replied that they were “delighted,” “pleased,” or“mostly satisfied” on a 7-point scale. Satisfaction with treat-ment decisions was assessed at follow-up visits by determiningif patients would still choose their initial treatment. Disabilityand work status were assessed at follow-up and are reportedaccording to whether or not the patient was receiving Workers’Compensation at baseline. Finally, first operations for patientsinitially treated nonsurgically and reoperations for thosetreated surgically were assessed for all patients from physicianoffice records, state hospital discharge data, and patient re-sponses to follow-up surveys and telephone contact.

Analysis. The rating of current findings at follow-up used cat-egorical responses, so distributions were directly compared be-tween treatment groups using �2 tests or Fisher’s exact tests.Although baseline clinical features of the treatment groups dif-fered, there was considerable overlap. To adjust for baselinedifferences between the two treatment groups, logistic regres-sion models were used to estimate the marginal effect of surgi-cal compared with nonsurgical treatment for the predominantsymptom and satisfaction with the current state outcomes at 10years. Changes in symptoms and functional status were as-sessed by subtracting results at 10-year follow-up from those atbaseline, and linear regression models were used to examine theeffect of treatment group after controlling for baseline score. Inall analyses, the effect of the patient’s initial treatment decision,either surgical or nonsurgical care, was assessed. Reoperationrates among patients initially undergoing surgery and surgicalcrossover rates among those initially receiving nonsurgicaltreatment were assessed using survival analysis methods (in thiscase, time-to-an-event analysis) over 10 years. Secondary anal-yses of treatment effect considered the actual treatment re-ceived at 10-year follow-up.

To examine the pattern of change over time, repeated-measures analysis was performed using data from all follow-ups between 2 and 10 years (n � 468, 92%). Mixed-effectsmodels and logistic regression models with Generalized Esti-mating Equations were used to model the correlation structureof the repeated-measures within each patient.15 The treatment-by-time interaction was used to test how treatment effect dif-fered over time in these models. All analyses were performedusing a commercial software package (Statistical Analysis Sys-tem, SAS Institute, Cary, NC).

Results

Thirty of 507 eligible patients enrolled in the study diedduring follow-up; a 93.4% 10-year survival rate (usingsurvival analysis). Among those alive after 10 years, out-comes were available for 400 of 477 (84%) patients: 217

Table 2. Change in Symptoms and Functional Status From Baseline to 10-Yr Follow-up According to Initial Treatment

Surgical Treatment* Nonsurgical Treatment*

Variable �mean score (SD)�† Baseline‡ 10 Yr‡ Change§ Baseline‡ 10 Yr‡ Change§ P¶

Low back pain in the past weekFrequency score 4.4 (2.0) 1.9 (1.8) �2.5 (2.3) 3.7 (2.1) 2.3 (1.8) �1.4 (2.4) �0.001Bothersome score 4.2 (2.0) 1.9 (2.0) �2.3 (2.5) 3.5 (2.0) 2.3 (1.8) �1.2 (2.4) 0.007

Leg pain in the past weekFrequency score 5.1 (1.4) 1.5 (1.8) �3.6 (2.1) 3.5 (2.1) 1.8 (1.7) �1.7 (2.3) �0.001Bothersome score 5.3 (1.4) 1.5 (1.9) �3.8 (2.2) 3.6 (2.2) 1.7 (1.8) �1.9 (2.4) 0.002

Sciatica IndexFrequency 17.8 (5.2) 6.1 (6.6) �11.8 (7.6) 12.2 (6.6) 6.7 (6.0) �5.5 (7.9) 0.01Bothersome 17.6 (5.1) 5.7 (6.7) �11.9 (7.4) 11.8 (6.4) 6.1 (6.0) �5.8 (7.4) 0.004Modified Roland scale 17.7 (4.0) 6.0 (7.0) �11.7 (7.2) 13.3 (5.9) 7.6 (7.0) �5.8 (7.6) �0.001SF-36 general health perceptions** 2.2 (1.1) 2.4 (1.0) 0.26 (1.2) 2.4 (1.0) 2.6 (0.9) 0.25 (1.1) 0.20

*The no. of patients reporting each variable ranged from 188 to 217 for surgical patients and 152 to 183 for nonsurgical patients.†Low back and leg pain scores range from 0 to 6. Sciatica frequency and bothersome indices are the sum of 4 questions (each response scored from 0 to 6): legpain, leg or foot weakness, leg numbness, and pain in the back or leg while sitting. Sciatica Index scores range from 0 to 24, modified Roland Scale scores rangefrom 0 to 23, and SF-36 general health perceptions range from 1 to 5.‡Higher mean scores at baseline and at 10 yr indicate worse symptoms or function for all variables.§Change is calculated as score at 10 yr minus score at baseline. Negative values indicate improvement for all variables.¶P values compare the change between surgical and nonsurgical treatment groups using multiple linear regression models that control for baseline score.**Assessed using the SF-36 question: “In general, would you say your health is” “excellent” (1) to “poor” (5).

929Ten-Year Outcomes of Sciatica • Atlas et al

of 255 (85%) initially treated surgically and 183 of 222(82%) initially treated nonsurgically.

Patients’ Global Evaluation at 10 YearsPatients were asked to contrast the current status of theirlow back and leg pain symptoms at 10 years with that atbaseline. A slightly higher percentage of patients treatedsurgically reported improved low back pain comparedwith patients treated nonsurgically (69% vs. 59%, P �0.11), but leg pain improvement was similar in bothtreatment groups (Table 3). The predominant painsymptom, either low back or leg pain based on the pa-tient’s report at entry, was improved in 69% of surgicalpatients compared with 61% of nonsurgical patients(P � 0.19). However, the predominant symptom wasreported to be “much better” or “completely gone” in56% of surgical patients compared with 40% of nonsur-gical patients (P � 0.006).

Although improvement in the primary symptom mea-sures was not significantly different between treatmentgroups, patient satisfaction at 10-year follow-up was sig-nificantly better among those treated surgically (Table

3). Satisfaction with the patient’s current state was re-ported by 70.5% of surgically treated patients versus55.5% of nonsurgically treated patients (P � 0.002).Eighty-seven percent of patients treated surgically wouldstill definitely or probably choose their initial treatmentcompared with 76% of those treated nonsurgically (P �0.006).

Work and disability outcomes were similar regardlessof initial treatment received (Table 3). Among patientsreceiving Workers’ Compensation at study entry, 19%of surgically treated and 17% of nonsurgically treatedpatients were receiving some form of disability compen-sation at 10 years (P � 0.83). Patients who were workingat study entry were very unlikely to be receiving disabil-ity compensation at 10 years (5%), regardless of treat-ment. The percentage of patients working at 10-year fol-low-up was also similar regardless of initial work,disability, or treatment status.

Change in Symptoms and Functional Status Over10 Years

Patients rated the frequency and bothersomeness of lowback pain and leg symptoms in the past week at baselineand 10 year follow-up (Table 2). For both low back painand leg pain, surgically treated patients rated the fre-quency and severity of symptoms as worse at baselineevaluation and better at 10-year follow-up comparedwith patients treated nonsurgically. The frequency andbothersomeness of sciatica symptoms were also worse atbaseline and better at follow-up for surgical patients. Foreach symptom question, the change in symptoms at 10years was significantly greater among surgically treatedpatients than among nonsurgically treated patients (allP � 0.01). Similarly, surgically treated patients reportedsignificantly greater improvement in back-specific func-tional status using the modified Roland scale (11.7 pointimprovement vs. 5.8 point improvement, P � 0.001).Although back-related symptoms and function im-proved over 10 years, general health perceptions wors-ened slightly to a comparable degree in both treatmentgroups.

Independent Predictors of Symptom Improvement andSatisfaction at 10 Years

Initial treatment was not randomly assigned. At baseline,surgical patients on average had more severe symptomsand findings than nonsurgical patients. Variables in allmodels included treatment group, age, and sex as well asindependent baseline predictors of outcome. Logistic re-gression models examining the independent effect oftreatment on the patients’ predominant symptom andsatisfaction with the current state at 10 years confirmedthe unadjusted findings in Table 3 (Table 4). Specifically,when “improved” was defined as a patient response of“better,” “much better,” or “completely gone,” surgicaltreatment was not a statistically significant independentpredictor of the predominant symptom outcome (oddsratio [OR], 1.4; 95% confidence interval [CI], 0.9–2.3).However, if symptom improvement was restricted to re-

Table 3. Patient Reported Improvement in Symptoms andSatisfaction and Disability Status at10-Year Follow-up

10-Year Outcomes (%) Surgical Nonsurgical P*

Low back pain vs. baseline† (n � 212) (n � 179) 0.11Improved 68.9 58.7Same 20.3 27.9Worse 10.9 13.4

Leg pain compared to baseline† (n � 208) (n � 171) 0.56Improved 69.2 64.3Same 19.2 23.4Worse 11.5 12.3

Predominant symptom comparedto baseline†‡

(n � 207) (n � 175) 0.19

Improved 69.1 61.1Completely gone 32.9 16.6Much better 22.7 23.4Better 13.5 21.1Same 18.4 25.7Worse 12.6 13.1

Satisfied with current state, yes§ 70.5 55.5 0.002Still choose the same initial

treatment, yes¶86.6 75.7 0.006

If receiving Workers’Compensation at entry,current status

(n � 63) (n � 88)

Receiving any disabilitycompensation

19.1 17.1 0.83

Working 81.0 75.0 0.43If employed at entry, current

status(n � 130) (n � 92)

Receiving any disabilitycompensation

5.4 5.5 1.00

Working 81.5 82.6 0.86

*P values assessed using Fisher’s exact or �2 tests for categorical variables.†Symptom severity was reported to be improved if the response was “bet-ter” to “completely gone,” the same if the response was “about the same”or “a little better,” and worse if the response was “a little worse” or “muchworse.”‡The predominant symptom, either back or leg pain, as rated by the patient atbaseline.§N � 217 for surgical and N � 182 for nonsurgical cohorts.¶N � 217 for surgical and N � 181 for nonsurgical cohorts.

930 Spine • Volume 30 • Number 8 • 2005

sponses of “much better” or “completely gone,” surgicaltreatment was significantly associated with a favorableoutcome (OR, 2.1; 95% CI, 1.2–3.7). Patients treatedsurgically were also more likely to be satisfied with theircurrent state (OR, 2.2; 95% CI, 1.4–3.6).

Depending on the outcome variable, several otherbaseline factors were independently associated with out-come (Table 4). In all models, receiving Workers’ Com-pensation at baseline was associated with worse out-comes (OR, 0.4 – 0.6). Greater severity of baselinesymptoms was also associated with worse outcomes, al-though the particular symptom measure (low back painor sciatica bothersomeness) varied depending on the out-come selected (OR, 0.7–0.8). On the other hand, betterSF-36 general health status at study entry was associatedwith improved symptoms, while SF-36 better mentalhealth status was associated with higher satisfaction at10 years (OR, 1.2 per 10-point change).

Time Course of OutcomesMean scores at baseline and at each follow-up over 10years for the frequency of sciatica symptoms and back-specific functional status are shown in Figure 1. The per-centage of patients reporting satisfaction with their cur-rent state at each follow-up is shown in Figure 2. Aspreviously reported,12 most of the improvement in out-come due to surgery was seen shortly after patients’ entryinto the study. From years 2 through 10, there was asmall but significant improvement in the frequency ofsciatica symptoms and satisfaction over time in both

Table 4. Independent Baseline Predictors of Patient Symptoms and Satisfaction at 10 Yr

Baseline Features*

Predominant Symptom†

Satisfied With Current State‡Improved � At Least Better Improved � At Least Much Better

OR (95% CI) P OR (95% CI) P OR (95% CI) P

Initial treatment, surgicalUnadjusted 1.4 (0.9–2.2) 0.10 1.9 (1.2–2.8) 0.003 1.9 (1.3–2.9) 0.002Adjusted model 1.4 (0.9–2.3) 0.17 2.1 (1.2–3.7) 0.01 2.2 (1.4–3.6) 0.001Education, college graduate NS 2.3 (1.4–3.9) 0.002 NSMarried or living together NS 2.0 (1.1–3.7) 0.02 NSReceiving Workers’ Compensation 0.6 (0.4–0.9) 0.02 0.4 (0.2–0.6) �0.001 0.5 (0.3–0.8) 0.002Comorbid illnesses, yes § NS NS 0.5 (0.3–0.9) 0.02Abnormal examination findings ¶ NS 1.3 (1.0–1.7) 0.01 NSLocation of pain, unilateral leg NS 2.6 (1.4–4.9) 0.003 NSLow Back Pain Score (0–12), per 2-point

increment0.8 (0.7–0.9) 0.002 NS 0.8 (0.7–0.9) 0.002

Sciatica Bothersome Index (0–24), per4-point increment

NS 0.7 (0.6–0.8) �0.001 NS

SF-36 General Health (0–100), per 10-point increment

1.2 (1.1–1.4) 0.004 1.2 (1.1–1.4) 0.005 NS

SF-36 Mental Health (0–100), per 10-point increment

NS NS 1.1 (1.0–1.3) 0.04

*Stepwise multiple logistic regression models were developed for the three outcomes reported. The table presents the final model for each outcome that adjustsfor initial treatment, age, gender, and other independent baseline predictors of outcome (c-statistic � 0.69, predominant symptom improved; 0.78, predominantsymptom much better; 0.74, satisfaction). NS � not (statistically) significant.†The predominant symptom, either back or leg pain, as indicated by the patient at baseline symptom was categorized as improved if the response was “better,”“much better,” or “completely gone.” To assess for a more definitive positive treatment outcome, symptom responses were also classified as “much better”or “completely gone.”‡Patient considered to be satisfied with their current state if they replied that they were “delighted,” “pleased,” or “mostly satisfied” on a 7-point scale.§Any self-reported chronic pulmonary disease, heart disease, stroke, cancer, or diabetes.¶ The no. of positive physical examination findings for a patient, including unilateral strength, sensation, or reflex abnormality (range, 0–3 findings).**Sum of low back pain frequency and bothersomeness questions in the past week at baseline (score range, 0–12; higher scores indicate worse low back pain).

Figure 1. Time course of symptom and functional status out-comes. Assessed at initial evaluation and at 3, 6, and 12 months,and then yearly follow-up for nonsurgical (-F-) and surgical (-E-)treatment. Mean � 2 SE. A, Sciatica Frequency Index. B, ModifiedRoland score.

931Ten-Year Outcomes of Sciatica • Atlas et al

treatment groups and to a similar extent. There was nochange in functional status between 2 and 10 years forpatients initially treated surgically, although there was asmall amount of improvement for nonsurgical patients(P � 0.2 for the interaction between time and treatmentgroup).

Lumbar Spine Surgery After Initial TreatmentAmong patients initially undergoing surgical treatment,the 10-year reoperation rate was 25% (n � 64; mediantime to reoperation, 24 months). Among patients ini-tially receiving nonsurgical treatment, the crossover rateto surgery between 3 months and 10 years was 25% (n �53; median time to crossover, 24 months). Baseline char-acteristics and findings were similar for surgical patientsundergoing a reoperation or not, and for nonsurgicalpatients having subsequent surgery or not (data notshown).

10-Year Outcomes According to ActualTreatment Received

Secondary analyses of treatment effect considered theactual treatment received at 10-year follow-up for the

138 patients treated completely nonsurgically, and forthe 262 patients treated surgically, including 51 patientsinitially in the nonsurgical group who crossed over tosurgical treatment. There were no significant outcomedifferences according to actual treatment received at 10-year follow-up for symptom improvement, satisfaction,work and disability status, and change in symptom fre-quency or bother. However, surgically treated patientshad greater improvement in Roland functional status(10.6 vs. 6.0 point change for surgical and nonsurgicalgroups, respectively, P � 0.03).

Ten-year outcomes were also compared according tothe patients’ initial and final treatment status: surgerywith or without a second spine operation (reoperation)and initial nonsurgical treatment with or without cross-over to surgical treatment after 3 months (Table 5). Pa-tients who remained in their initial treatment groupthroughout the study, especially those treated surgically,had better outcomes than those undergoing subsequentsurgery. Surgical patients having a reoperation or non-surgical patients having subsequent surgery had out-comes that were inferior to those initially receiving sur-gical treatment. For example, 51% and 40% of thosehaving a reoperation or crossing over to subsequent sur-gery were satisfied with the current state compared with76% and 61% of surgical or nonsurgical treatment con-tinuing without subsequent surgery. There was also atrend for patients undergoing subsequent surgical proce-dures to have worse disability and work outcomes thanthose continuing with their initial treatment, whethersurgical or nonsurgical.

Outcomes of Those Not Completing 10-Year Follow-upA total of 107 patients did not complete 10-year follow-up: 56 surgically and 51 nonsurgically treated. Eighty-six of these 107 patients (80%) returned at least 1 priorfollow-up survey (most recent follow-up: range, 3–108months; median, 60 months). Outcomes at the most re-cent follow-up reported by these patients were worse

Figure 2. Time course of satisfaction outcome. Assessed at 3, 6, and12 months, and then yearly follow-up for nonsurgical (-F-) and sur-gical (-E-) treatment. Percent satisfied with current state � 2 SE.

Table 5. 10-Yr Outcomes for Patients Based Upon Initial and Final Treatment Status

10-yr Outcomes

Initial Surgical Treatment Initial Nonsurgical Treatment

No Reoperation Reoperation Crossover to Surgery No Subsequent Surgery

(n � 166) (n � 51) P* (n � 45) P* (n � 138) P*

Predominant symptom, % improved† 71.8 60.8 0.14 48.8 0.006 64.9 0.21Satisfied with current state, % yes 76.5 51.0 �0.001 40.0 �0.001 60.6 0.003Sciatica Frequency Index, mean change (SD)‡ �12.5 (7.1) �9.3 (8.7) 0.008 �4.5 (7.4) �0.001 �5.9 (8.0) 0.02Roland scale, mean change (SD)‡ �12.6 (6.9) �8.4 (7.1) 0.002 �5.4 (7.6) �0.001 �6.0 (7.6) �0.001SF-36 general health perceptions, mean

change (SD)§0.16 (1.2) 0.60 (1.2) 0.02 0.22 (1.1) 0.08 0.26 (1.1) 0.19

If receiving Workers’ Compensation at entry,current status

(n � 48) (n � 15) (n � 22) (n � 66)

Receiving any disability compensation 12.5 40.0 0.02 27.3 0.14 13.6 0.86Working 85.4 66.7 0.12 68.2 0.10 77.3 0.28

*Using analysis of variance with Dunnett’s adjustment for multiple comparisons and using surgical with no reoperation group as reference.†The predominant symptom, either back or leg pain, as rated by the patient at baseline. Symptom severity was reported to be improved if the response was“better” to “completely gone.”‡Change is calculated as score at 10 yr minus the score at baseline. Negative values indicate improvement for all variables.§Assessed using the SF-36 question: “In general, would you say your health is” “excellent” (1) to “poor” (5).

932 Spine • Volume 30 • Number 8 • 2005

than those completing 10-year follow-up, but they weresimilarly worse for both surgical and nonsurgical pa-tients. If one includes outcomes from the last availablefollow-up survey (3–120 months) for the surgical (n �266, 97.4%) or nonsurgical (n � 220, 94.0%) cohort,the comparative results and conclusions are unchanged.For example, the predominant symptom was improved in66.4% of surgical and 58.2% of nonsurgical patients (P �0.06). In multiple regression models examining symptomimprovement and satisfaction with the current state thatincluded variables for treatment group, time of the last fol-low-up, and whether the patient died or not, only death wassignificantly associated with the outcome (patients lost tofollow-up because of death had better outcomes).

Discussion

Ten-year follow-up in this study was available for 84%of surviving patients with sciatica due to a herniatedlumbar disc initially treated either surgically or nonsur-gically. Although patients treated surgically had moresevere symptoms and worse functional status at baselinethan those treated nonsurgically, surgical patients re-ported better functional status and were more satisfiedwith their current state after 10 years. However, a similarpercentage of patients in both groups reported improvedsymptoms. Disability and work status were also similarregardless of initial treatment.

Although the Weber study has been criticized for notmeeting current standards for randomized trials,16 it isthe only randomized comparative trial of surgical versusnonsurgical treatment for patients with sciatica due to alumbar intervertebral disc herniation.8 Our results after5 years were similar to Weber’s 4-year results.12 At 10years, Weber reported good results in 63.6% of patientsinitially randomized to surgery and 56% of those ini-tially receiving conservative treatment, a difference thatwas not statistically significant. The narrowing of out-comes between 4 and 10 years represented a small im-provement for conservatively treated patients and worseoutcomes in surgically treated patients. In our study, fa-vorable 10-year outcomes were somewhat higher in bothtreatment groups (69% and 61%, respectively, amongsurgical and nonsurgical patients), but the absolute dif-ference in good results was nearly identical in the twostudies (see Figure 3 in accompanying manuscript).17

Like Weber’s findings, these overall results did notachieve statistical significance in unadjusted or adjustedanalyses, even with our larger sample size.

Few other prospective studies have reported long-term outcomes of discectomy, chemonucleolysis, or con-servative care for patients with an interververtebral discherniation. At 10 years, Gogan reported 77% of patientsinitially randomized to chymopapain and 38% of pa-tients receiving placebo injections were at least moder-ately improved.18 Other nonrandomized studies com-paring long-term outcomes of discectomy to othernonsurgical treatments have found modest differencesfavoring surgery.19,20

Although the relative benefit of surgical comparedwith nonsurgical treatment was small for improvementin the patient’s predominant pain symptom, other im-portant outcomes continued to favor surgery after 10years, including complete pain relief, back-specific func-tional status, and satisfaction. Few other comparablestudies have included the broad range of validated out-come measures used in the Maine Lumbar Spine Study.Nevertheless, these differences favoring surgical treat-ment did not translate into better disability and workoutcomes over 10 years. Few patients, irrespective ofinitial treatment or Workers’ Compensation status, werereceiving any disability compensation and approxi-mately 80% were working at 10 years. These resultssuggest that specific medical treatments may have lessimpact on disability and work status than factors such asworkplace accommodations, job characteristics (tasks,autonomy, satisfaction, etc.), and other income and localeconomic factors.21–24

Among those initially treated surgically, one in fourhad at least 1 additional spinal operation over 10 years.These reoperation rates are generally higher than thosereported in prior studies.19,25–27 For patients initiallytreated nonsurgically, another 1 in 4 had subsequentspine surgery between 3 months and 10 years. Few stud-ies have examined long-term rates of subsequent surgeryin patients initially managed conservatively.18,19 Al-though we lack detailed information about the indica-tions and findings that led to a subsequent operation,outcomes associated with these procedures were gener-ally poor. Prior studies have reported conflicting out-comes from patients undergoing subsequent sur-gery.28–31 Future research is needed to confirm thesegenerally poor findings and to perform interventionaltrials on patients considering these procedures.

Strengths of our study include prospective, long-termfollow-up in a large percentage of patients from contem-porary comparison groups treated in community-basedclinical practice using validated outcome measures, allmaking it likely that the results are broadly generaliz-able. The study’s major limitation is its observational,nonrandomized design.11,12 One cannot be certain thatthe differences in outcomes between treatment groupswere exclusively due to surgery, rather than unmeasuredconfounders. Important differences at baseline in demo-graphic, past history, and current symptoms, back-related disability, and objective findings existed amongthose patients treated surgically or nonsurgically. How-ever, the effect of surgical treatment on the predominantsymptom and satisfaction were similar after adjusting forother predictors of outcome in regression models. Also,the overall results are similar to those of Weber’s ran-domized trial, although both of our treatment groupshad somewhat better outcomes. Finally, although ourfollow-up rate is high, these 10-year results likely over-state the benefit of both surgery and conservative care toa similar extent based on dropouts in this study appearingto have worse outcomes than those remaining in the study.

933Ten-Year Outcomes of Sciatica • Atlas et al

Although our results support the relative benefit ofsurgery for properly selected patients, this does not meanthat nonsurgical treatment should no longer be a recom-mended initial treatment. Conservative treatment for atleast 4 weeks is recommended for patients with an inter-vertebral disc herniation without significant neurologicexamination findings.32 After 6 weeks, our results shouldnot be interpreted as requiring surgery in those withoutimprovement since many of the conservatively treatedpatients in this study had good outcomes. For those withvery bothersome symptoms, surgical treatment can has-ten recovery and may result in better long-term out-comes. However, pain outcomes narrow over time, asizeable number undergo additional surgery, and workand disability outcomes are not improved with surgicaltreatment. Important questions remain for patients andphysicians about optimal conservative treatment beyond6 weeks. Patients with persisting symptoms who are re-luctant to undergo surgery may continue conservativecare knowing that their symptoms will likely improve,but the magnitude of the improvement may not be asgreat as for surgical treatment. It is unclear whether thereis a time window beyond which the relative benefit ofsurgery diminishes and thus when conservative treat-ment should be deemed to have failed.

These results demonstrate how complex it is to inter-pret outcomes of studies that evaluate the relative impactof alternative treatments in patients with a lumbar discherniation. Patients need to better understand that theirspecific symptoms and concerns may determine how rel-ative treatment benefits and risks should be weighed.Differences in decision-making depend on whether themost important outcome is symptoms, function, satisfac-tion, disability, or work. Efforts are needed to integrateunbiased information that combines the patient’s perspec-tive with his or her clinical findings into the clinical deci-sion-making process.9,33 Additional work is needed to as-sess whether helping patients and their physicians translatethe complexity of these findings into individualized treat-ment plan results in improved outcomes of care.34,35

Key Points

● Outcomes of 400 patients with sciatica resultingfrom a lumbar disc herniation treated surgically ornonsurgically were followed over a 10-year period.● Because treatment was determined in a routineclinical manner by the patient and the physician,those surgically treated had more severe symptomsand worse functional status at baseline.● Patients initially treated surgically had morecomplete relief of leg pain and improved functionand satisfaction compared with patients initiallytreated nonsurgically over 10 years.● However, improvement in the patient’s predom-inant symptom and work and disability outcomeswere similar regardless of treatment received.

AcknowledgmentsThe authors thank YuChiao Chang, PhD, for assistancewith statistical analyses and Valerie Soucie for assistancewith managing long-term patient follow-up.

References

1. Frymoyer JW. Back pain and sciatica. N Engl J Med 1988;318:291–300.2. Andersson GBJ, Brown MD, Dvorak J, et al. Consensus summary on the

diagnosis and treatment of lumbar disc herniation. Spine 1996;24(suppl):75–8.

3. Cherkin DC, Deyo RA, Loeser JD, et al. An international comparison of backsurgery rates. Spine 1994;19:1201–6.

4. Keller RB, Atlas SJ, Soule DN, et al. Relationship between rates and out-comes of operative treatment for lumbar disc herniation and spinal stenosis.J Bone Joint Surg Am 1999;81:752–62.

5. Taylor VM, Deyo RA, Cherkin DC, et al. Low back pain hospitalization:recent United States trends and regional variations. Spine 1994;19:1207–13.

6. Volinn E, Mayer J, Diehr P, et al. Small area analysis of surgery for low backpain. Spine 1992;17:575–81.

7. Weinstein J, Birkmeyer J, eds. The Dartmouth Atlas of MusculoskeletalHealth Care. Chicago: American Hospital Association Press, 2000.

8. Weber H. Lumbar disc herniation: a controlled, prospective study with tenyears of observation. Spine 1983;8:131–40.

9. Birkmeyer NJ, Weinstein JN, Tosteson AN, et al. Design of the Spine Patientoutcomes Research Trial (SPORT). Spine 2002;27:1361–72.

10. Keller RB, Atlas SJ, Singer DE, et al. The Maine Lumbar Spine Study: I.Background and Concepts. Spine 1996;21:1769–76.

11. Atlas SJ, Deyo RA, Keller RB, et al. The Maine Lumbar Spine Study: II.1-year outcomes of surgical and nonsurgical management of sciatica. Spine1996;21:1777–86.

12. Atlas SJ, Keller RB, Chang Y, et al. Surgical and nonsurgical management ofsciatica secondary to a lumbar disc herniation: five-year outcomes from theMaine Lumbar Spine Study. Spine 2001;26:1179–87.

13. Patrick DL, Deyo RA, Atlas SJ, et al. Assessing health-related quality of lifein patients with sciatica. Spine 1995;20:1899–909.

14. Ware JE, Sherbourne CD. The MOS 36-item short-form survey (SF-36): 1.Conceptual framework and item selection. Med Care 1992;30:473–83.

15. Diggle PJ, Liang K, Zeger SL. Analysis of Longitudinal Data. New York:Oxford University Press, 1994.

16. Bessette L, Liang MH, Lew RA, et al. Classics in spine: surgery literaturerevisited. Spine 1996;21:259–63.

17. Keller RB, Atlas SJ. Is there a continuing role for prospective observationalstudies in spine research? Spine 2005;30:846–848.

18. Gogan WJ, Fraser RD. Chymopapain: a 10-year, double-blind study. Spine1992;17:388–94.

19. Nykbist F, Hurme M, Alaranta H, et al. Severe sciatica: a 13-year follow-upof 342 patients. Eur Spine J 1995;4:335–8.

20. Tregonning GD, Transfeld EE, McCulloch JA, et al. Chymopapain versusconventional surgery for lumbar disc herniation: 10-year results of treat-ment. J Bone Joint Surg Br 1991;73:81–6.

21. Dionne CD, Von Korff M, Koespell TD, et al. A comparison of pain, func-tional limitations, and work status indices as outcome measures in back painresearch. Spine 1999;24:2339–45.

22. Hadler NM. Back pain in the workplace: what you lift or how you liftmatters far less than whether you lift or when. Spine 1997;22:935–40.

23. Magora A. Investigation of the relation between low back pain and occupa-tion: V. Psychological aspects. Scand J Rehabil Med 1973;5:191–6.

24. Papageorgiou AC, Macfarlane GJ, Thomas E, et al. Psychosocial factors inthe workplace: do they predict new episodes of low back pain? Spine 1997;22:1137–42.

25. Keskimaki I, Seitsalo S, Osterman H, et al. Reoperations after lumbar discsurgery: a population-based study of regional and interspecialty variations.Spine 2000;25:1500–8.

26. Loupasis GA, Stamos K, Katonis PG, et al. Seven- to 20-year outcome oflumbar discectomy. Spine 1999;24:2313–7.

27. Moore AJ, Chilton JD, Uttley D. Long-term results of microlumbar discec-tomy. Br J Neurosurg 1994;8:319–26.

28. Cinotti G, Roysam GS, Eisenstein SM, et al. Ipsilateral recurrent lumbar discherniation: a prospective, controlled study. J Bone Joint Surg Br 1998;80:825–32.

29. Cinotti G, Gumina S, Giannicola G, et al. Contralateral recurrent lumbar

934 Spine • Volume 30 • Number 8 • 2005

disc herniation: results of discectomy compared with those in primary her-niation. Spine 1999;24:800–6.

30. Suk K-S, Lee H-M, Moon S-H, et al. Recurrent lumbar disc herniation:results of operative management. Spine 2001;26:672–6.

31. Vik A, Zwart JA, Hulleberg G, et al. Eight year outcome after surgery forlumbar disc herniation: a comparison of reoperated and not reoperated pa-tients. Acta Neurochir (Wien) 2001;143:607–10.

32. Deyo RA, Weinstein JN. Low back pain. N Engl J Med 2001;344:363–70.

33. Deyo RA. A key medical decision maker: the patient. Br Med J 2001;323:466–7.

34. Deyo RA, Cherkin DC, Weinstein J, et al. Involving patients in clinicaldecisions: impact of an interactive video program on use of back surgery.Med Care 2000;38:959–69.

35. Goldberg HI, Deyo RA, Taylor VM, et al. Can evidence change the rate ofback surgery? A randomized trial of community-based education. Eff ClinPract 2001;4:95–104.

935Ten-Year Outcomes of Sciatica • Atlas et al