TDW

SslmameMgttetdbwfisUstai

DA

1d

reatment of Lumbarisc Herniation: An Evidence-Based Review

ayne Moschetti, MD, Adam M. Pearson, MD, MS, and William A. Abdu, MD, MS

Several randomized controlled trials (RCTs) and prospective observational cohort studieshave compared surgical to conservative treatment for patients with sciatica caused bylumbar disc herniation. Whereas no RCT has been able to compare surgery with nonop-erative treatment without substantial crossover between treatment groups, multiple RCTsand observational studies have suggested that surgery resulted in faster improvement anda greater degree of improvement compared with nonoperative treatment. However, manypatients in these studies also experienced improvement with nonoperative care withoutadverse sequelae. This paper critically reviews the literature comparing surgery withnonoperative treatment for lumbar disc herniation.Semin Spine Surg 21:223-229 © 2009 Elsevier Inc. All rights reserved.

KEYWORDS Maine lumbar spine study, Spine Patient Outcomes Research Trial, sciatica,intervertebral disc herniation, low back pain, lumbar disc herniation

srormld

WRIpsntfwtFTdtScf“f

ciatica is defined as pain radiating in an area of the leg thatis served by a single nerve root in the lumbar or sacral

pine and may be associated with motor or sensory deficits. Aumbar intervertebral disc herniation (IDH) is the most com-

on cause of sciatica in working adults, with an estimatednnual incidence of 5 per 1000 adults.1,2 Diskectomy is theost common lumbar spine surgery, and more than 250,000

lective lumbar spine surgeries occur each year in the USA.3

any patients with sciatica will improve over time, yet sur-ery is frequently considered for patients with severe symp-oms or symptoms that persist. Most spine surgeons agreehat surgery should be offered only after a course of nonop-rative treatment for sciatica has failed. The most effectiveype and duration of conservative treatment has not beenetermined and varies substantially.4 In the USA, the rate ofack surgery was found to be approximately 40% higherhen compared with 11 other countries and was more thanve times the rate in England and Scotland.1 There is alsoignificant regional variation in the rate of diskectomy in theSA and internationally, suggesting that the indications for

urgery are also variable.3,5 Given this substantial variation inhe rate of surgery, it is clear that the appropriate timing ofnd indications for surgery are inconsistent. The economicmpact of back pain and sciatica is well known as demon-

epartment of Orthopaedics, Dartmouth Hitchcock Medical Center, Lebanon, NH.ddress reprint requests to Adam M. Pearson, MD, MS, Department of

Orthopaedics, Dartmouth Hitchcock Medical Center, 1 Medical Center

cDrive, Lebanon, NH 03756. E-mail: Adam.Pearson@hitchcock.org

040-7383/09/$-see front matter © 2009 Elsevier Inc. All rights reserved.oi:10.1053/j.semss.2009.08.005

trated by a Dutch study concluding that low back pain wasesponsible for more time off work and disability than anyther medical condition.6 In light of the uncertainty sur-ounding the outcomes of surgical and nonoperative treat-ent for lumbar IDH, this article aims to review the pertinent

iterature to assist spine care professionals in providing evi-ence-based recommendations to their patients.

eber’s Classicandomized Control Trial

n 1983, from a single referral center in Norway, Weber7

ublished the first randomized control trial (RCT) comparingurgery and nonoperative treatment for patients with a her-iated lumbar disc and radicular symptoms. Excluding pa-ients with “intolerable” pain and those with “no indication”or operative intervention, this paper looked at 126 patientsith “uncertain” indications for surgical treatment. These pa-

ients were randomized to surgical or nonoperative treatment.ollow-up examination was performed at 1, 4, and 10 years.hose undergoing surgery improved significantly more on aescriptive outcome scale (good, fair, poor, bad) compared withhose treated nonoperative at the 1-year follow-up examination.ixty-five percent of surgical patients had a “good” outcomeompared with 36% in the nonoperative group. At the 4-yearollow up, the surgical patients still showed better results (70%good” outcome vs 51% for nonoperative patients), but the dif-erence was no longer statistically significant. Only minor

hanges took place during the last 6 years of the study.

223

pmafrosWatirapRa

MIOcwcrts

pntomlpmoitpttdpmdcamovgl

ts

T

SP

I

E

T

L

O

F

SW

B

I

224 W. Moschetti, A.M. Pearson, and W.A. Abdu

Weber’s study represented the first effort to perform arospective RCT comparing surgery with nonoperative treat-ent for IDH. Similar to subsequent RCTs, this study was

ffected by a substantial number of patients crossing overrom nonoperative treatment to surgery. In the first year afterandomization, 17 of the 66 patients (26%) assigned to non-perative treatment underwent surgery while 1 patient as-igned to surgery refused and was treated nonoperatively.

eber performed both an intention to treat (ITT) and ans-treated analysis, with similar results for the 2 analyses. Byoday’s standards, this study would seem limited by vaguenclusion criteria, antiquated imaging techniques (magneticesonance imaging (MRI) was not routinely obtained in thessessment of patients), and a lack of validated outcomes andower analysis. However, 10-year follow-up for a similarCT has not been accomplished by any subsequent study,nd its findings are similar to more modern studies.

odern RCTsn a smaller, more recent, RCT with shorter follow-up,sterman et al8 attempted to assess the effectiveness of mi-

rodiskectomy for lumbar disc herniation. Fifty-six patientsith a lumbar disc herniation, clinical findings of nerve root

ompression, and radicular pain lasting 6-12 weeks wereandomized to microdiskectomy or an isometric physicalherapy program. In this study, no clinically or statisticallyignificant differences between the groups in leg or back

able 1 Surgical vs Prolonged Conservative Treatment for Sc

tudy design Multicenter RCT with ITT analysatients 283 total patients (141 early sur

41.7 for surgical patients andnclusion criteria Age 18-65, “incapacitating lumb

herniation correlating with symxclusion criteria Cauda equina, muscle paralysis,

radiculopathy in the past 12 mpregnancy, or severe coexistin

reatments compared Early (within 2 wks) microdiskecneeded, and physical therapy

oss to follow-up 4 of 141 lost to follow-up in theconservative treatment group.

utcome measures Primary: Rowland disability quesself-rating scale of global percassessed at 2, 4, 8, 12, 26, 38

Secondary: SF-36, Sciatica Freqperception, neurologic examin

indings No baseline differences between1.9 wks. 11% of early surgery39% of prolonged conservativewere no significant differencesrecovery was 4.0 wks for early(P < 0.001). Leg pain improvewas seen in all patient subgro

trengths Multicenter, prospective, RCT useaknesses Study compared optimal timing o

treatment was not specified, aottom line Early surgery resulted in faster r

TT, intention to treat; VAS, visual-analogue scale.

ain, Oswestry disability index (ODI), or quality of life wereoted at the 2-year follow-up. Compared with the nonopera-ive patients, the surgery group improved significantly moren the leg pain visual analog scale (VAS) at 6 weeks and wasore likely to be satisfied with their treatment at all fol-

ow-up visits other than at 1 year. In a subgroup analysis,atients with an L4-L5 disc herniation (n � 28) improvedore with surgery than with nonoperative treatment on all

utcome measures. Those with an L5-S1 herniation (n � 28)mproved to a similar degree with surgery and nonoperativereatment. Shortcomings of this study include the small sam-le size and the 36% crossover from nonoperative treatmento surgery. The authors reported that the study was poweredo detect a 15 point change on the VAS, while the observedifferences were 9 points on leg pain and 10 points on backain at 2 years. These differences may have been clinicallyeaningful, but the study was not sufficiently powered toetect them. In addition, 10 out of 28 (36%) patients in theontrol group crossed over to surgery and were analyzedccording to the ITT principle. As such, a beneficial treat-ent effect of surgery may have been obscured due to cross-

ver. The authors did note that an as-treated analysis re-ealed no significant differences, however, the nonoperativeroup included only 17 patients after the crossover occurred,imiting power even further.

A recent RCT by Peul et al9 compared early microdiskec-omy with prolonged nonoperative treatment followed byurgery if needed (Table 1). This study randomized 283 pa-

roup, 142 conservative treatment group). Average age wasr nonsurgical patients (range 18-65).l radicular syndrome” for 6-12 wks, MRI showing disc

s.ficient strength to move against gravity, similar episode ofvious spine surgery, bony stenosis, spondylolithesis,ase.s prolonged conservative care (education, analgesics ifse fearful of moving).al cohort. 3 of 142 were lost to follow-up in the

ire for Sciatica, 100-mm VAS for leg pain, 7-point Likertrecovery, defined as complete or nearly complete recoveryks.and Bothersome Index, 100-mm VAS for healthssessed at 8, 26 and 52 wks.

ps. Median time to surgery for early surgery group wasrecovered before surgery and did not undergo surgery.group underwent surgery at a median of 14.6 wks. Thereimary outcome measures at 52 wks. Median time tory group vs 12.1 wks for prolonged conservative care

ier in the early surgery group. The benefit of early surgerycept those whose sciatica was not provoked by sitting.lidated outcome measures.ery rather than efficacy of surgery. The nonsurgicalding was not possible. Follow-up was limited to 1 yr.

ry, though outcomes were no different by 1 yr.

iatica9

is.gery g43.4 foosacraptominsuf

o, preg disetomy vfor thosurgic

tionnaeived, 52 wuencyation a

grougroupcareon prsurge

d earlups exing vaf surg

nd blinecove

tPbtmVwps1gtdw

ssnttTlftffcht5upgw

ivsrlestcibstv

MDscpb5

pcscwtms“fsa“1lasrnewunmcwm

inaPogeatcgyn.oimttstsiebhsd

Treatment of lumbar disc herniation 225

ients to early surgery or prolonged nonoperative treatment.atients were 18-65 years old, had sciatica for 6-12 weeksefore enrollment and had MRI-confirmed disc herniationshat correlated with their symptoms. The primary outcomeeasures were the Roland Disability Questionnaire for Sciatica,AS for leg pain and 7-point Likert scale of perceived recovery,ith recovery defined as complete or nearly complete disap-earance of symptoms. Eighty-nine percent of patients as-igned to early surgery underwent surgery at a median of.9 weeks. The other 16 patients initially assigned to sur-ery had recovered before undergoing surgery. Of the pa-ients assigned to prolonged nonoperative care, 39% un-erwent surgery at a median of 14.6 weeks. All the patientsere followed for 52 weeks.There were no differences in the primary outcome mea-

ures at 1 year in the ITT analysis. Not surprisingly, the earlyurgery group improved more rapidly than the prolongedonoperative care group. This was quantified by comparinghe areas under the curves for the Roland Disability Ques-ionnaire and the leg pain VAS over the 52-week follow-up.his analysis revealed no significant differences for the Ro-

and Disability Questionnaire, whereas the results for leg painavored early surgery. The Kaplan–Meier curve comparingime with recovery also showed significantly faster recoveryor the early surgery group (median recovery time 4.0 weeksor early surgery vs 12.1 weeks for prolonged nonoperativeare, P � 0.001), though about 95% of patients in each groupad recovered by 1 year. It should be noted that about 10% ofhe nonoperative treatment group reached recovery between0 and 52 weeks, so the curves were substantially differentntil the final follow-up. In a subgroup analysis, a Cox pro-ortional hazards model demonstrated that the only sub-roup that did not benefit from early surgery was patientshose sciatica was not provoked by sitting.This study was a high quality RCT that had well-defined

nclusion and exclusion criteria, validated outcomes, andery low attrition. However, rather than comparing the re-ults of surgery and nonoperative treatment, it compared theesults of early surgery vs continued conservative care fol-owed by later surgery if necessary. This study focused onvaluating the optimal timing of surgery, not the efficacy ofurgery. The study does convincingly suggest that diskec-omy can be avoided in many patients who satisfy the indi-ations for surgery without any long-term harm. By contrast,t also indicates that patients who value a quicker recovery areest treated with early diskectomy. The limitations of thetudy included the inability to blind patients or researchers tohe treatment received, the lack of standardization of conser-ative care, and the high cross-over rate.

aine Lumbar Spine Studyespite the RCT being considered the most valid study de-

ign, the Maine lumbar spine study (MLSS), a prospectiveohort study, provides some of the best long-term data com-aring surgical and nonoperative treatment of sciatica causedy lumbar disc herniation (Table 2).10-13 The MLSS enrolled

07 (235 surgical and 272 nonoperative) patients from the m

ractices of 25 surgeons and 5 occupational medicine spe-ialists in Maine. Patients were enrolled if they suffered fromciatica, defined as pain radiating to below the knee, thoughonfirmatory imaging studies were not required. Given that itas an observational study, treatment was determined by the

reating physician and the patient. The primary outcomeeasure was self-reported improvement in the predominant

ymptom (leg pain or back pain). The exact definition ofimprovement” varied among the 3 reports (1, 5, and 10-yearollow-up), with patients reporting that their predominantymptom was “much better” or “completely gone” qualifyings improved in the 1-year report, whereas those answeringbetter” were also included in the “improved” group at 5 and0 years. Many other outcome measures, including back and

eg pain frequency and bothersomeness, sciatica frequencynd bothersomeness indexes, Roland disability scale, SF-36cores, and work status, were also recorded. Results wereeported at 1, 5, and 10 years, and statistical modeling tech-iques were used to control for the significant baseline differ-nces between the 2 groups. Substantial crossover occurred,ith 15% of patients who initially chose nonoperative treatmentndergoing surgery within 3 months, and 25% of the remainingonoperative patients undergoing surgery between 3 and 120onths. The authors addressed this by assigning patients who

rossed over within the first 3 months to the surgery group,hile analyzing those who subsequently crossed over after 3onths with the nonoperative group.The MLSS demonstrated that surgical patients were signif-

cantly more likely to report improvement in their predomi-ant symptom compared with the nonoperative patients at 1nd 5 years (71% vs 43%, P � 0.001 at 1 year; 70% vs 56%,� 0.001 at 5 years). By 10 years, the difference on this

utcome measure was no longer significant (69% of the sur-ical patients reported improvement vs 61% for the nonop-rative patients, P � 0.2). However, if only patients whonswered that their predominant symptom was “much bet-er” or “completely gone” were included in the improvedategory (as was the case for the 1-year results), the surgicalroup continued to have significantly better results at 10ears compared with the nonoperative group (56% “defi-itely improved” with surgery vs 40% nonoperative, P �

006). The surgical group also had significantly better resultsn most secondary outcome measures at all follow-up times,ncluding low back pain improvement, leg pain improve-

ent, sciatica frequency and bothersomeness indexes, andhe Modified Roland Scale. The proportion of patients re-urning to work and receiving disability compensation wasimilar for the 2 treatment groups at all follow-up times. Theiming of improvement varied between the 2 groups, with theurgery group reaching maximal improvement on the Mod-fied Roland Scale within the first year, whereas the nonop-rative group continued to make small gains on this outcomeetween 2 and 10 years. By 10 years, 25% of surgical patientsad undergone at least 1 additional spine operation, and aimilar percentage of patients in the nonoperative group un-erwent surgery between 3 months and 10 years. In sum-

ary, the surgery patients had greater improvement of pain,

fi

snmtsofwaotftcpi

nnitdetwwImriArpMc

T

SP

I

E

T

L

O

F

S

W

B

226 W. Moschetti, A.M. Pearson, and W.A. Abdu

unction, and satisfaction, whereas there were no differencesn work status between the 2 groups.

The MLSS was the first large-scale study to compareurgical and nonoperative outcomes for lumbar disc her-iation. Its strengths include its size, prospective nature,ulticenter involvement, long-term follow-up with rela-

ively low attrition, and use of validated outcome mea-ures. However, its limitations must be considered. Itsbservational design contributed to marked baseline dif-erences between the surgical and nonoperative groups,ith the surgical group generally having worse symptoms

nd fewer workers’ compensation patients than the non-perative group. Although statistical modeling could con-rol for these measured baseline differences, the potentialor confounding by unmeasured variables existed. Addi-ionally, specific radiographic findings correlating withlinical findings were not required as inclusion criteria, soatients without actual disc herniations may have been

able 2 Maine Lumbar Spine Study10-12

tudy design Prospective cohort study.atients 507 patients initially enrolled in th

nonsurgical patients.nclusion criteria Persistent and definitive symptom

studies were not required.xclusion criteria Prior lumbar spine surgery, cauda

infection, tumor, inflammatory sreatments compared Surgical vs nonoperative treatmen

most frequently included back enarcotics, and epidural steroids.

oss to follow-up In the 5 yr follow-up, outcomes wand 182 (78.4%) nonoperative. Isurviving patients (84%): 217 of

utcome measures Primary: improvement in primary sSecondary: leg and back pain freq

bothersomeness indexes, modifidisability, work status.

indings Surgical patients had worse sympmore on the pain, function, andgroups narrowed over time. At 1patients reporting that their predgone” (69% vs 61%, P � 0.2) btheir predominant symptom wasWork and disability outcomes wSurgical patients who underwentreatment who subsequently und

trengths Large, prospective study, long-tercommunity-based cohort genera

eaknesses Observational nonrandomized despatients without herniations maynonsurgical group, surgical patimailed questionnaire, attrition msurgery and nonoperative treatmnonoperative treatment to surge

ottom line Surgically treated patients improv10 yrs, though there were no disurgery was greater in the shortover time.

ncluded. There was a substantial crossover from initial b

onoperative treatment to surgery after 3 months (25% ofonoperative patients), and these patients were included

n the nonoperative group for analysis. A 10-year as-reated analysis was performed that showed no significantifferences between the two as-treated groups, with thexception of greater improvement on the Roland score forhe surgical group. This suggests that the benefit of surgeryas not underestimated by including patients who under-ent surgery beyond 3 months in the nonoperative group.

n fact, patients who initially chose nonoperative treat-ent and subsequently underwent surgery had the worst

esults of all, with only 40% of these patients reportingmprovement in their predominant symptom at 10 years.

final concern was the use of mail-in questionnairesather than actual clinical follow-up as this precluded re-eat physical examinations. Despite these limitations, theLSS currently offers the best long-term follow-up data

omparing surgery with nonoperative treatment for lum-

y. Average age was 43.2 for surgical patients and 42.6 for

iatica based on history and physical examination. Imaging

a, developmental spine deformities, vertebral fractures,lopathy, pregnancy, or severe comorbid conditions.gery consisted of open diskectomy. Nonsurgical treatmentes, physical therapy, bed rest, spinal manipulation,

ailable for 402 (79.3%) of 507 patients: 220 (80%) surgical10 yr follow-up, outcomes were available for 400 of 47785%) surgical and 183 of 222 (82%) nonsurgical patients.m (leg or back pain).and bothersomeness scores, sciatica frequency and

wland disability scale, SF-36, satisfaction with treatment,

t baseline. At 1, 5, and 10 yrs, surgical patients improvedction measures, though the differences between the twothere was no significant difference in the proportion ofnt symptom was “better”, “much better,” or “completely

n the two groups, though more surgical patients reportedh better” or “completely gone” (56% vs 40%, P � 0.002).milar regardless of initial treatment at all follow-up times.eration and patients initially choosing nonoperativent surgery had the worst outcomes.w-up, detailed data using validated outcome measures,to most practices.aging not required to diagnose disc herniation so somebeen included, more workers’ compensation patients inith more severe baseline symptoms, data gathered byve resulted in an overestimate of the benefits of boths dropouts had worse outcomes, 25% crossover fromween 3 mo and 10 yrs.re on pain, function and satisfaction outcomes at 1, 5 andes in work status between the two groups. The benefit ofand the differences between the two groups narrowed

e stud

s of sc

equinpondyt. Surxercis

ere avn the255 (ymptouencyed Ro

toms asatisfa0 yrs,omina

etwee“muc

ere sit reoperwe

m follolizableign, imhave

ents way haent ary beted mofferencterm,

ar disc herniation.

SOTmcutt1hfiwvtiaMdcS

tscittlc

rsystotnpba

T

S

P

I

E

T

L

O

F

SW

B

N

Treatment of lumbar disc herniation 227

pine Patientutcomes Research Trial

he largest study comparing surgical and nonoperative treat-ent for lumbar disc herniation is the Spine Patient Out-

omes Research Trial (SPORT) (Table 3).14-17 This study wasnique in that it included both randomized and observa-ional arms, which allowed patients to be included even ifhey did not agree to randomization. Patients were enrolled at3 multidisciplinary spine practices in 11 states. All patientsad radicular pain, neurological findings, and an MRI-con-rmed disc herniation that coincided with their symptoms,hich were at least 6 weeks in duration. The surgical inter-ention was a standard open diskectomy, and nonoperativereatment consisted of “usual care” that was suggested tonclude at least education and counseling, physical therapy,nd nonsteroidal anti-inflammatory medications if tolerated.any nonoperative patients also received narcotics and epi-

ural injections. The study was powered to detect a 10 pointhange from baseline on its primary outcome measures; the

able 3 Spine Patient Outcomes Research Trial15-17

tudy design RCT and observational cohort anathese cohorts were combined aregression was used to control

atients 501 total patients (RCT) and 743surgical patients and 43 for nonobservational cohort.

nclusion criteria >18 yrs old with intervertebral dispain and evidence of nerve rootwks.

xclusion criteria Prior lumbar surgery, cauda equintumor, inflammatory spodyloarth

reatments compared Standard open diskectomy vs nonPT and NSAIDs if tolerated).

oss to follow-up 85% available at 1 yr, 80% at 2 ynonoperative patients.

utcome measures Primary: SF-36 BP and PF scoresSecondary: self reported improvem

Bothersome index.indings In the RCT, 60% assigned to surg

surgery by 2 yrs. The ITT analysmeasures, though the surgery gtreated analysis showed clinicalat all follow up times on all outc

In the observational cohort, surgework status at 1 and 2 yrs.

The 4-yr data from the combinedsurgery on all outcomes other t

trengths Large, multicenter, included RCTeaknesses High level of crossover (45% by 2

treatment was “usual care” rathblinding of patients or providers

ottom line Surgery resulted in greater improvoperative treatment was effectivmeaningful ITT analysis of the R

SAID, nonsteroidal anti-inflammatory drug; PT, physical therapy; Bintention to treat.

hort Form-36 bodily pain and physical function scales18 and p

he ODI.19 Secondary outcome measures included patientelf-reported improvement, satisfaction with symptoms andare, work status, and the Sciatica Bothersome Index. Thenitial reports detailed the first 2 years of follow-up,15,17 whilehe 4-year follow-up data were recently reported.16 Addi-ional subgroup analyses evaluated the effect of herniationocation, morphology, and intervertebral level on out-omes.20,21

The most surprising result of the SPORT RCT was the highate of protocol nonadherence (ie, crossover from the as-igned treatment group to the other group). In the first 2ears, 40% of patients assigned to surgery never underwenturgery, whereas 45% of patients assigned to nonoperativereatment underwent surgical intervention. While the cross-ver from nonoperative treatment to surgery was expected,he high rate of crossover in the other direction (ie, surgery toonoperative treatment) was not. This high rate of crossoverrecluded meaningful analysis of the data on an ITT basisecause the 2 groups were very similar in treatment receivedt 2 years. Nonetheless, the ITT analysis was presented as the

separately at 1 and 2 yrs. Due to protocol nonadherence,as-treated analysis was performed at 4 yrs. Longitudinalseline differences.atients (observational cohort). Average age was 41.7 foral patients in RCT and 40.5 and 43.7, respectively, in the

niation confirmed by imaging that correlated with radicularion or neurologic deficit. Symptoms were present for >6

liosis >15°, segmental instability, fracture, infection,hy, pregnancy, comorbidities precluding surgery.tive treatment (“usual care”, including at least education,

65% at 4 yrs. Attrition was similar for surgery and

work status, satisfaction with symptoms and care, Sciatica

d 45% assigned to nonoperative treatment had undergonewed no significant differences on the primary outcomemproved more on the Sciatica Bothersome Index. The as-statistically significant advantages for the surgery groupeasures other than work status.

ients improved more on all outcome measures other than

nd observational cohorts showed a persistent benefit ofork status.bservational cohort, validated outcome measures.recluded meaningful ITT analysis of RCT, nonsurgical

n specific regimen, substantial attrition at 4 yrs, no

t compared with nonoperative treatment out to 4 yrs. Non-a large proportion of patients. Crossover preventedta, so the benefits of randomization were lost.

ily pain; PF, physical function; ODI, Oswestry disability index; ITT,

lyzednd anfor batotal psurgic

c herirritat

a, scoropatopera

rs, and

, ODI.ent,

ery anis shoroup ily andome mry pat

RCT ahan wand oyrs) p

er tha.emene forCT da

P, bod

rimary analysis. As would be expected, there were no sig-

nDtaapdfsn

to2pOWgnpaaaggwg

ttpttebstaha

gttmlpoeWcbntttes

tcgsGss

cncabpetnor

DAtif(peaitcsvstsSemffadtttitty

sstr

228 W. Moschetti, A.M. Pearson, and W.A. Abdu

ificant differences on any of the primary outcome measures.espite the homogeneity of treatment across the 2 groups,

he surgery group did improve significantly more on the sci-tica bothersome index over the first 2 years (�10.1 vs �8.5t 2 years, P � 0.003). A secondary as-treated analysis waserformed with adjustment for potential confounders. Thisemonstrated large and statistically significant treatment ef-ects of surgery on the 3 primary outcome measures (ie, theurgery group improved 15 points more on the ODI than theonoperative group at 1 year).Similar to the as-treated analysis of the RCT, the observa-

ional trial demonstrated a large, significant benefit of surgeryn all primary and secondary outcome measures over the firstyears after controlling for potential confounders. For exam-le, the surgery patients improved 13 points more on theDI at 2 years compared with the nonoperative patients.ork status was the 1 outcome measure on which the sur-

ery patients did not improve significantly more than theonoperative patients at 2 years. Given the aforementionedroblems with the ITT analysis, the 4-year data from the RCTnd observational cohorts were combined, and an adjusteds-treated analysis was performed. The differences observedt the 2-year follow-up persisted at 4 years, with the surgeryroup improving significantly more than the nonoperativeroup on all primary and secondary outcomes other thanork status. Unlike the MLSS, the differences between the 2roups did not appear to decrease over time.

As-treated subgroup analyses were performed to evaluatehe effect of herniation location, morphology, and interver-ebral level on outcomes. These demonstrated that althoughatients with central disc herniations had worse back painhan patients with lateral herniations, and patients with pro-rusions had less severe symptoms at baseline than those withxtrusions or sequestrations, the change from baseline inack pain scores was similar across location and morphologyubgroups.21 Another subgroup analysis reported that pa-ients with upper (L2-3 or L3-4) lumbar disc herniations hadgreater treatment effect of surgery than patients with L5-S1erniations.20 Patients with L4-5 herniations had intermedi-te treatment effects.

SPORT represents the largest study to date comparing sur-ery with nonoperative treatment for lumbar disc hernia-ions. Its strengths include the use of stringent inclusion cri-eria, validated back-specific and general health outcomeeasures, powerful statistical modeling to account for base-

ine differences between the groups, and the inclusion ofatients from 13 different spine centers. The major limitationf the SPORT RCT was the high rate of protocol nonadher-nce that prevented a meaningful ITT analysis of the data.

hile the as-treated analysis was carefully adjusted to ac-ount for the baseline differences between the 2 groups, theenefits of randomization were lost. As a result, there is stillo Level I evidence comparing surgery with nonoperativereatment for the treatment of lumbar disc herniations. Givenhat it is not possible to control for unmeasured confounders,he as-treated analysis may have overstated the treatmentffect of surgery. Another limitation of SPORT and all other

tudies of disc herniation was the lack of blinding of patients a

o treatment received. Sham surgery is the only way to over-ome this threat to validity, and although it has been sug-ested, it seems unlikely that patients would enroll in such atudy or that institutional review boards would approve it.22

iven the lack of blinding, perhaps the placebo effect ofurgery may have contributed to the better outcomes for theurgery patients.

Proponents of various nonoperative therapies have criti-ized studies like SPORT for failing to specify a well-definedonoperative regimen.23 The advantage of specifying “usualare” as the nonoperative treatment was the increased gener-lizability this provided as it more closely reflected what waseing offered to patients. However, patients may have im-roved more with a specific nonoperative treatment regimen,specially as they had failed to improve with “usual care” overime leading up to enrollment. Unfortunately, the best type ofonoperative care for disc herniations has not been rigor-usly defined, so it was not possible to select a more specificegimen based on the available scientific literature.

iscussionll the major studies comparing surgery with nonoperative

reatment of lumbar disc herniation have reached the follow-ng conclusions: (1) surgery tends to improve symptomsaster and to a greater degree than nonoperative treatment,2) surgery is safe and the complication rates are low, (3)atients who initially meet the indications for surgery butlect nonoperative treatment may eventually improve to ancceptable level of pain and function, (4) surgery does notmprove the return to work rate compared with nonoperativereatment, and (5) nonoperative treatment is safe. The surgi-al outcomes of SPORT and MLSS patients were remarkablyimilar, though the nonoperative outcomes were not as fa-orable in the MLSS, which resulted in SPORT having amaller treatment effect of surgery. While all enrolled pa-ients met the indications for surgery, 55% of patients as-igned to nonoperative treatment avoided surgery in thePORT RCT as did 61% of the patients assigned to nonop-rative treatment in the study by Peul et al.9 This suggests thatany patients who can tolerate living with their symptoms

or a period may be able to improve to an acceptable level ofunction without surgery if they choose. Despite consistentgreement across studies about the benefit of surgery, theuration of benefit is still unclear. While Weber’s study andhe MLSS suggested that the treatment effect of surgery tendso decrease with time, the SPORT data gave no indication thathe treatment effect was decreasing between 2 and 4 years. Its anticipated that SPORT will continue to follow patients outo 10 years, however, the MLSS and Weber’s study suggestshat there is very little change in outcomes between 5 and 10ears.

Another striking similarity across all of these studies thatpan 4 decades is the relatively high rate of crossover tourgery for patients who initially chose or were randomizedo nonoperative treatment. Weber reported a 26% crossoverate, Osterman et al. 36%, Peul et al. 39%, the MLSS 40%,

nd SPORT 45%.7-9 These results indicate that it is may not

bnswetgtcs

tgmUsnptTpttaeaan

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

Treatment of lumbar disc herniation 229

e possible to conduct an effective RCT comparing surgery toonoperative treatment for lumbar disc herniation. At theame time, these results are not surprising given that patientsho fail nonoperative treatment are unlikely to continue in-

ffective treatment. As would be expected, the SPORT pa-ients who crossed over from nonoperative treatment to sur-ery had worse baseline symptoms and the perception thatheir symptoms were becoming worse, whereas those whorossed over from surgery to nonoperative treatment had lessevere symptoms that were improving.

Based on the evidence, most spine providers now agreehat, on average, diskectomy results in a more rapid andreater degree of improvement, though nonoperative treat-ent can be successful in a large proportion of patients.nfortunately, our ability to predict which patients will be

uccessful with nonoperative treatment is quite limited. Theext step for researchers is to use the available data to developredictive models to improve our ability to determine whichreatment is the most appropriate for individual patients.hese studies should allow a shared decision model whereatients can determine their preference for care based onheir symptoms, values, and the available evidence.24 Such aool would allow many patients to avoid unnecessary surgerynd others to avoid prolonged periods of ineffective nonop-rative treatment. While all the studies reviewed here providenswers for the “average” patient, we now have to work onpplying evidence to the treatment of individual patients,one of whom are “average.”

eferences1. Cherkin DC, Deyo RA, Loeser JD, et al: An international comparison of

back surgery rates. Spine 19:1201-1206, 19942. Frymoyer JW: Back pain and sciatica. N Engl J Med 318:291-300, 19883. Taylor VM, Deyo RA, Cherkin DC, et al: Low back pain hospitalization.

Recent United States trends and regional variations. Spine 19:1207-1212, 1994; discussion: 13

4. Andersson GB, Brown MD, Dvorak J, et al: Consensus summary of thediagnosis and treatment of lumbar disc herniation. Spine 21:75S-78S,1996

5. Weinstein JN, Lurie JD, Olson PR, et al: United States’ trends andregional variations in lumbar spine surgery: 1992-2003. Spine 31:2707-2714, 2006

6. van Tulder MW, Koes BW, Bouter LM: A cost-of-illness study of backpain in The Netherlands. Pain 62:233-240, 1995

7. Weber H: Lumbar disc herniation. A controlled, prospective study with

ten years of observation. Spine 8:131-140, 1983

8. Osterman H, Seitsalo S, Karppinen J, et al: Effectiveness of microdis-cectomy for lumbar disc herniation: a randomized controlled trial with2 years of follow-up. Spine 31:2409-2414, 2006

9. Peul WC, van Houwelingen HC, van den Hout WB, et al: Surgeryversus prolonged conservative treatment for sciatica. N Engl J Med356:2245-2256, 2007

0. Atlas SJ, Deyo RA, Keller RB, et al: The Maine Lumbar Spine Study, PartII. 1-year outcomes of surgical and nonsurgical management of sciatica.Spine 21:1777-1786, 1996

1. Atlas SJ, Keller RB, Chang Y, et al: Surgical and nonsurgical manage-ment of sciatica secondary to a lumbar disc herniation: five-year out-comes from the Maine Lumbar Spine Study. Spine 26:1179-1187,2001

2. Atlas SJ, Keller RB, Wu YA, et al: Long-term outcomes of surgical andnonsurgical management of sciatica secondary to a lumbar disc herni-ation: 10 year results from the Maine Lumbar Spine Study. Spine 30:927-935, 2005

3. Keller RB, Atlas SJ, Singer DE, et al: The Maine Lumbar Spine Study,Part I. Background and concepts. Spine 21:1769-1776, 1996

4. Birkmeyer NJ, Weinstein JN, Tosteson AN, et al: Design of the SpinePatient Outcomes Research Trial (SPORT). Spine 27:1361-1372, 2002

5. Weinstein JN, Lurie JD, Tosteson TD, et al: Surgical vs nonoperativetreatment for lumbar disk herniation. The Spine Patient OutcomesResearch Trial (SPORT) observational cohort. JAMA 296:2451-2459,2006

6. Weinstein JN, Lurie JD, Tosteson TD, et al: Surgical versus nonopera-tive treatment for lumbar disc herniation: four-year results for the SpinePatient Outcomes Research Trial (SPORT). Spine 33:2789-2800, 2008

7. Weinstein JN, Tosteson TD, Lurie JD, et al: Surgical vs nonoperativetreatment for lumbar disc herniation. The Spine Patient OutcomesResearch Trial (SPORT): a randomized trial. JAMA 296:2441-2445,2006

8. Ware JE Jr, Sherbourne CD: The MOS 36-item short-form health sur-vey (SF-36). I. Conceptual framework and item selection. Med Care30:473-483, 1992

9. Daltroy LH, Cats-Baril WL, Katz JN, et al: The North American spinesociety lumbar spine outcome assessment instrument: reliability andvalidity tests. Spine 21:741-749, 1996

0. Tosteson AN, Skinner JS, Tosteson TD, et al: The cost effectiveness ofsurgical versus nonoperative treatment for lumbar disc herniation overtwo years: evidence from the Spine Patient Outcomes Research Trial(SPORT). Spine 33:2108-2115, 2008

1. Pearson AM, Blood EA, Frymoyer JW, et al: SPORT lumbar interverte-bral disk herniation and back pain: does treatment, location, or mor-phology matter? Spine 33:428-435, 2008

2. Flum DR: Interpreting surgical trials with subjective outcomes: Avoid-ing UnSPORTsmanlike conduct. JAMA 296:2483-2485, 2006

3. Brox JI, Sorensen R, Friis A, et al: Randomized clinical trial of lumbarinstrumented fusion and cognitive intervention and exercises in pa-tients with chronic low back pain and disc degeneration. Spine 28:1913-1921, 2003

4. Weinstein JN: The missing piece: embracing shared decision making to

reform health care. Spine 25:1-4, 2000