Evidence-Based Recommendations for Spine Surgery

Evidence-Based Recommendations for Spine SurgeryAlexander R. Vaccaro, MD, PhD; Charles G. Fisher, MD; Peter G. Whang, MD; Alpesh A. Patel, MD;Srinivas K. Prasad, MD; Peter D. Angevine, MD, MPH; Kishore Mulpari, MBBS, MHSc; Ken C. Thomas, MD, MHSc

Imaging strategies for low-back pain:systematic review and meta-analysis.Chou R, Fu R, Carrino JA, et al. Lancet 2009;373:463–72.

STUDY SUMMARY

Lumbar spinal imaging is commonly utilized in theevaluation of low back pain. The rationale for imag-

ing, either plain radiographs, computed tomography(CT), or magnetic resonance imaging (MRI), is primarilybased on identifying anatomical sources of pain.Unfortunately, the correlation between findings on im-aging and clinical symptoms can be limited. A number ofstudies have been preformed to elucidate the value ofspinal imaging in the setting of acute back pain. The defi-nition of “value”, however, varies from study to study.Investigations have focused separately on diagnostic infor-

mation, treatment interventions, patient outcomes, orpatient satisfaction. Additionally, the inclusion and ex-clusion criteria for these studies have not been uniform.Specifically the definition of “red flags” for serious dis-ease (fevers, weight loss, neurological deficits, etc.) aresubjective and, in some instances, not defined. Nonethe-less the clinical question remains: is immediate routinelumbar spine imaging more effective than usual carewithout imaging in patients with low back pain and nosuggestion of “red flags.” Chou et al. used methods ofsystematic review and meta-analysis to address thisquestion.

METHODOLOGICAL REVIEW

The authors conducted a systematic review and meta-analysis with sound and reproducible methodology. Themethods used to select articles were well described withclear inclusion and exclusion criteria. Relevant outcomemeasures formed part of the inclusion criteria. Random-ized controlled trials that compared immediate lumbarimaging to routine care for low back pain patients with-out indication of serious underlying conditions wereconsidered. Trials were included if their outcomes re-ported on pain, function, mental health, quality of life,patient satisfaction and overall patient reported im-provement. Two reviewers independently assessed titlesand abstract for study inclusion. Two reviewers ab-stracted data from included papers and ranked each pa-per as higher or lower quality according to accepted andpre-specified criteria. Two primary outcomes were iden-tified: pain and function. Secondary outcomes includedmental status, quality of life, patient satisfaction andoverall improvement. Outcomes were categorized asshort-term (�3 months), long-term (3–6 months) andextended (�1 year). Given that different trials used dif-ferent scales for measurement of specific outcomes, out-come scores were standardized to allow for pooling. Theauthors defined clinically important differences in outcomemeasures a priori, diminishing the likelihood of bias. Pointestimates and their confidence intervals were reported usinga random effects model. Meta-regression was performedfor pain and function using duration of pain, imaging tech-nique and trial quality as independent variables.

Peter G. Whang, MD, Department of Orthopaedics andRehabilitation, Yale University School of Medicine, NewHaven, CT, USAAlpesh A. Patel, MD, Departments of Orthopaedic Surgeryand Neurosurgery, University of Utah, Salt Lake City, UT,USAAlexander R. Vaccaro, MD, PhD, Srinivas K. Prasad,MD, Departments of Orthopaedic Surgery and NeurologicalSurgery, Thomas Jefferson University and The RothmanInstitute, Philadelphia, PA, USAPeter D. Angevine, MD, MPH, Department of NeurologicalSurgery, Columbia University College of Physicians andSurgeons, New York, NY, USAKishore Mulpari, MBBS, MHSc, British ColumbiaChildrens Hospital, Vancouver, BC, CanadaKen C. Thomas, MD, MHSc, Division of Spine,Department of Orthopaedics, University of British Columbia,Vancouver, BC, CanadaCharles G. Fisher, MD, Departments of OrthopaedicSurgery and Neurological Surgery, Thomas JeffersonUniversity and The Rothman Institute, Philadelphia, PACorresponding Author: Alexander R. Vaccaro, Departmentof Orthopaedic Surgery and the Rothman Institute, ThomasJefferson University, 925 Chestnut Street, 5th Floor,Phildelphia, PA 19107, USA. Phone: (215) 955-3458;E-mail: [email protected]

Spine

E178 SPINE � Volume 35 � Number 6 � 2010

CLINICAL INTERPRETATION

Six randomized controlled trials capturing 1804 pa-tients were included; 5 of the 6 studies were evaluated ashigher quality. The results of the meta-analysis demon-strate that immediate spinal imaging did not affect pri-mary clinical outcomes at either short- or long-term du-rations. The authors’ state that, based upon these results,“clinicians should refrain from routine, immediate lum-bar imaging in patients with low-back pain and withoutfeatures of a serious underlying condition.” While thismeta-analysis demonstrates the potential shortcomingsof routine lumbar imaging in back pain, the limitationsof the study make generalization of the results somewhatguarded.

The meta-analysis performed by Chou et al. is, like allstudies of similar methodology, inherently limited by thequality of the primary sources. As the authors note, ex-clusion criteria of the included trials varied with littleconsistency based upon age, patient history, patientsymptoms, or exam findings. Most notably, the presenceof neurological symptoms (e.g. leg pain, weakness,numbness, etc.) was not consistently addressed across allstudies. Additionally, the type of lumbar imaging (plainradiographs, CT, MRI) investigated was not consistentwith four trials utilizing plain radiographs and two usingCT and/or MRI. Lastly, as noted by the authors, the

reported outcomes provide only a limited perspective onthe value of imaging. The diagnostic value of these stud-ies as well as their effects on medical decision-making,changes in treatment, and patient satisfaction are notadequately addressed. The authors also state, withoutreferenced evidence, that profit motive is a driving factorbehind the use of lumbar spinal imaging suggests a po-tential for pre-existing bias against spinal imaging. Giventhe heterogeneity in patient populations, inclusion/exclusion criteria, investigational treatment (imagingmodality), and limitations in measured outcomes, gener-alization of the results is limited.

RECOMMENDATION ON IMPACT TO CLINICAL PRACTICE

The fundamental question of whether “immediateroutine lumbar spine imaging is more effective than usualcare without imaging in patients with low back pain andno suggestion of red flags” is important not only from aclinical but societal perspective. The exhaustive numberof factors that influence the answer to this question makeresearch in this area extremely difficult. In considerationof this study, which probably represents the best avail-able evidence to date, and clinical experience aroundcosts, burdens and risks we would make a weak recom-mendation to not carry out immediate routine lumbarspine imaging in patients with low back pain and no “redflags”.

Surgical compared with nonoperativetreatment for lumbar degenerativespondylolisthesis. Four-year results inthe Spine Patient Outcomes ResearchTrial (SPORT) randomized andobservational cohorts.Weinstein JN, Lurie JD, Tosteson TD, et al. J Bone JointSurg Am 2009;91:1295–304.

STUDY SUMMARY

The SPORT studies have provided high quality, pro-spective randomized and cohort data on the treat-

ment of various spinal conditions. In 2007, Weinstein etal. described the short-term (min 2 year) clinical out-comes among a group of patients with degenerativespondylolisthesis.iv The study design allowed for the892 eligible patients to consent for randomizaton (304patients) or observational (303 patients) cohort of pa-tients undergoing either surgical versus nonsurgicaltreatment. 285 (32%) patients declined to participate,thus bringing the total to 558(66%) who could not berandomized.

The high rate of crossover among randomized pa-tients, nearly 40%, compromised the intent-to-treatanalysis. The as-treated analysis of both cohorts demon-strated statistically significant improvements in the oper-ative versus the non-operative group at 2-year follow-up.Questions about the long-term efficacy of surgical treat-ment remained as deterioration of surgical results hasbeen reported previously.v,vi Late complications fromsurgical treatment as well as secondary operations maynot be captured within the short term and may negativelyinfluence late results. Longer-term results may allow foridentification of patients who would be optimally treatedwith non-surgical management. To address this, Wein-stein et al. reported on the same group of patients ini-tially described in 2007.vii The authors attempted toidentify any differences in pain and functional outcomein patients with degenerative spondylolisthesis 4 yearsafter they were either treated nonoperatively or withsurgery.


The authors have performed a methologically soundrandomized controlled trial. Setting, patient population,interventions and outcomes were stated very clearly. The

Evidence-Based Recommendations for Spine Surgery

SPINE � Volume 35 � Number 6 � 2010 E179

Spine Patient Outcomes Research Trial (SPORT) wasconducted at 13 medical centers with multidisciplinaryspine practices in 11 states in United States strengtheninggeneralizability. The SPORT included both a random-ized cohort and a concurrent observational cohort ofpatients who declined randomization ,reflecting patientpreference. All patients who had neurogenic claudicationor radicular leg pain with associated neurological signs,spinal stenosis seen on cross-sectional imaging, degener-ative spondylolisthesis seen on standing lateral radio-graphs, symptoms that had persisted for at least twelveweeks, and physician confirmation that they were a sur-gical candidate were eligible for inclusion. Patients withadjacent levels of stenosis were eligible but those withspondylolysis and isthmic spondylolisthesis were not.

The protocol surgery consisted of a standard posteriordecompressive laminectomy with or without bilateralsingle level fusion (autogenous iliac crest bone-graftingwith or without posterior pedicle screw instrumenta-tion). The non-operative protocol was “usual recom-mended care,” which at least includes active physicaltherapy, education and counseling with instructions re-garding home exercise, and non-steroidal anti-inflammatory drugs if the patient could tolerate them.

The primary outcomes were the Short Form-36 (SF36),bodily pain and physical function scores, and the AmericanAcademy of Orthopaedic Surgeons MODEMS (Musculo-skeletal Outcomes Data Evaluation and Management Sys-tem) version of the Oswestry Disability Index measured at6 weeks, 3 months, 6 months, and yearly up to 4 years.

Although the details of a power analysis and samplesize calculation were not stated in this publication, it wasdone in their previous publication. One of the problemswith any surgical trial is patient recruitment. The authorsgave a good description of patient selection, randomiza-tion, crossover and follow-up. Only 34% of eligible pa-tients agreed to randomization with the characteristics ofthe remaining prospective cohort and declining patientscarefully analyzed and compared.

The authors should be commended for this essentialmethodological principle. As this was an effectivenesstrial, they performed an intention to treat analysis whichbiases towards the null hypothesis. They also performedan “as treated analysis” (sensitivity analysis) becausethere was crossover of 54% and 33% from the random-ized and observational cohorts, respectively reflectingthe impact of patient preference on surgical trials.


In the as treated analysis, combining the randomizedand observational cohorts of patients with spinal steno-sis secondary to degenerative spondylolisthesis, thosemanaged surgically were found to have significantlygreater improvement in scores for pain, function, satis-faction, and self-rated progress over 4 years comparedwith patients treated nonoperatively. The authors re-ported sustained improvement in outcomes among pa-tients treated surgically in an as-treated analysis at both 3and 4 year follow-up.

The high rate of crossover, which increased up to54% in the non-operative randomized group, con-founded the results with no significant differences notedin the intent-to-treat analysis. The as-treated analysiscombining the randomized and observational cohortsthat adjusted for potential confounders demonstratedthat the clinically relevant advantages of surgery that hadbeen previously reported at 2 years were maintained at 4years, with treatment effects of 15.3 (95% confidenceinterval, 11 to 19.7) for bodily pain, 18.9 (95% confi-dence interval, 14.8 to 23) for physical function, and214.3 (95% confidence interval, 217.5 to 211.1) for theOswestry Disability Index. The authors reported a re-operation rate of 15% at 4 years among all surgicallytreated patients with a re-stenosis rate of 5%.

The study, like its predecessor, is most significantlylimited by low recruitment and nonadherence to the ran-domized treatment arm with crossover reported to be46% in the surgical group and 54% in the non-surgicalgroup at 4 years. The study is somewhat limited by theheterogeneity of the treatments provided with no stan-dardization of operative or non-operative care; howeverthe tradeoff of providing broader generalization winsout. Despite these limitations, this study provides strongintermediate-term (3 to 4 year) evidence that surgicalmanagement for degenerative spondylolisthesis providessuperior results to non-operative care and results consis-tent with previous studies.

RECOMMENDATION TO IMPACT ON CLINICAL PRACTICE

A weak recommendation can be made for the surgicaltreatment of degenerative spondylolisthesis in the settingof high quality literature such that a majority of surgeonsand patients would choose surgery because of superiorresults but some would not based on personal prefer-ences or clinical circumstances.



Posterior short-segment fixation withor without fusion for thoracolumbarburst fractures. A five to seven-yearprospective randomized study.Dai LY, Jiang LS, Jiang SD. J Bone Joint Surg Am 2009May;91(5):1033– 41.

STUDY SUMMARY

Thoracolumbar burst fractures represent 10–20% ofall spinal fractures and are highly variable in their

presentation. The appropriate management of these frac-tures is controversial and ranges from non-operativemanagement to combined anterior-posterior proceduresand is typically individualized according to a number offactors including fracture morphology, neurology, sur-geon and patient preference, coincident injuries andmore. Numerous thoracolumbar injury classificationsystems have been developed to systematize this heterog-enous group of fractures in an effort to facilitate commu-nication and develop management recommendations.Sethi et al.viii recently reviewed the evolution of thora-columbar injury classification systems from inceptionixthrough the most recent iteration, TLISS/TLICS.x Thisevolution in classification systems has been fueled byprogressive understanding of the importance of biome-chanical stability, injury mechanism and neurologicalcondition. Despite this, there remains significant vari-ability in management recommendations for these pro-tean injuries.

One of the principal determinations made in the eval-uation of thoracolumbar burst fractures is whether op-erative intervention is warranted at all. While there is awealth of literature exploring this question, this remainsa central controversy within thoracolumbar injury man-agement models. Although Thomas et al. embarked on asystematic review of outcomes in non-operative vs. op-erative management of patients with thoracolumbarfractures who are neurologically intact, they were unableto abstract meaningful conclusions, at least partially be-cause of heterogeneous injury classification and manage-ment criteria.xi The complex landscape of thoracolum-bar injury management is ripe with opportunity formeaningful clinical research, but will depend heavily onthe adoption of a more universal classification system.

Within this broad landscape, Dai et al. sought to ad-dress the important – if narrow – relevance of fusion inthe setting of posterior short-segment fixation for thora-columbar burst fractures Denis type-B between T11 andL2.xii The authors invoked the Denis classification sys-temxiii and the Load Sharing Classification system pro-posed by McCormack in 1994.xiv This investigationrepresents a prospective, randomized study comparingshort-segment fixation with or without autologous ICBG

in patients in burst fractures with load-sharing scores�6. Addressing available evidence suggesting that fusionbetter preserves kyphosis correction, and emergingshort-term evidence that fusion may not be essential,?,??this study presents valuable evidence examining the im-portance of fusion within this narrow clinical context.


The authors identified 91 possible patients for enroll-ment in this study from their Level 1 Trauma Center overa two-year period. Nine patients were excluded due toco-morbid conditions. Nine patients refused surgery.This left 73 patients between the ages of 18 and 60,fifty-six of which were male. Inclusion criteria stipulatedthat only burst fractures between T11 and L2 with iso-lated fracture of the superior endplate and a load sharingscore of less than or equal to 6 be included.

Simple randomization was used to place 36 and 37patients in the non-fusion and fusion groups respec-tively. The technique of posterior short segment fixationwas described. Surgical treatment was similar in bothgroups with the exception of the harvest of autogenousiliac crest bone graft and its application to the postero-lateral aspect of the spine in the fusion group. Twenty-five patients with neurologic deficits were equally di-vided between treatment groups. Only one patientunderwent decompression in the form of a hemilaminec-tomy.

The postoperative management was the same be-tween groups. Patients were kept in bed for 3 days fol-lowing surgery. A brace was not prescribed. Follow-upwas at 1, 3, 6, 12 months and then annually thereafter toa minimum of 5 years. Both functional and radiographicoutcomes were evaluated by independent observers. Fu-sion status was evaluated using plain x-rays according toa pre-defined set of criteria. When fusion was in questioncomputerized tomography was assessed.

Simple parametric and non-parametric statistics pro-vide the basis for hypothesis testing. A priori power cal-culations are described for the primary outcomes loss ofkyphosis correction and visual analog scale for backpain; although the origin of 3 degrees as a clinically im-portant loss of kyphosis correction was not provided andthis parameter is probably not valid or reliable. Otheroutcomes assessed in this study included neurologic im-provement (as measured by Frankel grade and ASIA mo-tor score) as well as visual analog scale for pain over thebone graft site, SF-36 quality of life measure and fusionstatus.

With respect the study’s primary outcome measure,no significant differences were found in the loss of ky-photic correction or the VAS for back pain, with mini-mum 5 year follow-up. Although no significant differ-ence was identified in long-term back pain VAS, 25 of 37



patients in the fusion group continued to report non-trivial donor-site pain even at last follow-up. Moreover,the authors report significantly more operative time andblood loss in the fusion group. No significant differenceswere reported in the peri and early post-operative coursewith respect to surgical timing, return to weight-bearingstatus or complication rates. No significant differencesare reported in functional outcome between the twotreatment groups. Notably, the authors report that nohardware malfunction was identified and that no hard-ware revision or removal was warranted throughout thestudy period. Finally, the authors report that all 37 pa-tients in the fusion group had solid fusion as measured byX-Ray, with CT use as necessary. Based on their findings,they concluded that the addition of a fusion to shortsegment posterior fixation for non-comminuted thoraco-lumbar burst fractures does not confer added patientbenefit, as measured by kyphotic deformity or VAS forback pain at minimum 5 year follow-up.


The authors should be congratulated for completingthis randomized controlled trial in which all consecutivepatients were accounted for at enrollment and in whichthey had 100% follow-up at each time point over 5 to 7years. A power calculation was performed which provedto be imperative given the null result.

While this study was well designed and executed, Daiet al. do not mention blinding of the clinical assessment.The radiographic analysis was performed by indepen-dent evaluators but not blinded to treatment group. In-deed, it would be interesting for the reader to know howmany patients in the non-fusion group fused their frac-tures, radiographically.

The generalizability of this study’s findings to all tho-racolumbar bursts is somewhat limited. Patients withDenis type B burst fractures (superior endplate fractures)with a load sharing score of 6 or less were included. Themore comminuted thoracolumbar burst fractures wereexcluded as the load sharing classification would suggestthey are better treated anteriorly but this is controversial.Neurological function was not a critical consideration inthis study, with only one patient in each group gettingany neurological decompression (unilateral laminec-tomy in both). As the authors acknowledge, many ofthese fractures may have been managed non-operativelyaltogether but this is not the question being addressed inthis study. For many clinicians, the presence of a neuro-logical deficit with a thoracolumbar burst fracture sug-gests the need for a possible anterior or posterolateraldecompression. The surgical construct under scrutinywas not studied in the face of decompression which lim-its the generalizability of this investigation to this sce-nario.


The term “thoracolumbar injury” remains a widelyvariable entity with equally variable management op-tions. This investigation introduces Level I evidence withrelatively few limitations for a narrow segment withinthis universe, but raises interesting questions about therelevance of posterior lateral fusion in the managementof thoracolumbar burst fractures. While this study waswell designed and executed, the generalizability of theirfindings beyond Denis B type burst fractures is limited.Thus, we propose a weak recommendation not to in-clude posterolateral fusion for the operative treatment ofDenis Type B Thoracolumbar burst fractures.

Percutaneous VertebroplastyCompared to Conservative Treatmentin Patients With Painful Acute orSubacute Osteoporotic VertebralFractures: Three-Months Follow-up ina Clinical Randomized Study.Rousing R, Andersen MO, Jespersen SM, et al. Spine2009;34:1349 –54.

STUDY SUMMARY

Osteoporotic vertebral body compression fractures(VCFs) are routinely treated with vertebral body

augmentation techniques such as percutaneous vertebro-plasy (PVP) and kyphoplasty. While these interventions

often confer significant clinical benefit, there is a growingbody of literature seeking to establish the comparativeeffectiveness of vertebral augmentation within thebroader spectrum of VCF management, most notablyusing medical management as the benchmark treatment.The VERTOS study represents the first prospective ran-domized controlled trial comparing PVP to medical man-age-ment.xvii In a study population of 34 patients whofailed 6-weeks to 6-months of conservative treatment,Voormolen et al. revealed that PVP was associated withsignificantly greater pain reduction, decreased analgesicuse and improvement in disability at the 1-day and2week time points. This study was limited in that cross-over was permitted at 2 weeks and 14 of 16 patientsrandomized to medical management elected to have PVPperformed after 2 weeks; this impaired any further anal-ysis and the study was prematurely aborted accordingly.



There have been 3 Level II studies comparing PVPwith conservative management, all of which show ashort-term benefit with PVP that diminishes within 12months.xviii,xix,xx Rousing et al. have conducted a pro-spective, randomized, controlled trial comparing PVPwith conservative management for the treatment of acute(�2wks) and subacute (2–8wks) osteoporotic compres-sion fractures.xxi This represents only the second Level Istudy examining PVP versus conservative treatment.


The study was performed at a university hospital inOdense, Denmark from January 2001 to January 2008.50 subjects were included in the study, 26 were allocatedto PVP group and 24 to conservative treatment group.The criteria for inclusion were intractable pain due toeither acute (40 patients) or subacute (10 patients) osteo-porotic fractures, inability of the patient to perform ac-tivities of daily living due to the pain, and sufficient cog-nitive function to complete the study. Simplerandomization with envelopes was used. The authors didnot specify blocks or stratification by acute/subacute orthe number of eligible patients (rejection log).

PVP was performed in the operating theatre and un-der local anaesthetics by orthopaedic surgeons specializ-ing in spine surgery. Bone cement (PMMA) was injectedunder continuous fluoroscopy. Both groups were offeredpain medication and physiotherapy if necessary until dis-charge. In addition, the patients in the conservativegroup were offered brace treatment.

A visual analogue scale (VAS) measuring self-reportedpain was the primary outcome measure on which thesample size calculation was done. Secondary outcomemeasures for physical and mental outcome included theSF 36, Dallas Pain Questionnaire, EuroQol (EQ5D), anda modified mini-mental state examination (MMSE). Ra-diographic data was collected in an effort to characterizedevelopment of adjacent level fractures although thisstudy was not powered to isolate this outcome.

The reduction in pain from the initial visit to the3month follow-up was comparable in the 2 groups (p �0.33) from approximately 8.0 to 2.0 visual analoguescale. The intragroup difference was significant (P �0.00). While pain relief in the PVP group most oftenoccurred within 12 to 24 hours after the procedure (P �0.00), the other secondary parameters appeared similarbetween the two cohorts. They observed 2 adjacent frac-tures in the PVP group and none in the conservativegroup.

This study implies that the majority of patients withacute or subacute painful osteoporotic compression frac-tures in the spine will recover after a few months ofconservative treatment and that further research is nec-

essary to characterize the risk of adjacent fractures. Nomajor adverse events were observed in either group


The authors should be congratulated for completingthis randomized controlled trial in which they concludedthere is no difference in pain in patients who receivedPVP or conservative treatment at a 3 month follow-up.This is a superiority trial so the power and sample sizecalculations were performed to find a difference. We can-not assume equivalence or non-inferiority because wewould be committing a type 2 error; a large sample size isneeded to establish equivalence or non-inferiority.

A few structural weaknesses in this study are concern-ing. The study had a long enrollment period so the au-thors need to explain why it took them 7 years to recruit50 subjects. They also did not list the number of eligiblesubjects approached for recruitment. They should havemaintained and reported a rejection log to track eligiblepatients that did not consent to see if they differed fromthe participating patients– this could have been part ofFig 1. There were only 36 of 50 patients who completedthe baseline VAS which is the primary outcome, i.e. al-most 30% of included subjects did not record baselinescores. Furthermore, there are too many p-values re-ported; as an option they may have only subjected theirprimary outcome measure to hypothesis testing or theycould have used a corrective factor such as the Bonfer-roni correction. We recognize that this would not havechanged their null result.

The main concern we have relates to their first keypoint: “PVP is comparable to conservative treatment fortreating painful osteoporotic vertebral compression frac-tures” whereas it should read “immediate PVP is notsuperior to conservative treatment in the treatment ofacute or subacute osteoporotic vertebral compressionfractures at 3 month follow up”. The initial statement isnot justified by their study. The most notable finding ofthis investigation is a statistically significant reduction inVAS pain score within 12–24 hrs of intervention in thepopulation undergoing PVP, a potentially clinically rel-evant finding. Although this symptomatic benefit waspreserved, the conservative treatment group also enjoyedgradual improvement so that by 3-months there was nostatistically significant difference between treatmentgroups. Moreover, there was no significant difference inmental or physical outcome at 3 months. This paper hasdefined the natural history of osteoporotic fractures inthat most patients feel better fairly quickly. At this point,many practitioners use conservative care as first-linetreatment for osteoporotic compression fractures, re-serving PVP for those patients who do not respond toconservative care. This study has not addressed the com-



parative effectiveness of PVP in those who fail this initialperiod which is arguably a more important question.

As patient involvement in clinical decision-makinggrows, it is imperative that physicians present a balancedperspective, drawing on experience and best availableclinical evidence. The findings of Rousing et al. supportthe early use of PVP in the setting of acute or sub-acuteosteoporotic VCFs if early pain control is a central pri-ority or if conservative measures are not a viable option.Both groups enjoyed considerable improvement at 3months without any significant difference in measuredpain or functional outcome.


This is moderate quality or level 2 evidence that whenput in the context of previous studies does not seem tohave an impact on clinical practice. When taking this andprevious evidence into consideration along with patientpreference and clinical expertise there is a weak recom-mendation not to use PVP for acute VCFs as the resultsof conservative care at 3 months is favorable. This inves-tigation suggests that PVP patients get better fasterwhich could provide certain clinical advantages butstudy limitations prevent any recommendations regard-ing this outcome.

RhBMP-2 versus iliac crest bone graftfor lumbar spine fusion in patients over60 years of age. A cost-utility study.Carreon LY, Glassman SD, Djurasovic M, et al. Spine2009;34:238 – 43.

STUDY SUMMARY

Carreon et al. report the results of a cost-utility anal-ysis (CUA) of rhBMP-2 compared to autologous il-

iac crest bone graft (ICBG) in patients over 60 years oldbased on a previously-published randomized clinicaltrial (RCT).xxii Fifty patients were randomized to re-ceive rhBMP-2 and 52 were treated with ICBG. All sub-jects underwent single or multilevel instrumented lumbararthrodesis without interbody fusion. Patients were eval-uated at regular intervals using multiple outcome mea-sures including SF-6D utility scores and complicationswere recorded contemporaneously. Direct costs werecalculated by considering the reimbursement amountsfor all inpatient and outpatient events during the 2-yearstudy period; however, indirect and non-health relatedcosts were not considered. According to their study data,the authors constructed a decision-analysis tree withnodes for clinical improvement, complications, and re-operations.

The authors report that the mean total 2-year cost forcare excluding complications and additional spinal treat-ments was $34,235 for the ICBG group and $36,530 forthe rhBMP-2 cohort. Other mean costs for the entirestudy population were $10,888 for a major complica-tion, $46,852 for a revision surgery to manage apseudarthrosis, and $5892 for additional spinal treat-ments. Rollback of the decision analysis tree yielded atotal cost of rhBMP-2 of $39,967 with a 0.11 mean im-provement in SF-6D scores and a total cost of $42,286for ICBG with a mean improvement of 0.10. Based onthis analysis, the authors calculated that the higher initial

costs of rhBMP-2 were more than made up for by adecrease in the need for additional treat ments by theindividuals in this cohort and they concluded that the useof rhBMP-2 was both more effective and less expensivethan ICBG.

While a number of studies have demonstrated thatrecombinant human bone morphogenetic proteins suchas rhBMP-2 may represent an effective method for pro-moting spinal fusion, one major consideration that mayultimately determine whether these materials gain wide-spread acceptance as substitutes for autogenous bone inthis era of limited health care resources is their significantexpense. For this reason, several authors have performedeconomic analyses to determine the costs associated withthese products. Ackerman et al. and Polly et al. com-pared the estimated expenditures of stand-alone anteriorlumbar interbody arthrodesis procedures using either au-tograft procured from the iliac crest or rhBMP2.xxiii,xxivThese investigations suggested that the higher charges aris-ing from rhBMP-2 may be offset by savings incurred byfewer complications and more rapid functional recovery.These results were recently corroborated by a similar eval-uation conducted in Europe.xxv Unfortunately, these pre-vious reports were based on largely theoretical models de-rived from peer-reviewed literature and expert opinion. Inan effort to overcome these limitations, Carreon et al. em-ployed actual cost data obtained from a prospective, ran-domized clinical trial in their cost-utility study quantifyingthe relative value of rhBMP-2 for instrumented posterolat-eral fusions.


Many of the key elements of a high-quality CUA areincorporated in this evaluation. The authors presenttheir decision tree with the relative probabilities associ-ated with each chance node. In addition, the costs andincremental utility for each branch are also listed. Theprobabilities, costs, and utilities were all determinedfrom a single study which serves to minimize potential



sources of error. Health-state utilities were also deter-mined using the validated SF-6D instrument. A 2-yearfollow-up period for the study is probably valid based onthe assumption that symptomatic nonunions are identi-fied within that time; however a longer follow-up wouldbe desirable.

The model seems to be a reasonable representation ofthe intervention in terms of both its possible sequelae andoutcomes. Nevertheless, one concern is that the baselineanalysis indicates an overall improved outcome for therhBMP-2 cohort (0.11) compared to the ICBG group(0.10) yet Figure 2 shows that the overall mean utility forthe rhBMP-2 group was lower at each time point. Afalloff of the mean rhBMP-2 utility at 2 years does notseem to support the claim that the incremental utilitybenefit for this group was greater compared to the ICBGcohort. This discrepancy raises concerns about the valid-ity of the model itself that may compromise the overallstudy.

The investigators in this study used patient-level reim-bursements in their model which represents the amountspaid to a health care provider by a government or privateinsurer. A more accurate reflection of the true economiccost, however, is the value of the resource utilization forproviding that service. In general, Medicare reimburse-ment rates are employed to calculate the costs for physi-cian services whereas hospital expenses are estimated byapplying the ratio of costs-to-charges (RCC) method tothe institutional reports prepared for the Centers forMedicare & Medicaid Services (CMS). Simply consider-ing reimbursement data will tend to bias the results de-pending on the insurance status of the patients. In thisstudy population, most patients were likely covered byMedicare but the reimbursement values do not excludethe hospital “markup” for these services.

Probably the most significant shortcoming of thisstudy is that it does not include an assessment of thesensitivity of the model to changes in input parameters.As noted by the authors, the patients’ costs exhibited asubstantial degree of variability; as an example, the costsof ICBG ranged from $19,988 to $139,583 and $21,955to $109,027 for rhBMP-2. In these situations, mean val-ues do not always give a comprehesive picture of themodel. Due to sampling error, the “true” cost mightactually be significantly higher or lower than the calcu-lated mean value. Additional measures of dispersionsuch as the median values, interquartile ranges, and stan-dard deviations would allow the reader to assess the va-lidity of the mean values as summary point estimates.Even with randomization of a moderately large patientcohort, inequalities between the groups may occur bychance which is even more likely to occur with relativelyrare events. For instance, the mean cost of readmissionwithin 3 months of surgery for the ICBG group was

almost twice that of the rhBMP-2 group which wasbased on a total of 7 readmissions.xxvi Because the esti-mated utility benefit of an intervention may vary signif-icantly depending on the instrument used, extensive sensitivity testing comprising both one-way (i.e. assessingthe impact of changing one variable at a time) and two-way (i.e. altering input values in pairs) analyses should bepart of every CUA.xxvii

A review of this model reveals significant sensitivity tothe probability of improvement after additional treat-ments were administered to the subset of patients in therhBMP-2 cohort who had no complications. All 7 pa-tients in this group improved in the baseline model but if2 of those individuals had not responded favorably theprobability of improvement would have been similar tothat observed for the corresponding ICBG subgroupsuch that the difference between the interventions interms of both cost and outcome would have essentiallydisappeared. This finding suggests that the results may bevery sensitive to small changes in the input parameters.Similar to subgroup analyses in clinical trials, sensitivityanalyses should be selected a priori if possible in order tominimize the likelihood of bias.

Finally, it is important to point out that this investi-gation involves the use of rhBMP-2 for posterolaterallumbar fusions which is an indication that is not cur-rently approved by the Food and Drug Administration.As with all industry-supported research it is also criticalfor the reader to recognize that one or more of the au-thors acknowledged a financial relationship that has thepotential to create a conflict of interest.


Cost-utility analysis is widely considered the goldstandard design for comparative effectiveness studies be-cause it allows for the comparison of the costs and effec-tiveness of different interventions for a single or multipleconditions. Because of their allure, high regard, and po-tential for significant impact from a clinical and reim-bursement perspective it is critical that the results ofthese types of studies are not taken out of context and areappropriately reviewed . In this case, it would be inap-propriate to conclude that spinal fusion is not cost-effective because its cost per quality adjusted life year(QALY) is roughly $400,000 based on the findings ofthis investigation. The relevant metric is the incrementalcost-effectiveness ratio (ICER) that compares the costs oftwo different treatments.xxviii


Although there are several significant methodologicalshortcomings in this study, this study should be viewedas an important first step in an iterative and evolving



process which will hopefully be refined in the future bythe original authors or by other groups. Sensitivity anal-yses should be performed on the existing data to identifykey areas requiring additional examination in an attemptto improve the estimates of cost, outcome, and probabil-ity which would be expected to yield more valid modelsand a better understanding of the rel-rhBMP-2 as a re-

placement for iliac crest bone graft for ative cost-effectiveness of various surgical options for fusion appli-cations; thus, we do not currently recomthis patientpopulation. The findings of this investiga-mend anychanges to clinical practice based on the evition do notjustify the widespread incorporation of dence providedby this study.

Does incorrect level needle localizationduring anterior cervical discectomy andfusion lead to accelerated discdegeneration?Nassr A, Lee JY, Bashir RS, et al. Spine 2009;34:189 –92.

STUDY SUMMARY

In a recent retrospective cohort study, Nassr et al.attempted to elucidate the risk of accelerated disk

degeneration associated with the improper placementof a localizing needle during anterior cervical discec-tomy and fusion with plating (ACDFP) procedures byreviewing the preoperative, intraoperative, and post-operative cervical radiographs of 247 individuals whowere managed with either 1- or 2-level operations.xxix160 patients were excluded because of non-degenerativepathology, a history of previous cervical spine interven-tions, inadequate radiographic studies, or insufficientfollow-up; of the remaining 87 subjects, 15 (17%) werenoted to have had the 22-gauge localizing needle inad-vertently placed into the segment rostral to the intendedfusion construct. Follow-up x-rays obtained at a mean of21.8 to 24.6 months after surgery demonstrated that agreater proportion of patients with incorrect needleplacement (9/15) exhibited radiographic evidence of pro-gressive degeneration according to a 3-tier scale com-pared to those whose needles had been inserted into anappropriate disk space (23/72) which corresponded toan approximately 3-fold increased risk. There were nostatistically significant differences between the incidencesof progressive degeneration exhibited by subjects under-going either 1- or 2-level procedures (28% and 43%,respectively).

The incidence and etiology of adjacent segment de-generation continue to be a matter of some debate, withmany studies suggesting that it is a consequence of in-creased stresses arising from a contiguous arthrodesiswhereas others have attributed this pathology to the nat-ural history of cervical spondylosis. In a landmark paper,Hilibrand et al. retrospectively examined a large series ofanterior cervical fusions and calculated the incidence ofsymptomatic adjacent segment disease to be 2.9% annu-

ally.xxx Many surgeons elect to insert a needle within anintervertebral disk as a means of confirming the correctoperative level during these procedures; however, withthis method it is inevitable that the incorrect disk spacewill be marked incorrectly in some cases. Puncturing theannulus with a needle has been shown to trigger a num-ber of biomechanical and biochemical alterations thatcollectively bring about progressive disk degeneration inanimals. Elliott et al. examined the effect of needle diam-eter on this phenomena in an animal model and estab-lished that a needle:disk height ratio of less than 0.4 isunlikely to give rise to significant degeneration.


Although this is a retrospective review, the authorsincluded a few design elements to minimize the potentialbiases to which these types of studies are susceptible.First of all, the raters of the pre- and postoperative ra-diographs were blinded to the patient’s group (i.e. incor-rect vs. correct needle placement). Furthermore, the au-thors calculated k values which represent a measure ofinterobserver agreement. Together these two factorsserve to increase the validity of these findings.

The authors mention a few potentially confoundingoperative factors that may have also precipitated adja-cent level degeneration such as the additional muscle andsoft tissue dissection that occurred as part of the expo-sure of an adjacent disk space level. The authors alsomeasured the distance between the upper edge of theanterior plate and the rostral disk space because anyinstrumentation in close proximity to uninvolved spinalsegments has been shown to predispose these levels toprogressive degeneration.

At first glance, the control and experimental groups inthis study appear to be comparable; however, there werenon-statistically significant differences between the meanages, follow-up times and the proportion of each groupthat was male and female between these populations. It isalso important to consider what other factors might haveled to incorrect localization that also may have influ-enced the risk of developing degenerative disk disease.One such variable is the distribution of operative levelstreated in each group. For example, there may have beena greater proportion of subjects with improper needle



placement who underwent surgery for C6–7 pathologyin which case the puncture injury would have occurred atthe C5–6 level which is known to be one of the mostcommon segments to be affected by cervical spondylo-sis.2 Similarly, body habitus is another potentially con-founding characteristic that was not addressed by theauthors.

The statistical analysis performed in this study alsoraises several questions. The sizes of the groups are rela-tively small which may affect the overall validity of theresults. Although the odds ratio was reported to be 3.2,the 95 percent confidence interval is quite wide (1.02 to10.05). It is impossible to determine from a single, smallstudy whether the point estimate (i.e. 3.2) is the mostaccurate reflection of the true underlying value. There-fore, the authors’ suggestion that “the results do reveal asignificant increase in the radiographic evidence of diskdegeneration with an odds ratio of 3.2 resulting fromincorrect level needle localization” appears to overstatethe strength of the evidence.

The finding that there were no differences in the like-lihood of adjacent level disk degeneration between 1-and 2-level constructs is also limited by the study size. Inthis series 28% and 43% of 1- and 2-level fusion pa-tients, respectively, demonstrated progression of degen-eration of at least 1 grade which was associated with aP-value of 0.143. Unfortunately post-hoc power analysisis not possible because the number of individuals in eachgroup is not provided but it is likely that a larger samplesize would give rise to a statistically significant P value ifthese reported proportions were maintained. Given theinsufficient power of this investigation, the authors’ con-clusions must be considered accordingly.


The results of this investigation are quite striking inthat incorrect needle localization had occurred in 17% ofthese cases. More importantly, these individuals werefound to be approximately 3 times more likely to developradiographic evidence of adjacent segment degenerationthan those whose needles were introduced into an appro-priate disk space. The authors attributed this finding toeither a puncture injury to the annular tissue or unnec-essary surgical dissection.

The small cohorts and retrospective design of thestudy significantly limit its ability to establish a causallink between improper needle placement and the radio-graphic degeneration. There is significant potential forconfounding and bias in this study that may account forthe observed difference in degeneration between thegroups. The retrospective design limitations could havebeen minimized with an appropriate sample size as it isunlikely this research question will be addressed througha prospective design.

While it may still be prudent to develop alternativetechniques for marking the cervical spine intraopera-tively such as fixing a retractor pin in the vertebral bodieswhich may serve to minimize the development of second-ary spondylotic changes at the levels that are not in-tended to be included into the fusion construct, the in-sufficient power of this investigation provides only lowquality evidence to make any definitive conclusionsabout the effects of inadvertent disk injury. Based uponthe results of this study along with the animal data andthe insignificant clinical implications of marking the ver-tebral body, we offer a weak recommendation not toplace a needle in the disk space to confirm levels duringanterior cervical spine surgery.

REFERENCES1. Boden, SD, Davis DO, Dina TS, et al. Abnormal magnetic-reso-

nance scans of the lumbar spine in asymptomatic subjects. A pro-spective investigation. J Bone Joint Surg Am 1990;72:403–8.

2. Kerry S, Hiltons S, Dundas D, et al. Radiography for low backpain: a randomised controlled trial and observational study inprimary care. Br J Gen Pract 2002;52:469–74.

3. Chou R, Fu R, Carrino JA, et al. Imaging strategies for low-backpain: systematic review and meta-analysis. Lancet 2009;373:463–72.

4. Weinstein JN, Lurie JD, Tosteson TD, et al. Surgical versus non-surgical treatment for lumbar degenerative spondylolisthesis.N Engl J Med 2007;356:2257–70.

5. Atlas SJ, Keller RB, Robson D, et al. Surgical and nonsurgicalmanagement of lumbar spinal stenosis: four-year outcomes fromthe maine lumbar spine study. Spine 2000;25:556–62.

6. Fischgrund JS, Mackay M, Herkowitz HN, et al. 1997 VolvoAward winner in clinical studies. Degenerative lumbar spon-dylolisthesis with spinal stenosis: a prospective, randomized study

comparing decompressive laminectomy and arthrodesis with andwithout spinal fusion. Spine 1997;22:2807–12.

7. Weinstein JN, Lurie JD, Tosteson TD, et al. Surgical comparedwith nonoperative treatment for lumbar degenerative spondylolis-thesis. four-year results in the Spine Patient Outcomes ResearchTrial (SPORT) randomized and observational cohorts. J BoneJoint Surg Am 2009;91:1295–304.

8. Sethi MK, Schoenfeld AJ, Bono CM, et al. The evolution of tho-racolumbar injury classification systems. Spine J 2009;9:780–8.

9. Boehler L. Die Techniek der Knochenbruchbehandlung imGrieden und im Kriege. Vienna, Austria: Verlag von WilheimMaudrich, 1930. 9–11.

10. Vaccaro AR, Lehman RA Jr, Hurlbert RJ, et al. A new classifica-tion of thoracolumbar injuries: the importance of injury mor-phology, the integrity of the posterior ligamentous complex, andneurologic status. Spine 2005;30:2325–33.

11. Thomas KC, Bailey CS, Dvorak MF, et al. Comparison of oper-ative and nonoperative treatment for thoracolumbar burst frac-tures in patients without neurological deficit: a systematic review.J Neurosurg Spine 2006;4: 351–8



12. Dai LY, Jiang LS, Jiang SD. Posterior short-segment fixation withor without fusion for thoracolumbar burst fractures. A five toseven-year prospective randomized study. J Bone Joint Surg Am2009; 91:1033–41.

13. McCormack T, Karaikovic E, Gaines RW. The load sharing clas-sification of spine fractures. Spine 1994;19:1741–4.

14. Denis F. The three column spine and its significance in the clas-sification of acute thoracolumbar spinal injuries. Spine 1983;8:817 -31.

15. Sanderson PL, Fraser RD, Hall DJ, et al. Short segment fixation ofthoracolumbar burst fractures without fusion. Eur Spine J 1999;8:495–500.

16. Wang ST, Ma HL, Lu CL, et al. Is fusion necessary for surgicallytreated burst fractures of the thoracolumbar and lumbar spine? Aprospective,randomized study. Spine 2006;31:2646–52.

17. Voormolen MH, Mali WP, Lohle PN, et al. Percutaneous verte-broplasty compared with optimal pain medication treatment:short-term clinical outcome of patients with subacute or chronicpainful osteoporotic vertebral compression fractures. The VER-TOS study. AJNR Am J Neuroradiol 2007;28:555–60.

18. Diamond TH, Champion B, Clark WA. Management of acuteosteoporotic vertebral fractures: a non-randomised trial compar-ing percutaneous vertebroplasty with conservative therapy. Am JMed 2003;114:257–65.

19. Diamond TH, Bryant C, Browne L, et al. Clinical outcomes afteracute osteoporotic vertebral fractures: a 2-year non-randomisedtrial comparing percutaneous vertebroplasty with conservativetherapy. Med J Aust 2006;184:113–7.

20. Alvarez L, Alcaraz M, Perez-Higueras A, et al. Percutaneous ver-tebroplasty: functional improvements in patients with osteopo-rotic compression fractures. Spine 2006;31: 1113–8.

21. Rousing R, Andersen MO, Jespersen SM, et al. Percutaneousvertebroplasty compared to conservative treatment in patientswith painful acute or subacute osteoporotic vertebral fractures.Spine 2009;34:1349–54.

22. Carreon LY, Glassman SD, Djurasovic M, et al. RhBMP-2 versusiliac crest bone graft for lumbar spine fusion in patients over 60years of age. A cost-utility study. Spine 2009;34:238–43.

23. Ackerman SJ, Mafilios MS, Polly DW Jr. Economic evaluationof bone morphogenetic protein versus autogenous iliac crest

bone graft in single-level anterior lumbar fusion. Spine 2002;27:S94–9.

24. Polly DW Jr, Ackerman SJ, Shaffrey CI, et al. A cost analysis ofbone morphogenetic protein versus autogenous iliac crest bonegraft in single-level anterior lumbar fusion. Orthopedics 2003;26:1027–37.

25. Alt V, Chhabra A, Franke J, et al. An economic analysis of usingrhBMP-2 for lumbar fusion in Germany, France and UK from asocietal perspective. Eur Spine J 2009;18:800–6.

26. Glassman SD, Carreon LY, Campbell MJ, et al. The perioperativecost of Infuse bone graft in posterolateral lumbar spine fusion.Spine J 2008;8:443–8.

27. Marra CA, Marion SA, Guh DP, et al. Not all “quality-adjustedlife years” are equal. J Clin Epidemiol 2007;60:616–24.

28. Torrance GW, Siegel JE, Luce BR: Framing and designing the costeffectiveness analysis, in Gold MR, Siegel JE, Russell LB, Wein-stein MC (eds): Cost-effectiveness in health and medicine. NewYork, Oxford University Press, 1996, pp 54–81.

29. Nassr A, Lee JY, Bashir RS, et al. Does incorrect level needlelocalization during anterior cervical discectomy and fusion leadto accelerated disc degeneration? Spine 2009;34:189–92.

30. Hilibrand AS, Carlson GD, Palumbo MA, et al. Radiculopathyand myelopathy at segments adjacent to the site of a previousanterior cervical arthrodesis. J Bone Joint Surg Am 1999;81:519–28.

31. Masuda K, Aota Y, Meuhleman C, et al. A novel rabbit model ofmild, reproducible disc degeneration by an annulus need punc-ture: correlation between the degree of disc injury and radiolog-ical and histological appearances of disc degeneration. Spine2004;30:5–14.

32. Sobajima S, Kompel JF, Kim JS, et al. A slowly progressive andreproducible animal model of intervertebral disc degenerationcharacterized by MRI, x-ray, and histology. Spine 2004;30:15–24.

33. Kim SK, Yoon ST, Li J, et al. Disc degeneration in the rabbit: abiochemical and radiological comparison between four disc in-jury models. Spine 2004;30:33–7.

34. Elliott DM, Yerramalli CS, Beckstein JC, et al. The effect of rel-ative needle diameter in puncture and sham injection animalmodels of degeneration. Spine 2008;33:588–96.



Evidence-Based Recommendations for Spine Surgery

Documents