Lumbar spinal stenosis (LSS) is an anatomical impairment

Transcription

1 Annals of Internal Medicine ORIGINAL RESEARCH Versus Nonsurgical Treatment of Lumbar Spinal Stenosis A Randomized Trial Anthony Delitto,, PhD; Sara R. Piva,, PhD; Charity G. Moore, PhD, MSPH; Julie M. Fritz,, PhD; Stephen R. Wisniewski, PhD; Deborah A. Josbeno,, PhD; Mark Fye, MD; and William C. Welch, MD Background: Primary care management decisions for patients with symptomatic lumbar spinal stenosis (LSS) are challenging, and nonsurgical guidance is limited by lack of evidence. Objective: To compare surgical decompression with physical therapy () for LSS and evaluate sex differences. Design: Multisite randomized, controlled trial. (ClinicalTrials.gov: NCT ) Setting: Neurologic and orthopedic surgery departments and clinics. Participants: Surgical candidates with LSS aged 50 years or older who consented to surgery. Intervention: Surgical decompression or. Measurements: Primary outcome was physical function score on the Short Form-36 Health Survey at 2 years assessed by masked testers. Results: The study took place from November 2000 to September A total of 169 participants were randomly assigned and stratified by surgeon and sex (87 to surgery and 82 to ), with 24-month follow-up completed by 74 and 73 participants in the surgery and groups, respectively. Mean improvement in physical function for the surgery and groups was 22.4 (95% CI, 16.9 to 27.9) and 19.2 (CI, 13.6 to 24.8), respectively. Intention-to-treat analyses revealed no difference between groups (24-month difference, 0.9 [CI, 7.9 to 9.6]). Sensitivity analyses using causal-effects methods to account for the high proportion of crossovers from to surgery (57%) showed no significant differences in physical function between groups. Limitation: Without a control group, it is not possible to judge success attributable to either intervention. Conclusion: Surgical decompression yielded similar effects to a regimen among patients with LSS who were surgical candidates. Patients and health care providers should engage in shared decision-making conversations that include full disclosure of evidence involving surgical and nonsurgical treatments for LSS. Primary Funding Source: National Institutes of Health and National Institute of Arthritis and Musculoskeletal and Skin Diseases. Ann Intern Med. 2015;162: doi: /m For author affiliations, see end of text. Lumbar spinal stenosis (LSS) is an anatomical impairment characterized by narrowing of the spinal canal or nerve root foramen (1). When a person is symptomatic, LSS causes pain; weakness in the lower back, buttocks, and thighs; and claudicating pain (1). Anatomical LSS, however, is also common in older patients who are asymptomatic (2), which underscores the importance of corroborative findings between patient history and examination (3). Even after careful examination, management decisions for persons with symptomatic LSS remain a challenge that has aptly been described as a balancing act (4), which is compounded by a lack of clear, evidence-based nonsurgical treatment options (5). remains an option for patients with persistent and severe LSS symptoms that include both back and leg pain (6 8). In fact, LSS is now the most often cited cause for lumbar surgery in the United States (9). Studies comparing surgical with nonsurgical treatment of LSS have been done but remain unclear about optimal nonsurgical treatment options (7, 10 12). The surgical approach in these studies has been highly standardized. In contrast, the nonsurgical comparator groups commonly lack structure and detail and have inconsistent follow-through to ensure that the most basic evidence-based approaches are included, including activation, exercise, and discouragement of passive agents. In SPORT (Spine Patient Outcomes Research Trial) (6), the largest randomized, controlled trial (RCT) comparing surgical and nonsurgical treatment of LSS, the surgical group had a standard posterior decompressive laminectomy. In contrast, the nonsurgical group received usual care in which surgeons were encouraged to recommend active physical therapy (), education or counseling with home exercise instruction, and nonsteroidal anti-inflammatory medication as initial management strategies, but participants could receive any additional conservative treatments deemed appropriate by the surgeon. Crossover rates were 67% for the surgical group (that is, crossed over to nonsurgical intervention) and 43% in the nonsurgical group (crossed over to surgery), both of which were sufficiently high to preclude intention-to-treat (ITT) interpretations. Instead, interpretation was based on astreated analyses for virtually all of their outcomes (11, 13 16). Secondary analysis of patients in the nonsurgi- See also: Editorial comment Summary for Patients....I-32 Web-Only Supplements 2015 American College of Physicians 465

2 ORIGINAL RESEARCH Versus Nonsurgical Treatment of Lumbar Spinal Stenosis EDITORS' NOTES Context Evidence is limited about optimal nonsurgical alternatives to surgical decompression for management of patients with symptomatic lumbar spinal stenosis (LSS). Contribution Patients with LSS who were surgical candidates and who provided consent for surgery were randomly assigned to physical therapy () for 6 weeks or surgical decompression. Physical functioning, the primary outcome, was assessed after treatment and during the 2-year follow-up. Caution Half of patients in the group crossed over to receive surgery. Implication Patients with LSS who were offered an evidence-based program or surgical decompression achieved similar symptom relief and improvements in physical functioning. cal group found that only 10% were provided education or counseling and 37% received within 6 weeks. Those receiving had higher self-ratings of improvement and were less likely to cross over to surgery than those who did not receive (17). Comparisons between surgery and standardized application of a program for patients with LSS have not been done. Therefore, we wanted to compare surgical decompression with a specified nonsurgical regimen in patients considered surgical candidates for symptomatic and degenerative LSS. Our study also evaluated sex differences in outcomes after treatment of LSS. To reduce crossover in at least 1 group of the study, we randomly assigned patients after they consented to surgery. METHODS Patients were randomly assigned to 1 of 2 treatment groups: surgical decompression or. The primary outcome of physical function was evaluated through patient-reported outcomes. We followed patients both in the short term (first 10 weeks) and long term (at 6, 12, and 24 months). Targeted Patient Population We enrolled patients with a diagnosis of LSS identified by computed tomography using the criteria of Wiesel and colleagues (18) or magnetic resonance imaging using the criteria of Boden and colleagues (2). All patients were considered by a spine surgeon to be candidates for surgical decompression and had consented to surgery. Consent to surgery was a key inclusion criterion to ensure all participants were surgical candidates and to minimize crossover from the surgery group to the group. Patients also had to meet the following eligibility criteria: presence of neurogenic claudication (for example, self-reported inability to walk more than a quarter mile because of lower-extremity pain or cramping); consent to be randomly assigned to surgery or a specified clinic for twice-weekly exercise sessions; and no previous surgery for LSS at the level being considered for decompression. We excluded patients who were younger than 50 years, had signs of serious dementia, were diagnosed with severe vascular disease or had a recent history of myocardial infarction, had concomitant spondylolisthesis requiring spinal fusion (defined as >5 mm of slippage), had compression fractures at the level being considered for decompression, or were diagnosed with metastatic cancer. Recruitment and Consent Procedures Five neurosurgeons and 1 orthopedic spine surgeon from 2 medical centers participated in participant recruitment. All potential participants were examined initially by spine surgeons and judged to be surgical candidates. After patients consented to surgery, the surgeon introduced the study. If patients indicated an interest in the study, a study coordinator then fully explained the study and obtained consent to participate in the RCT, after which they would be randomly assigned to have surgery or attend. The study, conducted in western Pennsylvania, was approved by a Data and Safety Monitoring Board chartered by the National Institutes of Health and National Institute of Arthritis and Musculoskeletal and Skin Diseases and the Institutional Review Boards of University of Pittsburgh and Allegheny General Hospital (now called Allegheny Health Network). Study participants self-reported demographic and other descriptive information, including sex, age, ethnicity, past and present activity levels, present income level and employment status, and medical history. Participants also completed questionnaires about disability and other psychosocial factors. Randomization Patients were randomly assigned after the initial assessment procedures. Randomization was conducted using SAS software, version 6.2 (SAS Institute), with permuted blocks of random block sizes and stratification by sex and surgeon. Sequentially numbered and sealed envelopes with assignment allocation were prepared by the data center. A study coordinator not involved with preparing the allocation sequence opened the envelopes and assigned participants to interventions. Surgical Decompression Surgical treatment was performed similarly to that described by Rothman and Simeone (19). The protocol for surgery was identical to that used in other surgical trials for LSS (20) and included decompressive laminectomies, partial facet resection, and neuroforaminotomies performed at the levels of radiographic stenosis. 466 Annals of Internal Medicine Vol. 162 No. 7 7 April

3 Versus Nonsurgical Treatment of Lumbar Spinal Stenosis ORIGINAL RESEARCH No patients had spinal fusion. All surgical procedures were done by either fellowship-trained spine surgeons or surgeons with more than 20 years of experience dedicated to spine surgery. Hospital stay after surgery averaged 3 days. The postoperative course included a graduated ambulation program that began on postoperative day 1. Patients were encouraged to increase their level of walking activity as tolerated. No study funds were used to reimburse the expenses of surgical and postoperative care. The program emphasized lumbar flexion exercises, general conditioning exercises, and patient education. Each patient was evaluated by a physical therapist using an examination scheme described in our previous work (21). This examination scheme was designed to identify impairments in lower-extremity strength and flexibility that would be addressed during treatment. Treatment fidelity was assessed by the investigators. The following treatments were administered: instruction in lumbar flexion exercises, including posterior pelvic tilts and supine knee-to-chest and quadruped flexion exercises; general conditioning exercises, including stationary cycling or treadmill walking; lowerextremity strengthening exercises deemed appropriate for the patient after examination (for example, standing squats, seated knee extension, or supine straight-leg raises); lower-extremity flexibility exercises deemed appropriate for the patient on the basis of individual examination (for example, hamstring or hip flexor stretching); and patient education to avoid postures involving hyperextension of the lumbar spine. Full descriptions these treatments are available in Supplement 1 (available at Physical therapy was prescribed for 6 weeks, with a frequency of 2 visits per week, and was delivered by licensed physical therapists. Patients were instructed that they could cross over to surgery at any point in the trial on the basis of a shared decision-making process with the spine surgeon. No study funds were used to reimburse care. Outcome and Follow-up The primary outcome for the study was the physical function score on the Short Form-36 Health Survey (SF- 36) (22) at the 2-year end point. For treatment outcome scales, we administered the Oswestry Disability Index (ODI) (23) and the North American Spine Society outcome assessment instrument for pain, disability, neurogenic symptoms, and expectation (24). All evaluations were carried out by research assistants blinded to participants' group assignment. Patients were asked to wear T-shirts to cover surgical scars when reporting for follow-up visits. Statistical Analysis We predetermined that, with a sample size of 96 participants per group, we would have 93% power to detect an effect size of 0.5 SD between the surgery and groups for the primary outcome (2-sided level of 0.05). To evaluate sex differences in outcomes, we would have approximately 79% power to detect a difference in treatment effects of 0.8 SD between men and women split evenly by sex (48 participants per group). No adjustments to sample size were made for attrition or crossovers. Three interim analyses were planned after 48, 96, and 158 participants completed 12-month follow-up on the basis of Lan and DeMets spending functions with O Brien Fleming boundaries. Primary results are based on the ITT approach. The ITT analysis provides an estimated effect of being offered compared with being offered surgery. Function was measured on the SF-36 at baseline; 10 weeks; and 6, 12, and 24 months; therefore, we used linear mixed-effects models (25) to assess for differences in the improvement in function over time between the 2 groups (controlling for sex and surgeon). Baseline and all follow-up measurements of a patient were fitted using maximum likelihood methods with unstructured covariance matrix to account for the repeated measurements within an individual (SAS PROC MIXED with REPEATED statement) (25). The linear mixed-effects model included fixed effects for group, time, and group-by-time interaction. The same approach to the analysis was used for secondary outcomes of ODI and the North American Spine Society instrument for pain, disability, neurogenic symptoms, and expectation. A naive as-treated analysis can result in biased estimates due to the factors associated with crossovers. Therefore, we conducted complier average causal effect (CACE) (26, 27) and inverse probability weighting (IPW) (28) analyses because of the high rate of crossovers in the group. Last, we compared the rate of successful outcomes among the participants from the surgery group, crossovers from to surgery, and participants from the group who did not cross over. We defined an improvement of more than 0.5 SD at 2-year follow-up relative to baseline as a clinically meaningful effect. Subgroup analyses were only done for sex because this was prespecified as a goal for our study to assess the effect of sex on treatment effects. We only formally tested for a sex interaction for the primary outcome based on the ITT approach. We did sensitivity analyses for physical function using multiple and simple imputation based on nonignorable missingness to evaluate the effect of missing data on the primary ITT results. All tests were 2-sided, with an level of 0.05, and all analyses were done in SAS, version 9.3. Role of the Funding Source The funding source did not have a role in the design, conduct, and analysis of the study or the decision to submit the manuscript for publication. RESULTS Recruitment took place from November 2000 to October 2005, and the last follow-up was in September Of the 481 patients who met eligibility criteria, 312 declined to participate, with most preferring not to Annals of Internal Medicine Vol. 162 No. 7 7 April

4 ORIGINAL RESEARCH Versus Nonsurgical Treatment of Lumbar Spinal Stenosis Figure 1. Study flow diagram. Assessed for eligibility (n = 5119) Enrollment Excluded (n = 4950) Did not meet inclusion criteria: 4638 Aged <50 y: 436 No imaging examination: 2769 Inability to walk: 32 Not a surgical candidate: 361 Did not consent to surgery: 684 Dementia: 7 Metastatic cancer: 6 Severe osteoporosis: 21 Cardiopulmonary disease: 37 Spondylolisthesis: 408 Not an English speaker: 2 Declined to participate: 312 Included in baseline examination and randomization (n = 169) Allocation Allocated to surgery (n = 87) Received allocated intervention: 85 Did not receive allocated intervention (declined surgery): 2 Allocated to physical therapy (n = 82) Received allocated intervention: 69 Did not receive allocated intervention: 13 Cost of copayment: 3 No transportation to and from clinic: 1 Wanted surgery soon: 9 10-wk follow-up Lost to follow-up (n = 4) Declined further participation: 4 Crossed over to (n = 2) 10-wk follow-up Lost to follow-up (n = 3) Declined further participation: 3 Crossed over to surgery (n = 31) Follow-up 6-mo follow-up Lost to follow-up (n = 5) Declined further participation: 3 Died: 2 Crossed over to : 0 12-mo follow-up Lost to follow-up (n = 2) Died: 2 Crossed over to : 0 6-mo follow-up Lost to follow-up (n = 3) Declined further participation: 2 Died: 1 Crossed over to surgery: 9 12-mo follow-up Lost to follow-up (n = 1) Dementia: 1 Crossed over to surgery: 5 24-mo follow-up Lost to follow-up (n = 2) Declined further participation: 2 Crossed over to : 0 24-mo follow-up Lost to follow-up (n = 2) Died: 1 Declined further participation: 1 Crossed over to surgery: 2 Analysis Analyzed (n = 87) Excluded from analysis: 0 Analyzed (n = 82) Excluded from analysis: 0 = physical therapy. 468 Annals of Internal Medicine Vol. 162 No. 7 7 April

5 Versus Nonsurgical Treatment of Lumbar Spinal Stenosis ORIGINAL RESEARCH risk being randomly assigned to the nonsurgical group and instead going straight to surgery (Figure 1). A total of 169 participants was randomly assigned. Study recruitment was halted in October 2005 before meeting recruitment goals because of slow recruitment and limited resources to carry out the remainder of the trial. Results from planned interim analyses conducted after 48, 96, and 158 participants completed 12-month follow-up were not significant. Patients Patient characteristics were similar in both groups (Table 1), except that patients in the surgery group were, on average, 3 years younger than those in the group. Patients in both groups entered the study with numerical pain ratings of 7 out of 10. Crossovers All but 2 patients assigned to the surgery group received surgery. These crossovers occurred before week 10. In contrast, 47 (57%) of the 82 participants in the group crossed over to surgery over the 2-year period (Figure 1). Of these, 79% had received, with an average of 7.8 visits. In addition, 66% (n = 31) of crossovers had surgery before week 10; among these, 74% (n = 23) had at least 1 session, with an average of 7.5 visits. Most demographic and clinical characteristics of participants in the group who crossed over to surgery were similar to those who did not cross over, except cross overs had higher pain and lower education levels (Appendix Table 1, available at Treatment Fidelity of The average number of sessions attended was 8.4 (SD, 4.6). Fifty-four patients (66%) attended at least 50% of the prescribed 12 sessions. Thirteen patients (16%) did not attend at least 1 session, and 77% of them (n = 10) had surgery. Primary Treatment Effects Mean changes in physical function for the surgery and groups were 22.4 (95% CI, 16.9 to 27.9) and 19.2 (CI, 13.6 to 24.8), respectively. The ITT analyses revealed no difference between the surgery and groups at all points of follow-up (P > 0.50), including the 2-year primary end point (adjusted mean difference, 0.85 [CI, 7.9 to 9.6]) (Table 2 and Figure 2). Sensitivity analyses for missing data did not result in different findings for physical function. We controlled for design features of the trial (surgeon, sex, baseline physical function, and imbalance in age) and found that the difference in physical function at 2 years using the CACE approach, with predictors of adherence (sex, baseline pain, and education), was nonsignificant (estimate, 8.5 [CI, 14.4 to 31.4]). Further, the CACE approach resulted in a wider CI than did the ITT approach. The IPW estimate with the same predictors for adherence and 2 predictors for censoring (treatment and depression) was also not significant (estimate, 3.2 [CI, 5.9 to 12.1]), but it had a CI similar to the ITT analyses. Supplement 2 (available at provides a technical description of CACE Table 1. Baseline Demographic and Clinical Characteristics of Participants* Variable (n 87) (n 82) Mean age at enrollment (SD), y 66.6 (10.5) 69.8 (9.0) Men, n (%) 44 (51) 44 (54) Mean height (SD), cm (12.5) (13.1) Mean weight (SD), kg 89.1 (19.9) 91.0 (19.3) Mean BMI (SD), kg/m (6.8) 31.1 (5.4) Race, n (%) White 83 (95) 77 (94) African American 4 (5) 5 (6) Hispanic ethnicity 0 (0) 0 (0) Education level, n (%) High school or less 38 (44) 40 (49) Some college 23 (26) 21 (26) College degree 26 (30) 21 (26) Marital status, n (%) Single 1 (1) 3 (4) Married 63 (72) 59 (72) Divorced/separated 14 (16) 8 (10) Widowed/other 9 (10) 12 (15) Employment situation, n (%) Currently working 22 (25) 17 (21) Unemployed 4 (5) 1 (1) Retired 50 (57) 52 (63) Disabled and/or retired due 11 (13) 12 (15) to health problems Government programs in which currently enrolled, n (%) Social Security 54 (62) 63 (77) Disability 9 (10) 10 (12) Workers' compensation 1 (1) 0 (0) Present activity level, n (%) Inactive/mildly active 83 (95) 74 (90) Moderately/very active 4 (5) 8 (10) Median comorbid conditions (IQR), n 1.0 ( ) 1.0 ( ) Cigarette smoker, n (%) 8 (9) 5 (6) Mean current level of pain (SD) 6.8 (2.2) 6.6 (2.2) Mean BDI score (SD) 9.3 (6.3) 7.9 (4.5) Median length of back pain 233 ( ) 282 ( ) episode (IQR), d Prior back surgeries not for the same spinal segment, n (%) 0 79 (91) 65 (79) 1 7 (8) 12 (15) >1 1 (1) 5 (6) Mean prior series of for LBP (SD) 4.4 (5.7) 4.4 (4.5) BDI = Beck Depression Inventory; BMI = body mass index; IQR = interquartile range; LBP = low back pain; = physical therapy. * Percentages may not sum to 100 due to rounding. P < Measured by a numeric pain scale (0 to 10). A 21-item scale that measures symptoms of depression. Each item is scored from 0 to 3, with a possible range of scores between 0 and 63. Higher numbers indicate greater severity of depressive symptoms. and IPW analyses and results. We also provide the trajectories of physical function according to the different categories of treatment that participants received during the trial, including participants randomly assigned to surgery who had surgery, those randomly assigned to surgery who had, those randomly assigned to who had surgery, and those randomly assigned to who did not have surgery (Appendix Table 2 and Appendix Figure, available at Analysis of Success and Failure In the group assigned to surgery, 45 of 74 participants (61%) with available data achieved a successful Annals of Internal Medicine Vol. 162 No. 7 7 April

6 ORIGINAL RESEARCH Versus Nonsurgical Treatment of Lumbar Spinal Stenosis Table 2. Changes in Outcome Over Time in the and Groups* Outcome Baseline 10 wk 26 wk Participants, n Participants, n Participants, n Primary SF-36 physical function (23.2 to 30.4) (37.1 to 47.9) (41.1 to 53.3) (23.9 to 32.5) (35.3 to 46.7) (39.3 to 51.5) Secondary ODI (39.5 to 45.7) (29.1 to 37.1) (24.5 to 32.5) (36.9 to 43.5) (29.6 to 37.5) (23.3 to 31.1) NASS pain and disability scale (2.5 to 2.8) (1.8 to 2.2) (1.5 to 1.9) (2.3 to 2.6) (1.8 to 2.2) (1.4 to 1.8) NASS neurogenic symptoms scale (3.4 to 4.0) (2.2 to 2.8) (2.2 to 2.7) (3.6 to 4.1) (2.4 to 3.1) (2.3 to 3.0) NASS treatment expectation scale** (3.6 to 4.0) (2.1 to 2.6) (2.2 to 2.8) (3.5 to 4.0) (2.4 to 2.9) (2.4 to 3.0) NASS = North American Spine Society; ODI = Oswestry Low Back Pain Disability Questionnaire; = physical therapy; SF-36 = Short Form-36. * Mean scores and 95% CIs are not adjusted unless otherwise noted. The adjusted difference between the surgery and groups over time was compared using linear mixed-effects models with adjustments for sex, surgeon, and baseline age. P values for group time interaction effect were >0.50 for the primary outcome and all secondary outcomes. Adjusted mean differences and 95% CIs were derived from contrasts from the linear mixed models. A self-administered questionnaire that assesses the concept of physical functioning. Scores range from 0 to 100. A self-administered questionnaire based on 10 areas of performance: pain intensity, lifting, sitting, standing, walking, traveling, personal hygiene, social activity, sex life, and sleeping. Scores range from 0 to 100. A self-administered questionnaire based on 11 items of performance: back pain, buttock pain, dressing, lifting, sitting, standing, walking, traveling, social activity, sex life, and sleeping. Scores range from 1 to 6. A self-administered questionnaire based on 6 items of performance, including leg pain, numbness/tingling, and weakness in leg/foot. Scores range from 1 to 6. ** A self-administered questionnaire based on 6 items of performance: relief from symptoms, do more every day, sleep more comfortably, go back to my usual job, exercise and do recreational activities, and prevent future disability. Scores range from 1 to 5. outcome at 2-year follow-up. Among the 44 participants who crossed over from to surgery, 24 (55%) achieved a successful outcome. Among the 29 participants in the group who did not cross over, 15 (52%) had a successful outcome. Sex-Related Findings Subgroup analysis of sex revealed no difference in treatment effects over time between men and women (sex-by-group-by-time interaction, P = 0.066; group-bytime interaction for men, P = 0.10; group-by-time interaction for women, P = 0.50) (Figure 3 and Appendix Table 3, available at Complications Thirty-three surgery-related complications occurred, 11 of which were in participants who crossed over from to surgery (Appendix Table 4, available at The most common surgery-related complication was reoperation, which included delay in wound healing and surgical site infection. All 9 related complications were reports of worsening symptoms. Although 6 participants died during the study (4 in the surgery group and 2 in the group), the deaths were not related to study participation. The other complications recorded during study implementation were not directly related to study interventions and were similar in both groups. The most common were worsening symptoms; injuries, such as broken bones and falls; and neurologic and neuromuscular complications, such as vestibular disorders and stroke. Other Therapies The use of therapies outside of study regimens, such as injections, exercises, orthotics, and visits to a pain clinic, were similarly low in both groups (Appendix Table 5, available at More than 50% of participants in both groups used pain medication throughout the study. The rate of visits to health care providers was also similar between both groups. Participants continued to seek spine surgeons throughout the study; during year 2, approximately 10% of participants sought this service. Visits to the general practitioner were around 20% throughout the study. The other health care providers were not visited as frequently. In addition, 21 participants (24%) in the surgery group had some form of after the surgery. Thereafter, outside of the study regimen was used by 18% in the surgery group and 20% in the group during year 1 and by 22% in both groups during year 2. DISCUSSION In patients with LSS who were considered surgical candidates and who consented to surgery, no differences were detected between and simple decompression surgery with respect to relieving symptoms 470 Annals of Internal Medicine Vol. 162 No. 7 7 April

7 Versus Nonsurgical Treatment of Lumbar Spinal Stenosis ORIGINAL RESEARCH Table 2 Continued 52 wk 104 wk Participants, n Participants, n Mean Difference Adjusted Mean Difference 1.9 ( 7.3 to 11.2) 0.9 ( 7.9 to 9.6) (44.9 to 57.3) (43.1 to 55.9) (41.4 to 54.4) (40.7 to 54.4) 1.7 ( 7.3 to 3.8) 1.8 ( 7.3 to 3.7) (25.1 to 33.4) (21.5 to 29.0) (25.2 to 33.8) (22.8 to 31.1) 0.1 ( 0.4 to 0.2) 0.1 ( 0.3 to 0.2) (1.5 to 2.0) (1.4 to 1.8) (1.5 to 2.0) (1.4 to 1.9) 0.2 ( 0.6 to 0.3) 0.1 ( 0.6 to 0.3) (2.3 to 2.9) (2.1 to 2.7) (2.2 to 2.9) (2.2 to 2.9) 0.1 ( 0.6 to 0.4) 0.1 ( 0.5 to 0.4) (2.3 to 2.9) (2.5 to 3.1) (2.3 to 2.9) (2.6 to 3.2) and improving function. Results of the ITT analyses showed that participants in both groups began to improve at 10 weeks; continued to improve through 26 weeks; and maintained improvement at the remaining points of follow-up, including the 2-year end point. Compared with baseline, the magnitude of improvement at the 2-year follow-up for all groups was well beyond the minimum clinically important differences for measures of function (29). Although the results showed no statistically significant differences in outcome between groups, the confidence bounds for the between-group difference in physical function improvement show us that our results are consistent with as much as a 7.9-point benefit with assignment to and as much as a 9.6-point benefit with assignment to surgery. If a meaningful difference in SF-36 physical function score is considered to be less than 9.6 points, then these results could be interpreted as indeterminate. Although we were able to minimize crossover in the surgery group, the high percentage of crossover in the group presents a challenge in interpretation. But alternative analytic techniques (for example, CACE or IPW) resulted in similar interpretations of the data namely, that any differences between the groups were not significant. Similar to the observations associated with the ITT results, the CIs of the IPW and CACE analyses were wide, included the thresholds of meaningful differences, and should be interpreted as indeterminate (30 32). By design, participants were required to consent to surgery before consenting to the study, which resulted in 65% of eligible patients declining to participate. This large percentage may include patients with attributes that would limit the study's generalizability, including those who were inherently less risk-averse or those who cannot participate in. Given that 43% of the patients assigned to the group (n = 35) avoided surgery at 2-year follow-up, it is reasonable to assume that many patients had not exhausted their nonsurgical options before consenting to surgery. This could be due to inaccurate self-reported attempts of nonsurgical care (for example, ) or that the received was inadequate because of suboptimal care standards in environments or barriers limiting adherence to (for example, burden of copayments). Unlike other LSS RCTs (6, 7, 11), our study participants were surgical candidates and had consented to surgery before being randomly assigned. This criterion successfully reduced crossover from surgery to to less than 3%, although the crossover rate from to surgery (57%) was higher than that in SPORT (43%), possibly because of differences in patient eligibility. In Figure 2. Adjusted means for physical function over time in the surgery and groups. Physical Function Score on SF Week Adjusted means and 95% CIs of the physical function scale of the SF-36 for the surgery and groups over time from linear mixed-effects models (adjusted for sex, surgeon, and baseline age). The SF-36 scale ranges from 0 to 100, with lower scores indicating more severe symptoms. = physical therapy; SF-36 = Short Form Annals of Internal Medicine Vol. 162 No. 7 7 April

8 ORIGINAL RESEARCH Versus Nonsurgical Treatment of Lumbar Spinal Stenosis Figure 3. Adjusted means for physical function over time using intention-to-treat analyses, by treatment and sex. Physical Function Score on SF in men in men in women in women Week Adjusted means of the physical function scale of the SF-36 for the surgery (solid lines) and (dotted lines) groups by sex over time from linear mixed-effects models (adjusted for surgeon and baseline age). Sex-by-group-by-time interaction (P = 0.066), group-by-time interaction for men (P = 0.10), and group-by-time interaction for women (P = 0.50). The SF-36 scale ranges from 0 to 100, with lower scores indicating more severe symptoms. = physical therapy; SF-36 = Short Form-36. contrast, Malmivaara and colleagues (7) excluded participants whose signs and symptoms suggested forthcoming treatment (for example, surgery), which was the opposite of our study. Despite differences in inclusion and exclusion criteria, Malmivaara and colleagues' study, SPORT (6), and our study reported similar baseline levels of disability based on the ODI. In our surgery groups (participants randomly assigned and crossovers), the ODI change over 2 years was also similar to that in SPORT and was slightly higher than that in Malmivaara and colleagues' study. The major difference was the greater degree of improvement in the group in our trial than in the nonoperative groups of SPORT and Malmivaara and colleagues' study; this might be partially explained by how most of our study's participants in the nonoperative group (the group) received exercise and instruction as opposed to SPORT (6), which did not control the nonoperative interventions. Although women improved in both groups, the magnitude of change lagged behind men, which is a sex difference that has been described in the LSS literature (33 35) but not reported in previous RCTs of LSS (Figure 3). Baseline measures related to demographic and clinical characteristics did not account for this difference. The possible sex difference in outcomes for LSS treatments requires further study. Most participants (84%) in the group attended at least 1 session, with 66% attending at least 50% of the prescribed sessions. We did not see a relationship between attendance in and outcome, including the probability of success. Although we did not track the reasons for nonattendance, several members of the study team reported that copayments were a major obstacle in attending. Of eligible patients, 65% declined to participate primarily because they did not wish to risk the 50% chance of being assigned to the nonsurgical group. We recognize that there could be key differences between the cohorts who did and did not give consent, which should be a strong consideration when generalizing the trial results. Although we successfully controlled for crossover in the surgery group, 57% of patients assigned to crossed over to surgery, which presents a challenge in interpreting ITT analyses despite that alternative analytic techniques were used. Functional outcome in both groups improved to levels of clinical meaningfulness. Without a control group, however, changes in functional outcome cannot be judged as attributable to either intervention or alternative explanations (for example, regression to the mean). Patients with LSS who were surgical candidates and who consented to surgery achieved similar long-term functional gains when offered surgical decompression compared with an evidence-based regimen. Although proportions of successes were similar with both groups, there were also similar proportions of patients who did not achieve a clinically meaningful level of improvement. Future research should focus on a more expanded approach to defining predictors of success and failure with surgical and nonsurgical approaches to symptomatic LSS. In the meantime, patients, surgeons, and their primary care providers should engage in shared decision-making conversations that include a full disclosure of the evidence involving surgical and nonsurgical interventions for LSS. From University of Pittsburgh and University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania; University of Utah, Salt Lake City, Utah; and University of Pennsylvania; Philadelphia, Pennsylvania. Grant Support: By the National Institutes of Health and National Institute of Arthritis and Musculoskeletal and Skin Diseases (grant AR-NS45622; Dr. Delitto). Disclosures: Disclosures can be viewed at /authors/icmje/conflictofinterestforms.do?msnum=m Reproducible Research Statement: Study protocol: See Supplement 3 (available at for more details. Statistical code: Available from Dr. Moore ( , Data set: Not available. Requests for Single Reprints: Anthony Delitto,, PhD, Department of Physical Therapy, School of Health and Rehabilitation Sciences, University of Pittsburgh, 6035 Forbes Tower, Pittsburgh, PA Current author addresses and author contributions are available at Annals of Internal Medicine Vol. 162 No. 7 7 April

9 Versus Nonsurgical Treatment of Lumbar Spinal Stenosis ORIGINAL RESEARCH References 1. Katz JN, Harris MB. Clinical practice. Lumbar spinal stenosis. N Engl J Med. 2008;358: [PMID: ] doi: /NEJMcp Boden SD. The use of radiographic imaging studies in the evaluation of patients who have degenerative disorders of the lumbar spine. J Bone Joint Surg Am. 1996;78: [PMID: ] 3. Katz JN, Dalgas M, Stucki G, Katz NP, Bayley J, Fossel AH, et al. Degenerative lumbar spinal stenosis. Diagnostic value of the history and physical examination. Arthritis Rheum. 1995;38: [PMID: ] 4. Deyo RA. Treatment of lumbar spinal stenosis: a balancing act. Spine J. 2010;10: [PMID: ] doi: /j.spinee Atlas SJ, Delitto A. Spinal stenosis: surgical versus nonsurgical treatment. Clin Orthop Relat Res. 2006;443: [PMID: ] 6. Weinstein JN, Tosteson TD, Lurie JD, Tosteson AN, Blood E, Hanscom B, et al; SPORT Investigators. Surgical versus nonsurgical therapy for lumbar spinal stenosis. N Engl J Med. 2008;358: [PMID: ] doi: /nejmoa Malmivaara A, Slätis P, Heliövaara M, Sainio P, Kinnunen H, Kankare J, et al; Finnish Lumbar Spinal Research Group. Surgical or nonoperative treatment for lumbar spinal stenosis? A randomized controlled trial. Spine (Phila Pa 1976). 2007;32:1-8. [PMID: ] 8. Atlas SJ, Keller RB, Wu YA, Deyo RA, Singer DE. Long-term outcomes of surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation: 10 year results from the Maine lumbar spine study. Spine (Phila Pa 1976). 2005;30: [PMID: ] 9. Weinstein JN, Lurie JD, Olson PR, Bronner KK, Fisher ES. United States' trends and regional variations in lumbar spine surgery: Spine (Phila Pa 1976). 2006;31: [PMID: ] 10. Atlas SJ, Keller RB, Wu YA, Deyo RA, Singer DE. Long-term outcomes of surgical and nonsurgical management of lumbar spinal stenosis: 8 to 10 year results from the Maine lumbar spine study. Spine (Phila Pa 1976). 2005;30: [PMID: ] 11. Weinstein JN, Tosteson TD, Lurie JD, Tosteson A, Blood E, Herkowitz H, et al. Surgical versus nonoperative treatment for lumbar spinal stenosis four-year results of the Spine Patient Outcomes Research Trial. Spine (Phila Pa 1976). 2010;35: [PMID: ] doi: /brs.0b013e3181e0f04d 12. Amundsen T, Weber H, Nordal HJ, Magnaes B, Abdelnoor M, Lilleâs F. Lumbar spinal stenosis: conservative or surgical management?: A prospective 10-year study. Spine (Phila Pa 1976). 2000;25: ; discussion [PMID: ] 13. Katz JN, Stucki G, Lipson SJ, Fossel AH, Grobler LJ, Weinstein JN. Predictors of surgical outcome in degenerative lumbar spinal stenosis. Spine (Phila Pa 1976). 1999;24: [PMID: ] 14. Pearson A, Lurie J, Tosteson T, Zhao W, Abdu W, Weinstein JN. Who should have surgery for spinal stenosis? Treatment effect predictors in SPORT. Spine (Phila Pa 1976). 2012;37: [PMID: ] 15. Radcliff KE, Rihn J, Hilibrand A, DiIorio T, Tosteson T, Lurie JD, et al. Does the duration of symptoms in patients with spinal stenosis and degenerative spondylolisthesis affect outcomes?: analysis of the Spine Outcomes Research Trial. Spine (Phila Pa 1976). 2011;36: [PMID: ] doi: /brs.0b013e edf 16. Rihn JA, Radcliff K, Hilibrand AS, Anderson DT, Zhao W, Lurie J, et al. Does obesity affect outcomes of treatment for lumbar stenosis and degenerative spondylolisthesis? Analysis of the Spine Patient Outcomes Research Trial (SPORT). Spine (Phila Pa 1976). 2012;37: [PMID: ] doi: /brs.0b013e31825e21b2 17. Fritz JM, Lurie JD, Zhao W, Whitman JM, Delitto A, Brennan GP, et al. Associations between physical therapy and long-term outcomes for individuals with lumbar spinal stenosis in the SPORT study. Spine J. 2014;14: [PMID: ] doi: /j.spinee Wiesel SW, Tsourmas N, Feffer HL, Citrin CM, Patronas N. A study of computer-assisted tomography. I. The incidence of positive CAT scans in an asymptomatic group of patients. Spine (Phila Pa 1976). 1984;9: [PMID: ] 19. Rothman RH and Simeone FA, eds. The Spine. 3 ed. Philadelphia: WB Saunders; Birkmeyer NJ, Weinstein JN, Tosteson AN, Tosteson TD, Skinner JS, Lurie JD, et al. Design of the Spine Patient outcomes Research Trial (SPORT). Spine (Phila Pa 1976). 2002;27: [PMID: ] 21. Fritz JM, Erhard RE, Vignovic M. A nonsurgical treatment approach for patients with lumbar spinal stenosis. Phys Ther. 1997;77: [PMID: ] 22. Ware JE Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care. 1992;30: [PMID: ] 23. Fairbank JC, Couper J, Davies JB, O Brien JP. The Oswestry low back pain disability questionnaire. Physiotherapy. 1980;66: [PMID: ] 24. Daltroy LH, Cats-Baril WL, Katz JN, Fossel AH, Liang MH. The North American spine society lumbar spine outcome assessment Instrument: reliability and validity tests. Spine (Phila Pa 1976). 1996;21: [PMID: ] 25. Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. 2 ed. Hoboken: Wiley; Knox CR, Lall R, Hansen Z, Lamb SE. Treatment compliance and effectiveness of a cognitive behavioural intervention for low back pain: a complier average causal effect approach to the BeST data set. BMC Musculoskelet Disord. 2014;15:17. [PMID: ] doi: / Skrondal A, Rabe-Hesketh S. Generalized Latent Variable Modeling: Multilevel, Longitudinal, and Structural Equation Models. Boca Raton, FL: Chapman & Hall/CRC; Toh S, Hernán MA. Causal inference from longitudinal studies with baseline randomization. Int J Biostat. 2008;4:Article 22. [PMID: ] doi: / Copay AG, Glassman SD, Subach BR, Berven S, Schuler TC, Carreon LY. Minimum clinically important difference in lumbar spine surgery patients: a choice of methods using the Oswestry Disability Index, Medical Outcomes Study questionnaire Short Form 36, and pain scales. Spine J. 2008;8: [PMID: ] doi: /j.spinee Lauridsen HH, Hartvigsen J, Manniche C, Korsholm L, Grunnet- Nilsson N. Responsiveness and minimal clinically important difference for pain and disability instruments in low back pain patients. BMC Musculoskelet Disord. 2006;7:82. [PMID: ] 31. Angst F, Aeschlimann A, Stucki G. Smallest detectable and minimal clinically important differences of rehabilitation intervention with their implications for required sample sizes using WOMAC and SF-36 quality of life measurement instruments in patients with osteoarthritis of the lower extremities. Arthritis Rheum. 2001;45: [PMID: ] 32. Davidson M, Keating JL. A comparison of five low back disability questionnaires: reliability and responsiveness. Phys Ther. 2002;82:8-24. [PMID: ] 33. Johnsson KE, Redlund-Johnell I, Udén A, Willner S. Preoperative and postoperative instability in lumbar spinal stenosis. Spine (Phila Pa 1976). 1989;14: [PMID: ] 34. Kim HJ, Suh BG, Lee DB, Park JY, Kang KT, Chang BS, et al. Gender difference of symptom severity in lumbar spinal stenosis: role of pain sensitivity. Pain Physician. 2013;16:E [PMID: ] 35. Shabat S, Folman Y, Arinzon Z, Adunsky A, Catz A, Gepstein R. Gender differences as an influence on patients' satisfaction rates in spinal surgery of elderly patients. Eur Spine J. 2005;14: [PMID: ] Annals of Internal Medicine Vol. 162 No. 7 7 April

10 Annals of Internal Medicine Current Author Addresses: Drs. Delitto, Piva, and Josbeno: Department of Physical Therapy, School of Health and Rehabilitation Sciences, University of Pittsburgh, Bridgeside Point 1, 100 Technology Drive, Suite 210, Pittsburgh, PA Dr. Moore: Department of Medicine, University of Pittsburgh, M240 Scaife Hall, 3550 Terrace Street, Pittsburgh, PA Dr. Fritz: Department of Physical Therapy, School of Health Professions, University of Utah, Health Profession Education Building, 520 Wakara Way, Salt Lake City, UT Dr. Wisniewski: Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, 127 Parran Hall, Pittsburgh, PA Dr. Fye: University of Pittsburgh Medical Center, 200 Lothrop Street, Pittsburgh, PA Dr. Welch: University of Pennsylvania, Penn Neurosurgery, Pennsylvania Hospital, Washington Square West Building, 235 South 8th Street, Philadelphia, PA Author Contributions: Conception and design: A. Delitto, J.M. Fritz, S.R. Wisniewski, W.C. Welch. Analysis and interpretation of the data: A. Delitto, C.G. Moore, S.R. Wisniewski, W.C. Welch. Drafting of the article: A. Delitto, S.R. Piva, C.G. Moore, J.M. Fritz, M. Fye, W.C. Welch. Critical revision of the article for important intellectual content: A. Delitto, S.R. Piva, C.G. Moore, D.A. Josbeno, W.C. Welch. Final approval of the article: A. Delitto, S.R. Piva, C.G. Moore, J.M. Fritz, M. Fye, W.C. Welch. Provision of study materials or patients: M. Fye, W.C. Welch. Statistical expertise: C.G. Moore, S.R. Wisniewski. Obtaining of funding: A. Delitto, J.M. Fritz, W.C. Welch. Administrative, technical, or logistic support: A. Delitto, S.R. Piva, D.A. Josbeno, M. Fye, W.C. Welch. Collection and assembly of data: J.M. Fritz, S.R. Wisniewski, D.A. Josbeno, M. Fye, W.C. Welch. Appendix Table 1. Baseline Demographic and Clinical Characteristics of Participants in the Group* Variable No Crossover (n 35) Crossover (n 47) Mean age at enrollment (SD), y 70.7 (9.1) 69.0 (8.9) Men, n (%) 22 (63) 22 (47) Mean height (SD), cm (13.0) (13.3) Mean weight (SD), kg 89.6 (18.9) 92.1 (19.7) Mean BMI (SD), kg/m (5.5) 31.5 (5.4) Race, n (%) White 34 (97) 43 (91) African American 1 (3) 4 (9) Hispanic ethnicity 0 (0) 0 (0) Education level, n (%) High school or less 12 (34) 28 (60) Some college 12 (34) 9 (19) College degree 11 (31) 10 (21) Marital status, n (%) Single 1 (3) 2 (4) Married 24 (69) 35 (74) Divorced/separated 4 (11) 4 (9) Widowed/other 6 (17) 6 (13) Employment situation, n (%) Currently working 7 (20) 10 (21) Unemployed 0 (0) 1 (2) Retired 24 (69) 28 (60) Disabled and/or retired due 4 (11) 8 (17) to health problems Government programs in which currently enrolled, n (%) Social Security 27 (77) 36 (77) Disability 3 (9) 7 (15) Workers' compensation 0 (0) 0 (0) Present activity level, n (%) Inactive/mildly active 31 (88) 43 (90) Moderately/very active 4 (12) 4 (9) Median comorbid conditions (IQR), n 1.0 ( ) 1.0 ( ) Cigarette smoker, n (%) 1 (3) 4 (9) Mean current level of pain (SD) 5.9 (2.3) 7.1 (2.0) Mean BDI score (SD) 7.4 (4.3) 8.2 (4.6) Median length of back pain 279 ( ) 308 ( ) episode (IQR), d Prior back surgeries not for the same spinal segment, n (%) 0 27 (77) 38 (81) 1 6 (17) 6 (13) >1 2 (6) 3 (6) Mean prior series of for LBP (SD) 4.7 (4.8) 4.1 (4.4) BDI = Beck Depression Inventory; BMI = body mass index; IQR = interquartile range; LBP = low back pain; = physical therapy. * Percentages may not sum to 100 due to rounding. P < P < Measured by a numeric pain scale (0 to 10). A 21-item scale that measures symptoms of depression. Each item is scored from 0 to 3, with a possible range of scores between 0 and 63. Higher numbers indicate greater severity of depressive symptoms. Annals of Internal Medicine Vol. 162 No. 7 7 April 2015

11 Appendix Table 2. Changes in SF-36 Physical Function Score Over Time According to the Treatment Received* Group Baseline 10 wk 26 wk 52 wk 104 wk Participants, n Participants, n Participants, n Participants, n Participants, n (22.8 to 30.3) (37.2 to 48.2) (41.1 to 53.5) (45.2 to 57.7) (43.3 to 56.3) Randomly assigned to and had only surgery ( to 255.0) ( 73.0 to 146.0) ( to 265.0) ( to 358.0) Randomly assigned to surgery and had (21.2 to 32.0) (32.5 to 47.7) (38.9 to 56.6) (40.6 to 58.3) (38.0 to 56.8) Randomly assigned to and had surgery (23.0 to 37.8) (33.0 to 51.1) (33.7 to 51.0) (35.8 to 55.8) (37.6 to 58.0) Randomly assigned to and had only = physical therapy; SF-36 = Short Form-36. * Mean SF-36 physical function scores and 95% CIs are not adjusted. The SF-36 physical function is a self-administered questionnaire that assesses the concept of physical functioning. Scores range from 0 to 100. Value pertains to 1 participant; thus, there is no CI. Appendix Figure. Physical function over time, by treatment received. Physical Function Score on SF Week Randomly assigned to and had only surgery (n = 85) Randomly assigned to surgery and had (n = 2) Randomly assigned to and had surgery (n = 47) Randomly assigned to and had only (n = 35) The SF-36 scale ranges from 0 to 100, with lower scores indicating more severe symptoms. = physical therapy; SF-36 = Short Form-36. Annals of Internal Medicine Vol. 162 No. 7 7 April