Surgical versus Non-Operative Treatment for Lumbar Disc Herniation: Four-Year Results for the Spine Patient Outcomes Research Trial (SPORT)

Abstract

Study Design

Concurrent prospective randomized and observational cohort study.

Objectives

To assess the 4-year outcomes of surgery vs. non-operative care.

Background

Although randomized trials have demonstrated small short-term differences in favor of surgery, long-term outcomes comparing surgical to non-operative treatment remain controversial.

Methods

Surgical candidates with imaging-confirmed lumbar intervertebral disc herniation meeting SPORT eligibility criteria enrolled into prospective randomized (501 participants) and observational cohorts (743 participants) at 13 spine clinics in 11 US states. Interventions were standard open discectomy versus usual non-operative care. Main outcome measures were changes from baseline in the SF-36 Bodily Pain (BP) and Physical Function (PF) scales and the modified Oswestry Disability Index (ODI - AAOS/Modems version) assessed at 6 weeks, 3 and 6 months, and annually thereafter.

Results

Non-adherence to treatment assignment caused the intent-to-treat analyses to underestimate the treatment effects. In the 4-year combined as-treated analysis, those receiving surgery demonstrated significantly greater improvement in all the primary outcome measures (mean change Surgery vs. Non-operative; treatment effect; 95% CI): BP (45.6 vs. 30.7; 15.0; 11.8 to 18.1), PF (44.6 vs. 29.7; 14.9; 12.0 to 17.8) and ODI (−38.1 vs. −24.9; −13.2; −15.6 to −10.9). The percent working was similar between the surgery and non-operative groups, 84.4% vs. 78.4% respectively.

Conclusion

In a combined as-treated analysis at 4 years, patients who underwent surgery for a lumbar disc herniation achieved greater improvement than non-operatively treated patients in all primary and secondary outcomes except work status.

Trial Registration

Spine Patient Outcomes Research Trial (SPORT): Intervertebral Disc Herniation; #NCT00000410; http://www.clinicaltrials.gov/ct/show/NCT00000410?order=2

INTRODUCTION

Lumbar disc surgery remains one of the most commonly performed operations, with rates exhibiting considerable geographic variation.¹ Two recent randomized trials demonstrated that surgery provides faster pain relief and perceived recovery in patients with herniated disc.^2–4 Outcomes were similar at 1 year for patients assigned to surgery and for those assigned to non-operative treatment. However, both trials included substantial numbers of surgical patients in the non-operative comparison arm due to treatment crossover, affecting the interpretation of the intent-to-treat analyses. This paper reports 4-year results for SPORT based on the continued follow-up of the herniated disc randomized and observational cohorts.

METHODS

Study Design

SPORT was conducted in 11 US states at 13 medical centers with multidisciplinary spine practices. The human subjects committees at each participating institution approved a standardized protocol for both the observational and the randomized cohorts. Patient inclusion and exclusion criteria, study interventions, outcome measures, and follow-up procedures have been reported previously.^3–5

Patient Population

Men and women who had symptoms and confirmatory signs of lumbar radiculopathy that persisted for at least six weeks, who had disc herniation at a corresponding level and side on imaging, and who were considered surgical candidates were eligible. The content of pre-enrollment non-operative care was not pre-specified in the protocol. ^3–5 Specific enrollment and exclusion criteria are reported elsewhere.^4,5

A research nurse at each site identified potential participants and verified eligibility. Participants were offered enrollment in either the randomized trial or the observational cohort. Enrollment began in March of 2000 and ended in November of 2004.

Study Interventions

The surgery was a standard open discectomy with examination of the involved nerve root.^5,6 The non-operative protocol was “usual care” recommended to include at least: active physical therapy, education/counseling with home exercise instruction, and non-steroidal anti-inflammatory drugs if tolerated. Non-operative treatments were individualized for each patient and tracked prospectively. ^3–5

Study Measures

Primary endpoints were the Bodily Pain (BP) and Physical Function (PF) scales of the SF-36 Health Survey⁷ and the AAOS/Modems version of the Oswestry Disability Index (ODI)⁸ as measured at 6 weeks, 3 and 6 months, and annually thereafter. If surgery was delayed beyond six weeks, additional follow-up data was obtained 6 weeks and 3 months post-operatively. Secondary outcomes included patient self-reported improvement; work status; satisfaction with current symptoms and care;⁹ and sciatica severity as measured by the sciatica bothersomeness index.^10,11 Treatment effect was defined as the difference in the mean changes from baseline between the surgical and non-operative groups.

Statistical Considerations

Initial analyses compared means and proportions for baseline patient characteristics between the randomized and observational cohorts and between the initial treatment arms of the individual and combined cohorts. The extent of missing data and the percentage of patients undergoing surgery were calculated by treatment arm for each scheduled follow-up. Baseline predictors of time until surgical treatment (including treatment crossovers) in both cohorts were determined via a stepwise proportional hazards regression model with an inclusion criterion of p < 0.1 to enter and p > 0.05 to exit. Predictors of missing follow-up visits at yearly intervals up to 4 years were separately determined via stepwise logistic regression. Baseline characteristics that predicted surgery or a missed visit at any time-point were then entered into longitudinal models of primary outcomes. Those that remained significant in the longitudinal models of outcome were included as adjusting covariates in all subsequent longitudinal regression models to adjust for potential confounding due to treatment selection bias and missing data patterns. ¹² In addition, baseline outcome, center, age and gender were included in all longitudinal outcome models.

Primary analyses compared surgical and non-operative treatments using changes from baseline at each follow-up, with a mixed effects longitudinal regression model including a random individual effect to account for correlation between repeated measurements within individuals. The randomized cohort was initially analyzed on an intent-to-treat basis.⁴ Because of crossover, subsequent analyses were based on treatments actually received. In these as-treated analyses, the treatment indicator was a time-varying covariate, allowing for variable times of surgery. Follow-up times were measured from enrollment for the intent-to-treat analyses, whereas for the as-treated analysis the follow-up times were measured from the beginning of treatment (i.e. the time of surgery for the surgical group and the time of enrollment for the non-operative group), and baseline covariates were updated to the follow-up immediately preceding the time of surgery. This procedure has the effect of including all changes from baseline prior to surgery in the estimates of the non-operative treatment effect and all changes after surgery in the estimates of the surgical effect. The six-point sciatica scales and binary outcomes were analyzed via longitudinal models based on generalized estimating equations¹³ using the same intent-to-treat and adjusted as-treated analysis definitions as the primary outcomes. The randomized and observational cohorts were each analyzed to produce separate as-treated estimates of treatment effect. These results were compared using a Wald test to simultaneously test all follow-up visit times for differences in estimated treatment effects between the two cohorts.¹² Final analyses combined the cohorts.

To evaluate the two treatment arms across all time-periods, the time-weighted average of the outcomes (area under the curve) for each treatment group was computed using the estimates at each time period from the longitudinal regression models and compared using a Wald test. ¹²

Kaplan-Meier estimates of re-operation rates at 4 years were computed for the randomized and observational cohorts and compared via the log-rank test. ^14,15 Crossover from assigned surgery to non-operative treatment and from assigned non-operative treatment to surgery was compared via McNemar’s test. ¹⁶

Computations were done using SAS procedures PROC MIXED for continuous data and PROC GENMOD for binary and non-normal secondary outcomes (SAS version 9.1 Windows XP Pro, Cary, NC). Statistical significance was defined as p < 0.05 based on a two-sided hypothesis test with no adjustments made for multiple comparisons. Data for these analyses were collected through April 10, 2008.

RESULTS

Overall, 1,244 SPORT participants with lumbar intervertebral disc herniation were enrolled (501 in the randomized cohort, and 743 in the observational cohort). (Figure 1) In the randomized cohort, 245 were assigned to surgical treatment and 256 to non-operative treatment. Of those randomized to surgery, 57% had surgery by 1 year and 59% by 4 years. In the group randomized to non-operative care, 41% of patients received surgery by 1 year and 45% received surgery by 4 years. In the observational cohort, 521 patients initially chose surgery and 222 patients initially chose non-operative care. Of those initially choosing surgery, 95% received surgery by 1 year; at 4 years no further surgeries had been reported. Of those choosing non-operative treatment, 20% had surgery by 1 year, 24% by 4 years. In both cohorts combined, 805 patients received surgery at some point during the first 4 years; 439 (35%) remained non-operative. Over the 4 years, 1,192 (96%) of the original enrollees completed at least 1 follow-up visit and were included in the analysis (randomized cohort: 94% and observational cohort 97%); between 65% and 87% of enrollees supplied data at each follow-up interval with losses due to dropouts, missed visits, or deaths (Figure 1).

Figure 1. Exclusion, Enrollment, Randomization and Follow-up of Trial Participants.

The values for surgery, withdrawal, and death are cumulative over 4 years. For example, a total of 1 patient in the group assigned to surgery died during the 4-year follow-up period. [Data set 04/10/2008]

Patient Characteristics

Baseline characteristics are compared in Table 1. Overall, the cohorts were similar. However, patients in the observational cohort had more disability, a strong preference for surgery, more often rated their problem as worsening, and were slightly more likely to have a sensory deficit.

Table 1.

Patient baseline demographic characteristics, comorbidities, and health status measures according to study cohort and treatment received.

	SPORT Study Cohorts			Randomized and Observational Cohorts Combined: Treatment Received*
	Randomized Cohort	Observational Cohort		Surgery	Non-Operative
	(n=473)	(n=719)		(n=788)	(n=404)
Mean Age (stdev)	42.3 (11.6)	41.4 (11.2)	0.16	40.7 (10.8)	43.9 (12.2)	<0.001
Female	194 (41%)	313 (44%)	0.42	340 (43%)	167 (41%)	0.59
Ethnicity: Not Hispanic	449 (95%)	688 (96%)	0.64	752 (95%)	385 (95%)	0.97
Race - White	400 (85%)	633 (88%)	0.10	694 (88%)	339 (84%)	0.056
Education - At least some college	356 (75%)	528 (73%)	0.52	571 (72%)	313 (77%)	0.072
Income - Under $50,000	207 (44%)	328 (46%)	0.57	367 (47%)	168 (42%)	0.11
Marital Status - Married	333 (70%)	502 (70%)	0.88	549 (70%)	286 (71%)	0.74
Work Status			0.70			0.004
Full or part time	291 (62%)	431 (60%)		455 (58%)	267 (66%)
Disabled	58 (12%)	100 (14%)		121 (15%)	37 (9%)
Other	124 (26%)	187 (26%)		211 (27%)	100 (25%)
Compensation - Any	76 (16%)	132 (18%)	0.35	161 (20%)	47 (12%)	<0.001
Mean Body Mass Index (BMI), (stdev)	28 (5.5)	27.9 (5.6)	0.87	28.2 (5.7)	27.5 (5.3)	0.037
Smoker	108 (23%)	174 (24%)	0.64	197 (25%)	85 (21%)	0.15
Comorbidities
Depression	62 (13%)	79 (11%)	0.31	93 (12%)	48 (12%)	0.96
Joint Problem	97 (21%)	124 (17%)	0.18	127 (16%)	94 (23%)	0.003
Other	221 (47%)	305 (42%)	0.16	328 (42%)	198 (49%)	0.018
Time since recent episode < 6 months	373 (79%)	557 (77%)	0.62	609 (77%)	321 (79%)	0.43
Bodily Pain (BP) Score	26.9 (17.9)	25.2 (18.3)	0.11	22.3 (16.2)	32.9 (19.7)	<0.001
Physical Functioning (PF) Score	39.5 (25.3)	36.6 (25.6)	0.056	32.3 (23.4)	48.2 (26.3)	<0.001
Mental Component Summary (MCS) Score	45.9 (12)	44.7 (11.2)	0.086	44.6 (11.4)	46.3 (11.8)	0.021
Oswestry (ODI)	46.9 (21)	51.2 (21.4)	<0.001	54.9 (19.6)	38.8 (20.4)	<0.001
Sciatica Frequency Index (0–24)	15.6 (5.5)	16 (5.3)	0.18	16.7 (5.1)	14.3 (5.6)	<0.001
Sciatica Bothersome Index (0–24)	15.2 (5.2)	15.8 (5.3)	0.057	16.4 (4.9)	13.9 (5.6)	<0.001
Satisfaction with symptoms - very dissatisfied	370 (78%)	584 (81%)	0.23	696 (88%)	258 (64%)	<0.001
Problem getting better or worse			<0.001			<0.001
Getting better	90 (19%)	89 (12%)		65 (8%)	114 (28%)
Staying about the same	221 (47%)	313 (44%)		338 (43%)	196 (49%)
Getting worse	161 (34%)	311 (43%)		379 (48%)	93 (23%)
Treatment preference			<0.001			<0.001
Preference for non-surg	193 (41%)	201 (28%)		120 (15%)	274 (68%)
Not sure	154 (33%)	43 (6%)		112 (14%)	85 (21%)
Preference for surgery	126 (27%)	472 (66%)		553 (70%)	45 (11%)
Pain Radiation	458 (97%)	704 (98%)	0.33	772 (98%)	390 (97%)	0.19
Straight Leg Raise Test - Ipsilateral	291 (62%)	459 (64%)	0.45	515 (65%)	235 (58%)	0.018
Straight Leg Raise Test - Contralateral/Both	67 (14%)	121 (17%)	0.25	149 (19%)	39 (10%)	<0.001
Any Neurological Deficit	351 (74%)	551 (77%)	0.38	617 (78%)	285 (71%)	0.004
Reflexes - Asymmetric Depressed	203 (43%)	278 (39%)	0.16	325 (41%)	156 (39%)	0.42
Sensory - Asymmetric Decrease	222 (47%)	381 (53%)	0.047	429 (54%)	174 (43%)	<0.001
Motor - Asymmetric Weakness	190 (40%)	311 (43%)	0.32	354 (45%)	147 (36%)	0.006
Herniation Level			0.087			<0.001
L2-L3/L3-L4	32 (7%)	56 (8%)		42 (5%)	46 (11%)
L4-L5	165 (35%)	291 (40%)		305 (39%)	151 (37%)
L5-S1	275 (58%)	372 (52%)		441 (56%)	206 (51%)
Herniation Type			0.85			0.30
Protruding	126 (27%)	196 (27%)		204 (26%)	118 (29%)
Extruded	314 (66%)	469 (65%)		530 (67%)	253 (63%)
Sequestered	32 (7%)	54 (8%)		54 (7%)	32 (8%)
Posterolateral herniation	378 (80%)	541 (75%)	0.071	626 (79%)	293 (73%)	0.009

^*Patients in the two cohorts combined were classified according to whether they received surgical treatment or only nonsurgical treatment during the first 2 years of enrollment.

^†Race or ethnic group was self-assessed. Whites and blacks could be either Hispanic or non-Hispanic.

^‡This category includes patients who were receiving or had applications pending for workers compensation, Social Security compensation, or other compensation.

^§The body-mass index is the weight in kilograms divided by the square of the height in meters.

^¶Other = problems related to stroke, diabetes, osteoporosis, cancer, fibromyalgia, CFS”, PTSD, alcohol, drug dependence, heart, lung, liver, kidney, blood vessel, nervous system, hypertension, migraine, anxiety, stomach or bowel.

^||The SF-36 scores range from 0 to 100, with higher score indicating less severe symptoms.

^**The Oswestry Disability Index ranges from 0 to 100, with lower scores indicating less severe symptoms.

^††The Stenosis Frequency Index ranges from 0 to 24, with lower scores indicating less severe symptoms.

^‡‡The Stenosis Bothersomeness Index ranges from 0 to 24, with lower scores indicating less severe symptoms.

^§§The Low Back Pain Bothersomness Scale ranges from 0 to 6, with lower scores indicating less severe symptoms

^¶¶The Leg Pain Bothersomeness Scale ranges from 0 to 6, with lower scores indicating less severe symptoms.

Summary statistics for the combined cohorts are also displayed in Table 1 according to treatment received. The study population had an overall mean age of 41.7 with a mean of 40.7 in the surgery group and a mean of 43.9 in the non-operative group. There were slightly more men than women. Subjects receiving surgery were: younger; less likely to be working; more likely to report being disabled and to be receiving compensation; had slightly greater BMI’s; fewer joint and other co-morbidities; more pain; frequent and bothersome sciatica; depression; dissatisfaction with their symptoms and more often rated them as getting worse; less function; and were more likely to prefer surgery. Subjects receiving surgery also had more ipsilateral and contralateral straight leg tests, and more neurologic, sensory, and motor deficits. Radiographically, their herniations were more likely to be at the L4–5 and L5-S1 levels and to be posterolateral in location.

Non-operative Treatments

Non-operative treatments within 4 years of enrollment were similar between the two cohorts. However, more observational patients reported visits to other practitioners (57% observational vs. 37% randomized, p = <0.001); and randomized patients had more (randomized vs. observational): injections (57% vs. 40%, p= <0.001), activity restriction (32% vs. 20% p=0.004) and narcotics (50% vs. 37% p=0.005).

Surgical Treatment and Complications

Overall surgical treatment and complications were similar between the two cohorts (Table 2). The average surgical time was slightly longer in the randomized cohort (80.6 minutes randomized vs. 74.9 minutes observational, p=0.049). Median (interquartile range) values for surgical time were 74.5 minutes (57.8, 90.0) for the randomized and 70 minutes (50.0, 90.0) for the observational cohort. The average blood loss was 67.5cc in the randomized cohort vs. 63.0cc in the observational, p=0.56. Median (25th percentile, 75^th percentile) for blood loss was 50cc (25, 75) in the randomized cohort and 50cc (25, 50) in the observational. Only 6 patients total required intra-operative transfusions. There were no perioperative mortalities. The most common surgical complication was dural tear (3% of cases). Re-operation occurred in a combined 6% of cases by 1 year, 8% at 2 years, 9% at 3 years, and 10% at 4 years post surgery. The rates of reoperation were not significantly different between the randomized and observational cohorts. Seventy-five of the 81 re-operations noted the type of re-operation; approximately 50% of these (48/75) were listed as recurrent herniations at the same level. One death occurred within 90 days post-surgery related to heart surgery at another institution; the death was judged to be unrelated and was reported to the Institutional Review Board and the Data and Safety Monitoring Board.

Table 2.

Operative treatments, complications and events.

	Randomized Cohort (n=253)	Observational Cohort (n=545)	p-value
Discectomy Level
L2-L3	2 (1%)	12 (2%)	0.27
L3-L4	7 (3%)	20 (4%)	0.69
L4-L5	97 (39%)	215 (40%)	0.98
L5-S1	149 (60%)	305 (56%)	0.32
Median time to surgery in months (95% CI)†	6.8 (4.5, 40.9)	0.5 (0.4, 0.7)	<0.001
Operation time, minutes (SD)	80.6 (41)	74.9 (35.5)	0.049
Blood loss, cc (SD)	67.5 (99.1)	63 (103.2)	0.56
Blood Replacement
Intraoperative replacement	4 (2%)	2 (0%)	0.16
Post-operative transfusion	0 (0%)	0 (0%)	-
Length of stay (SD)	1 (1.1)	0.94 (0.9)	0.26
Post-operative mortality (death within 6 weeks of surgery)	0 (0%)	0 (0%)
Post-operative mortality (death within 3 months of surgery)	0 (0%)	1‡ (0.2%)	0.70
Intraoperative complications§
Dural tear/spinal fluid leak	10 (4%)	14 (3%)	0.40
Nerve root injury	1 (0%)	1 (0%)	0.84
Vascular injury	1 (0%)	0 (0%)	0.69
Other	2 (1%)	1 (0%)	0.50
None	239 (94%)	530 (97%)	0.08
Postoperative complications/events¶
Nerve root injury	0 (0%)	1 (0%)	0.69
Wound hematoma	0 (0%)	4 (1%)	0.41
Wound infection	4 (2%)	14 (3%)	0.55
Other	9 (4%)	18 (3%)	1
None	237 (95%)	510 (94%)	0.67
Additional surgeries (1-year rate)||	10 (4%)	36 (7%)	0.13
Additional surgeries (2-year rate)||	14 (5%)	49 (9%)	0.082
Additional surgeries (3-year rate)||	18 (7%)	52 (10%)	0.23
Additional surgeries (4-year rate)||	22 (9%)	59 (11%)	0.32
Recurrent disc herniation	13 (5%)	36 (7%)
Complication or Other	5 (2%)	16 (3%)
New condition	2 (1%)	7 (1%)

^*259 RCT and 546 OBS patients had surgery, surgical information was available for 253 RCT patients and 545 observational patients.

^‡Patient died after heart surgery at another hospital, the death was judged unrelated to spine surgery.

^§None of the following were reported: aspiration, operation at wrong level, vascular injury.

^¶Any reported complications up to 8 weeks post operation. None of the following were reported: bone graft complication, CSF leak, paralysis, cauda equina injury, wound dehiscence, pseudarthrosis.

^||One-, two-, three- and four-year post-surgical re-operation rates are Kaplan Meier estimates and p-values are based on the log-rank test. Numbers and percentages are based on the first additional surgery if more than one additional surgery.

^†Median and 95% CI based on Kaplan-Meier estimates and p-value based on log-rank test.

Cross-Over

Non-adherence to treatment assignment affected both treatment arms: patients chose to delay or decline surgery in the surgical arm and crossed over to surgery in the non-operative arm. (Figure 1) Some characteristics of crossover patients were statistically different from patients who did not cross over (Table 3). Patients crossing over to non-operative care were older, had higher incomes and less pain and disability, were less likely to have an ipsilateral straight leg raise and to perceive their symptoms as getting worse, more likely to have a high level disc herniation and to express a baseline preference for non-operative care, and were less dissatisfied with their symptoms, Patients crossing over to surgery within 4 years had lower income, worse physical function and more self-rated disability, were more dissatisfied with their symptoms, perceived they were getting worse and expressed a baseline preference for surgery. While more patients crossed from non-operative treatment to surgery [112 (24%)] than crossed from surgery to nonoperative treatment [89 (19%)], this difference is not significant based on a McNemar’s test (p=0.12).

Table 3.

Statistically significant predictors of adherence to treatment among RCT patients.

	Assigned to Surgery			Assigned to Non-operative
	Treatment Received within 4 Years		p-value	Treatment Received within 4 Years		p-value
	Surgery	Non-operative		Surgery	Non-operative
	(n=143)	(n=89)		(n=112)	(n=129)
Mean Age (stdev)	40.2 (11)	44.1 (12.7)	0.01	41.9 (10)	43.8 (12.3)	0.21
Income - Under $50,000	66 (46%)	27 (30%)	0.02	61 (54%)	53 (41%)	0.052
Straight Leg Raise Test - Ipsilateral	96 (67%)	47 (53%)	0.04	74 (66%)	74 (57%)	0.21
Herniation Level			0.008			0.35
L2-L3/L3-L4	4 (3%)	12 (13%)		5 (4%)	11 (9%)
L4-L5	52 (36%)	28 (31%)		43 (38%)	42 (33%)
L5-S1	87 (61%)	49 (55%)		64 (57%)	75 (58%)
Bodily Pain (BP) Score	24.3 (16.7)	31.6 (20.4)	0.003	24.6 (16.8)	28.6 (17.7)	0.072
Physical Functioning (PF) Score	36.2 (24.1)	45.4 (25.3)	0.006	33.8 (23.2)	43.9 (26.8)	0.002
Oswestry (ODI)	51.2 (21)	41.5 (20.8)	<0.001	51.5 (19.5)	41.7 (20.5)	<0.001
Sciatica Frequency Index (0–24)	16.3 (5.2)	15.1 (6)	0.11	16.4 (5.5)	14.6 (5.4)	0.012
Sciatica Bothersome Index (0–24)	15.9 (4.8)	14.7 (5.6)	0.09	16 (5)	14.1 (5.3)	0.004
Satisfaction with symptoms - very dissatisfied	126 (88%)	58 (65%)	<0.001	96 (86%)	90 (70%)	0.005
Problem getting worse	60 (42%)	22 (25%)	0.01	45 (40%)	34 (26%)	0.032
Treatment Preference			<0.001			<0.001
Preference for non-surg	45 (31%)	51 (57%)		29 (26%)	68 (53%)
Not sure	48 (34%)	31 (35%)		36 (32%)	39 (30%)
Preference for surgery	50 (35%)	7 (8%)		47 (42%)	22 (17%)

^*The SF-36 scores range from 0 to 100, with higher score indicating less severe symptoms.

^†The Oswestry Disability Index ranges from 0 to 100, with lower scores indicating less severe symptoms.

^‡The Stenosis Frequency Index ranges from 0 to 24, with lower scores indicating less severe symptoms.

^§The Stenosis Bothersomeness Index ranges from 0 to 24, with lower scores indicating less severe symptoms.

Main Treatment effects

Intent-to-Treat Analysis

In the intent-to-treat analysis of the randomized cohort, all measures over 4 years favored surgery but there were no statistically significant treatment effects in any of the primary outcome measures at any time interval (Table 4 and Figure 2). The secondary outcomes (sciatica bothersomeness index and self-rated improvement) were statistically significant in favor of surgery in the intent-to-treat analysis at 1 year;⁴ significance was maintained out to 4 years only for the sciatica bothersomeness index (Table 4 and Figure 3).

Table 4.

Primary analysis results for years 3 and 4. Intent-to-treat for the randomized cohort and adjusted^* analyses according to treatment received for the randomized and observational cohorts combined.^‡‡

	Baseline	2-Year			3-Year			4-Year
	Overall Mean	Mean Change (SE) or percent			Mean Change (SE) or percent			Mean Change (SE) or percent
		Surgery	Non- Operative	Treatment Effect (95% CI)†	Surgery	Non- Operative	Treatment Effect (95% CI)†	Surgery	Non- Operative	Treatment Effect (95% CI)†
RCT Intent-to-treat
Primary Outcomes		(n = 187)	(n = 191)		(n = 180)	(n = 170)		(n = 149)	(n = 150)
SF-36 Bodily Pain (BP) (0–100) (SE)‡	26.9 (0.82)	40.5 (1.9)	37.5 (1.9)	3.1 (−2.2, 8.4)	39.6 (2)	36.2 (2)	3.4 (−2.1, 8.9)	41.3 (2.1)	36.8 (2.1)	4.5 (−1.2, 10.3)
SF-36 Physical Function (PF) (0–100) (SE)‡	39.5 (1.2)	36.2 (2)	35.7 (2)	0.5 (−4.9, 6)	37.2 (2)	34.1 (2)	3.1 (−2.5, 8.8)	36.6 (2.1)	34.4 (2.1)	2.2 (−3.7, 8)
Oswestry Disability Index (ODI) (0–100) (SE)§	46.9 (0.97)	−31.5 (1.7)	−28.8 (1.7)	−2.8 (−7.5, 1.9)	−31.5 (1.7)	−27.2 (1.7)	−4.3 (−9.1, 0.6)	−31.2 (1.8)	−27.6 (1.8)	−3.6 (−8.6, 1.4)
Secondary Outcomes
Sciatica Bothersomeness Index (0–24) (SE)¶	15.2 (0.24)	−10.2 (0.47)	−8.6 (0.47)	−1.6 (−2.9, −0.3)	−10.2 (0.48)	−8.4 (0.49)	−1.8 (−3.2, −0.5)	−10.5 (0.51)	−8.8 (0.5)	−1.8 (−3.2, −0.4)
Leg pain (0–6) (SE)||	4.6 (0.1)	−3.3 (0.2)	−2.8 (0.2)	−0.5 (−0.9, −0.1)	−3.4 (0.2)	−2.8 (0.2)	−0.6 (−1.1, −0.2)	−3.5 (0.2)	−2.9 (0.2)	−0.6 (−1.1, −0.2)
Low back pain bothersomeness (0–6) (SE)**	3.9 (0.1)	−1.9 (0.2)	−1.8 (0.2)	−0.1 (−0.6, 0.3)	−1.9 (0.2)	−1.6 (0.2)	−0.3 (−0.7, 0.2)	−1.8 (0.2)	−1.7 (0.2)	−0.1 (−0.6, 0.3)
Very/somewhat satisfied w/symptoms (%)	3.4 (1.8)	68.3	64.9	3.3 (−6.3, 13)	72.9	62.4	10.5 (0.5, 20.6)	64.7	61.3	3.4 (−7.7, 14.6)
Very/somewhat satisfied w/care(%)		87	84.5	2.4 (−4.8, 9.6)	85.2	85	0.2 (−7.4, 7.8)	89.3	81.2	8.1 (−0.1, 16.2)
Self-rated progress: major improvement (%)		75.7	68.5	7.2 (−2, 16.3)	72.1	66.8	5.3 (−4.6, 15.2)	72.5	65	7.5 (−3.2, 18.1)
Work status: working (%)	64.3 (4.8)	75	75.5	−0.6 (−9.1, 8)	71	74.5	−3.6 (−12.7, 5.6)	71.4	75.1	−3.8 (−13.3, 5.8)
RCT/OC As-treated
Primary Outcomes		(n = 662)	(n = 344)		(n = 581)	(n = 318)		(n = 511)	(n = 283)
SF-36 Bodily Pain (BP) (0–100) (SE)‡	26.6 (0.5)	43 (0.87)	31.6 (1.2)	11.4 (8.5, 14.2)	43.8 (0.9)	31.8 (1.2)	12 (9, 15)	45.6 (0.95)	30.7 (1.3)	15 (11.8, 18.1)
SF-36 Physical Function (PF) (0–100) (SE)‡	37.9 (0.7)	43.6 (0.82)	30.2 (1.1)	13.4 (10.7, 16)	44.1 (0.85)	30.5 (1.2)	13.6 (10.8, 16.4)	44.6 (0.89)	29.7 (1.2)	14.9 (12, 17.8)
Oswestry Disability Index (ODI) (0–100) (SE)§	49.3 (0.6)	−37.2 (0.68)	−24.6 (0.91)	−12.5 (−14.7, −10.4)	−37.5 (0.71)	−25 (0.96)	−12.5 (−14.8, −10.2)	−38.1 (0.74)	−24.9 (1)	−13.2 (−15.6, −10.9)
Secondary Outcomes
Sciatica Bothersomeness Index (0–24) (SE)¶	15.5 (0.1)	−10.9 (0.21)	−8 (0.3)	−2.8 (−3.6, −2.1)	−11.1 (0.22)	−8.3 (0.32)	−2.8 (−3.6, −2)	−11.5 (0.23)	−8.2 (0.33)	−3.3 (−4.1, −2.5)
Leg pain (0–6) (SE)||	4.7 (0)	−3.5 (0.1)	−2.7 (0.1)	−0.8 (−1, −0.5)	−3.6 (0.1)	−2.9 (0.1)	−0.7 (−1, −0.5)	−3.7 (0.1)	−2.8 (0.1)	−0.9 (−1.2, −0.6)
Low back pain bothersomeness (0–6) (SE)**	3.9 (0)	−2.1 (0.1)	−1.5 (0.1)	−0.7 (−0.9, −0.4)	−2.2 (0.1)	−1.5 (0.1)	−0.7 (−1, −0.5)	−2.2 (0.1)	−1.4 (0.1)	−0.8 (−1, −0.5)
Very/somewhat satisfied w/symptoms (%)	4.1 (2)	75	48.8	26.3 (19.4, 33.1)	74.3	49.9	24.4 (17.1, 31.6)	76.6	46.7	29.9 (22.4, 37.4)
Very/somewhat satisfied w/care(%)		91.7	79.3	12.5 (7.1, 17.8)	90.2	74.6	15.6 (9.4, 21.7)	91.7	77.3	14.4 (8.2, 20.7)
Self-rated progress: major improvement (%)		78.3	58.7	19.7 (12.8, 26.5)	75.9	53.9	22 (14.6, 29.3)	79.2	51.7	27.5 (19.9, 35)
Work status: working (%)	73.6 (4.4)	85.4	85.1	0.4 (−4.8, 5.5)	83.9	79.7	4.2 (−2.1, 10.5)	84.4	78.4	6 (−0.9, 12.9)

^*Adjusted for age, gender, marital status, smoking status, race, compensation, herniation location, working status, stomach comorbidity, depression, other*** comorbidity, self-rated health trend, duration of most recent episode, treatment preference and baseline score (for SF-36 and ODI), and center.

^†Treatment effect is the difference between the surgical and non-operative mean change from baseline.

^‡The SF-36 scores range from 0 to 100, with higher score indicating less severe symptoms.

^§The Oswestry Disability Index ranges from 0 to 100, with lower scores indicating less severe symptoms

^¶The Stenosis Bothersomeness Index ranges from 0 to 24, with lower scores indicating less severe symptoms

^||The Leg Pain Bothersomeness Scale ranges from 0 to 6, with lower scores indicating less severe symptoms

^**The Low Back Pain Bothersomeness Scale ranges from 0 to 6, with lower scores indicating less severe symptoms

^††The sample sizes for the as-treated analyses reflect the number of patients contributing to the estimate in a given time-period using the longitudinal modeling strategy explained in the methods section, and may not correspond to the counts provided for each visit time in Figure 1.

^***”Other comorbidities include: stroke, diabetes, osteoporosis, cancer, fibromyalgia, cfs, PTSD, alcohol, drug dependency, heart, lung, liver, kidney, blood vessel, nervous system, hypertension, migraine, anxiety, stomach, bowel

^‡‡The estimates for 1Y and 2Y for IDH RCT ITT differ slightly from those presented in JAMA paper (reference) due to modeling differences.

Figure 2. Primary Outcomes (SF-36 Bodily Pain and Physical Function, and Oswestry Disability Index) in the Randomized and Observational Cohorts during 2 Years of Follow-up.

The graphs show both the intent-to-treat and the as-treated analyses for the randomized cohort (column on the left) and the as-treated analysis for the observation cohort (column on the right). The horizontal dashed line in each of the 4 SF-36 graphics represents normal values adjusted for age and sex. The I bars represent 95% confidence intervals. At 0 months, the floating data points represent the observed mean scores for each study group, whereas the data points on plot lines represent the overall means used in the adjusted analyses.

Figure 3.

Secondary Outcomes in the Randomized and Observational Cohorts during 2 Years of Follow-up

As-Treated Analysis

The global hypothesis test (not shown) comparing the as-treated treatment effects between the randomized and observational cohorts over all time periods showed no difference between the randomized and observational cohorts (p = 0.44 for SF-36 BP, p = 0.76 for SF-36 PF, and p= 0.90 for the ODI). Treatment effects for the primary outcomes in the combined as-treated analysis were significant at 2 years and maintained out to 4 years: SF-36 BP 15.0 p<0.001 (95% CI 11.8 to 18.1); SF-36 PF 14.9 p<0.001 (95% CI 12.0 to 17.8); ODI -13.2 p<0.001 (95% CI −15.6 to −10.9. (Table 4) The footnote for Table 4 describes the controlling covariates for the final model.

Results from the intent-to-treat and as-treated analyses of the two cohorts are compared in Figure 2. The as-treated treatment effects significantly favored surgery in both cohorts. In the combined analysis, treatment effects were statistically significant in favor of surgery for all primary and secondary outcome measures (with the exception of work status) at each time point (Table 4 and Figure 3).

The treatment effects for the secondary measures of sciatica bothersomeness, satisfaction, and self-rated improvement narrowed between 3 months and 2 years but remained significant at all periods. Work status was significantly worse in the surgery group at 3 months due to surgery patients recovering from surgery; work status thereafter showed a small but non-significant benefit for surgery. At 4 years, the adjusted percentage of patients working was 84.4% surgical vs. 78.4% non-operative, treatment effect 6.0 (95% CI −0.9, 12.9). (Table 4 and Figure 3)

DISCUSSION

In patients with a herniated disc confirmed by imaging and leg symptoms persisting for at least six weeks, surgery was superior to non-operative treatment in relieving symptoms and improving function. In the as-treated analysis, the treatment effect for surgery was seen as early as 6 weeks, appeared to reach a maximum by 6 months and persisted over 4 years; it is notable that the non-operative group improved significantly and this improvement persisted throughout the 4-year period. Under reasonable assumptions, this mixing of treatments due to crossover can be expected to create a bias toward the null in the intent-to-treat analyses.^2,17 The large effects seen in the as-treated analysis after adjustments for characteristics of the crossover patients suggest that the intent-to-treat analysis underestimates the true effect of surgery.

The SPORT data are supported by the only other recent randomized trial for disc herniation, Peul et al.² In that study, 39% of those randomized to non-operative treatment crossed over to surgery at about five months. This is nearly identical to SPORT, in which 38% had crossed into surgery by 6 months. The estimated improvements 1 year after surgery in these two studies were similar (Peul vs. SPORT): SF-36 BP 59.3 vs. 43.7; SF-36 PF 50.3 vs. 44.4; and Sciatica Bothersomeness −11.5 vs. −11.2. In addition, the differences at 1 year between randomized groups in the intent-to-treat analyses were also quite similar: SF-36 BP 2.7 vs. 3.6; SF-36 PF 2.2 vs. 2.0; and Sciatica Bothersomeness −0.4 vs. −1.9. These results further validate the SPORT randomized cohort results but again highlight the need to also consider the as-treated analysis in this study population to estimate the true effect of surgery and to avoid bias towards the null.

SPORT Randomized vs. Observational Cohorts

Debate continues in the scientific literature regarding the optimal role of observational studies versus randomized trials.^18,19 The design of SPORT and its long-term follow-up provides an opportunity to compare over time the randomized trial results to its simultaneously enrolled observational cohort.^3–5,20,21 These two cohorts were remarkably similar at baseline. Patients in the observational cohort were relatively more symptomatic and functionally impaired than those in the randomized cohort; however, the absolute differences were quite small: 4.3 points on the ODI, 2.9 points on SF-36 PF, and 0.6 on the sciatica bothersomeness index. Given these similarities and our formal comparison of treatment effect between cohorts, the combined analyses are well justified.

Comparisons to Other Studies

There are no long-term studies reporting the same primary outcome measures as SPORT. As demonstrated above, the results of SPORT primary outcomes at 1 year and Peul et al secondary measures are quite similar, but longer follow up for the Peul study is necessary for further comparison. Unlike the suggestions made in the Weber study, the differences in the outcomes between treatment groups remained relatively constant after the first year, suggesting the importance of ongoing follow-up of the patients in SPORT (Table 4, Figures 2 and 3).

The results of SPORT are similar to the Maine Lumbar Spine Study (MLSS) ²² and the Weber study. ²³ The MLSS reported somewhat larger adjusted treatment effect differences at 5 years between patients who had received surgery within 3 months versus those that had not when compared to the SPORT 4-year data: −7.1 vs. − 3.3 (sciatica bothersomeness), −2.0 vs. −0.9 (leg pain in the past week) and −1.2 vs. −0.8 (low back pain in the past week). The differences are mainly related to greater improvements in the non-operative group in SPORT vs. MLSS; that is, the sciatica bothersomeness non-operative mean change from baseline for SPORT was −8.2 at 4 years vs. MLSS −4.6 at 5 years. While there are no validated outcome measures that can be directly compared between SPORT and the Weber study, their 4 year results of 70% more patients in the surgical group and 51% in the conservative treatment group with “Good” results is similar to SPORT’s 79.2% self-rated major improvement in the surgery group and 51.7% in the non-operative group at 4 years.

Given the increasing rates of disability and related cost for back conditions worldwide,²⁴ work status is thought to be an important measure of success in this population. However, return to work appears to be independent of treatment received and does not follow improvement in pain, function or satisfaction with treatment (percent working at 4 years: surgical 84.4%, non operative 78.4%; 6.0 treatment effect (95% CI, −0.9 to 12.9) (Table 4 and Figure 3-Panel E). The MLSS also showed no significant effect of surgery on return to work. From varied perspectives--payer, provider and patient--setting expectations regarding treatment and the issue of work status must be considered in light of these results. A formal cost-effectiveness analysis of SPORT disc herniation patients with workers compensation is underway.²⁵ We have performed a formal cost effectiveness analysis of the combined SPORT disc herniation cohorts over 2 years. ²⁶ It revealed that the cost per Quality-Adjusted-Life Year (QALY) gained for surgery relative to non-operative care was $69,403 (95% CI: $49,523, $94,999), concluding that surgery is cost-effective.

Over the 4 years there was little evidence of harm from either treatment. The 4-year rate of re-operation was 10%, which is lower than the 19.4% reported by MLSS at five years. ²²

Limitations

Although our results are adjusted for characteristics of cross over patients and control for important baseline covariates, the as-treated analyses presented do not share the strong protection from confounding that exists for an intent-to-treat analysis.^2–4 However, the as-treated analyses yielded results similar to prior studies and to a well designed randomized trial by Peul et al. Another limitation is the heterogeneity of the non-operative treatment interventions, discussed in our prior papers.^3,4

Conclusions

In the as-treated analysis combining the randomized and observational cohorts, which carefully controlled for potentially confounding baseline factors, patients treated surgically for intervertebral disc herniation showed significantly greater improvement in pain, function, satisfaction, and self-rated progress over 4 years compared to patients treated non-operatively. It appears, despite the improvement in symptoms, that except for the first 6 weeks after surgery, work status is not related to treatment.

KEY POINTS

• At 4 year follow-up, patients who had surgery for intervertebral disc herniation maintained greater improvement in all primary outcomes compared to those who remained non-operative based on as-treated analyses;

• Except for work status, all secondary measures retained a significant benefit for surgery at 4 years;

• Work status showed a non-significant benefit for surgery at 4 years.