Article Text

Comparative effectiveness of anesthetic technique on outcomes after lumbar spine surgery: a retrospective propensity score-matched analysis of the National Surgical Quality Improvement Program, 2009–2019
  1. Krizia Amoroso1,
  2. Ichiro Okano2,
  3. Michele Sarin1,
  4. Alexander P Hughes2,
  5. William D Zelenty2,
  6. Jennifer Shue2,
  7. Andrew A Sama2,
  8. Frank P Cammisa2,
  9. Federico P Girardi2 and
  10. Ellen M Soffin3
  1. 1 Hospital for Special Surgery, New York, New York, USA
  2. 2 Orthopaedic Surgery, Spine Care Institute, Hospital for Special Surgery, New York, New York, USA
  3. 3 Department of Anesthesiology, Critical Care and Pain Management, Hospital for Special Surgery, New York, New York, USA
  1. Correspondence to Dr Ellen M Soffin, Department of Anesthesiology, Critical Care and Pain Management, Hospital for Special Surgery, New York, NY 10021, USA; soffine{at}


Background The impact of anesthetic technique on spine surgery outcomes is controversial. Using a large national sample of patients, we compared outcomes after lumbar decompression under regional anesthesia (RA: spinal or epidural) or general anesthesia (GA).

Methods A retrospective population-based study of American College of Surgeons National Surgical Quality Improvement Program data (2009–2019). Patients were propensity score (PS) matched 3:1 (GA:RA) on demographic and surgical variables. The primary outcome was the association between anesthetic type and any complication (cardiac, pulmonary, renal, transfusion, stroke, infectious, deep vein thrombosis/pulmonary embolus). Secondary outcomes included the association between anesthetic type and individual complications, readmission and length of stay (LOS). Unadjusted comparisons (OR, 95% CI), logistic regression and adjusted generalized linear modeling (parameter estimate, PE, 95% CI) were performed before and after PS matching.

Results Of 1 51 010 cases, 149 996 (99.3%) were performed under GA, and 1014 (0.67%) under RA. After matching, 3042 patients with GA were compared with 1014 patients with RA. On unadjusted analyses, RA was associated with lower odds of complications (OR 0.43, 0.3 to 0.6, p<0.001), shorter LOS (RA: 1.1±3.8 days vs GA: 1.3±3.0 days; p<0.001) and fewer blood transfusions (RA: 3/1014, 0.3% vs GA: 40/3042, 1.3%; p=0.004). In adjusted analyses, RA was associated with fewer complications (PE −0.43, –0.81 to −0.06, p=0.02) and shorter LOS (PE −0.76, –0.90 to −0.63, p<0.001). There was no significant association between anesthetic type and readmission (PE −0.34, –0.74 to 0.05, p=0.09).

Conclusions Compared with GA, RA was associated with fewer complications, less blood transfusion and shorter LOS after spine surgery. Although statistically significant, the magnitude of effects was small and requires further prospective study.

  • Postoperative Complications

Data availability statement

Data are available upon reasonable request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Compared with general anesthesia, regional anesthesia improves outcomes after orthopedic surgery. It is unclear whether these benefits are present among patients undergoing spine surgery.


  • Regional anesthesia was associated with fewer complications after lumbar spine surgery. There was a lower incidence of blood transfusion and shorter length of stay among patients who received regional anesthesia, but no other differences in individual outcomes or complications between the two techniques.


  • These results suggest that regional anesthesia may be a modifiable factor to improve outcomes after lumbar spine surgery. However, given the rare use of regional anesthesia for spine surgery, more prospective research into technique-specific risks and benefits should be undertaken prior to widespread recommendations for use.


A long history and large volume of research supports regional anesthesia (RA) as equivalent or superior to general anesthesia (GA) to improve outcomes after orthopedic surgery.1 In contrast, the effect of anesthetic technique on outcomes after spine surgery is relatively understudied,2 with conflicting evidence to conclude the superiority of one technique over the other for individual outcomes, including bleeding and transfusion, hemodynamic stability and early postoperative pain scores.3–6 Three systematic reviews with meta-analysis on the topic urge caution in interpretation of the potential benefits of RA for spine surgery given the small number of prospective comparative studies with small sample sizes, significant risk of bias and high heterogeneity of included trials.7–9 Population-based research overcomes several of these methodologic limitations, but similar to the clinical studies, comparative research into GA versus RA in spine surgery cohorts is not yet established.10 11

In the current study, we used population-based data derived from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) data sets to compare the association of anesthetic type and outcomes after lumbar spine surgery, between 2009 and 2019. The primary outcome of interest was the association between anesthetic type and any complication within 30 days of surgery. We hypothesized that patients undergoing elective lumbar spine surgery under RA would comprise a minority of patients but that outcomes would be at least equivalent to patients who received GA.


The study was performed in accordance with the ethical standards of the Declaration of Helsinki (1964) and its subsequent amendments, under exempt status granted by the Institutional Review Board at Hospital for Special Surgery. The study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.12

Study population

Clinical information was extracted for patients who underwent lumbar spine surgery between 2009 and 2019 from the ACS-NSQIP data sets.13 ACS-NSQIP Participant Use Data Files were queried using current procedural terminology (CPT) codes for lumbar decompression, laminectomy, laminotomy, foraminotomy and combined procedures (22102, 62380, 63005, 63011, 63012, 63017, 63030, 63047, 63056, 63035, 63048, 63 057 and 0275T) (online supplemental table 1). All patients who received GA or RA (spinal anesthesia; epidural anesthesia) were included. Exclusion criteria were surgery for trauma, fracture, neoplasm, infectious disease; patients ≤age 18; and other types of anesthesia (not recorded; combined spinal-epidural; monitored anesthesia care).

Supplemental material

Baseline characteristics collected included: age, sex, race, body mass index (BMI), smoking status, American Society of Anesthesiologists (ASA) physical status classification, functional status (dependent, independent), comorbidities (diabetes mellitus, chronic obstructive pulmonary disease, heart failure, hypertension, dialysis, cancer and bleeding disorders) and surgical duration.


The primary outcome was complications within 30 days of surgery. Complications were defined in the NSQIP Participant Use Data Files13 and included: myocardial infarction, cardiac arrest, pneumonia, unplanned intubation, failure to wean from ventilator, pulmonary embolus, deep vein thrombosis, progressive renal insufficiency, acute renal failure, stroke, blood transfusion, wound dehiscence, urinary tract infection, infection (superficial wound, deep wound or organ space), sepsis, septic shock. Secondary outcomes included the differences between anesthetic type and (1) individual complications, (2) discharge destination (home or non-home), (3) length of stay (LOS), (4) readmission, (5) reoperation and (6) 30-day mortality.

Statistical analysis

Propensity score matching and covariate adjustment

Due to the observational (non-randomized) nature of the data set, we used propensity score (PS) matching to minimize the effects of confounding factors when assessing differences in patient demographics and outcomes between procedures performed with GA and RA. Three-to-one greedy nearest-neighbor PS matching was performed using anesthesia type as the treatment indicator to match patients with GA to patients with RA. The probability of undergoing a lumbar spinal procedure with GA was calculated based on clinically relevant covariates, including age, sex, BMI, ASA classification and procedure. Matching was considered sufficient if the standardized mean difference (SMD) in the means of PSs were less than 0.1.14

The SMD was calculated for comparisons of variables between anesthetic type for both the unmatched and matched cohorts. To assess postoperative factors, Fisher’s exact test was used for dichotomous and categorical variables; the Mann-Whitney (Wilcoxon rank-sum) test was used for continuous variables. For the matched cohort, unadjusted and adjusted logistic regression was used to assess the occurrence of any complication, readmission and length of hospital stay between the anesthetic types. Conditional logistic regression was performed with anesthesia group as the independent variable and ‘readmission’ and ‘complication’ as separate discrete outcome variables. Generalized linear models were used to measure the association between anesthesia type and ‘length of stay’ (LOS), ‘readmission’ and ‘complication’, with anesthesia group as a covariate and PS as a confounding factor. A log transformation was used to normalize LOS. For LOS=0, the small value of 0.01 was added in order to compute the natural log.

All statistical analyses were conducted using SAS software V.9.4 (SAS Institute, Cary, North Carolina) and significance was set at p<0.05 for all tests.


We identified 1 53 224 patients who underwent the included surgeries between 2009 and 2019. After excluding cases for ineligible CPT code (n=2083) or primary anesthetic type (n=131), 1 51 010 cases were available for analysis. Of these, 149 996 (99.3%) were performed under GA, and 1014 (0.67%) were performed under RA. There were 171 patients included in the GA group who also received spinal or epidural anesthesia (0.1%).

Table 1 shows the patient-related and surgical-related demographic variables by anesthesia type, before and after matching. For the unmatched cohort, significant differences were found in the SMD for age, race, BMI, baseline comorbidities, duration of surgery and number of lumbar levels operated.

Table 1

Patient demographics, preoperative and operative variables

After PS matching, 3042 (75%) patients with GA were compared with 1014 (25%) patients with RA (table 1). None of the variables used to PS match (age, sex, BMI, outpatient procedure and ASA class) was statistically significantly different after matching. Demographic and perioperative variables were well balanced, with the exception of mean operative time (18 min longer for GA, SMD 0.317) and race (fivefold fewer African American patients received RA; SMD 0.109).

Postoperative complications and discharge variables for the matched and unmatched cohorts are presented in table 2. The incidence of any complication was significantly higher among patients who received GA (n=8602/149,996, 5.7%) compared with RA (n=25/1014, 2.5%; p<0.001) on unmatched comparisons, but not after PS matching (GA:110/3042, 3.6%; RA: 25/1014, 2.5%; p=0.085). On unmatched comparisons, the incidence of deep vein thrombosis (0.4% vs 0; p=0.017), reoperation (2.5% vs 1.4%, p<0.03) and readmission (4% vs 2.5%, p=0.002) was all statistically higher in the GA versus RA cohorts, respectively, but not after PS matching.

Table 2

Postoperative complications, length of stay and discharge destination

The incidence of blood transfusion was significantly higher in the GA cohort, compared with RA in both the unmatched (3286/149,996, 2.2% vs 3/1014, 0.3%; p<0.0001) and PS-matched comparisons (40/3042, 1.3% vs 3/1014, 0.3%; p=0.004). Length of hospital stay was significantly longer among patients who received GA in both the unmatched (1.9±3.4 days vs 1.1±3.8 days; p<0.001) and PS-matched comparisons (1.3±3.0 days vs 1.1±3.8 days; p<0.001). There were no significant differences in other complications between the groups on either the unadjusted or PS-matched comparisons.

In unadjusted analysis, RA was associated with significantly lower odds of readmission (OR:0.57, 95% CI 0.38 to 0.84, p=0.005), any complication (OR:0.43, 95% CI 0.30 to 0.62, p<0.0001) and shorter length of hospital stay (−1.5, 95% CI −1.7 to −1.4, p<0.001). In adjusted analysis using the PS as a covariate, RA was associated with fewer complications (parameter estimate: −0.43, 95% CI −0.81 to −0.06, p=0.02) and shorter length of hospital stay (parameter estimate: −0.76, 95% CI −0.90 to −0.63, p<0.001). There was no statistically significant association between anesthetic type and readmission (parameter estimate: −0.34, 95% CI −0.74 to 0.05, p = 0.09) (table 3).

Table 3

Sensitivity and adjusted analyses


In this retrospective population-based analysis, we found that compared with GA, RA was associated with fewer complications, less blood transfusion and shorter LOS after lumbar spine surgery. Although statistically significant, the absolute magnitude of these effects was small, and of uncertain clinical significance.

Clinical and population-based studies which investigate the impact of anesthetic technique on outcomes after lumbar spine surgery are sparse and results are conflicting. There are three systematic reviews with meta-analysis of prospective clinical trials comparing outcomes between RA and GA for lumbar spine surgery. The earliest included eight studies (625 patients) and found that patients who received RA required less analgesia in postanesthesia care unit (PACU) and had less nausea and vomiting (PONV) over the first postoperative day but found no differences in intraoperative hypotension, bradycardia, blood loss or surgical duration between RA and GA.7 In the second, Zorrilla-Vaca et al included 15 studies (961 patients) and found several benefits of RA compared with GA, including lower incidence of PONV, shorter LOS and less blood loss. However, they did not find effects of anesthetic technique on early pain scores, analgesic requirements or incidence of urinary retention.8 The most recent analysis included 11 studies (896 patients) and concluded that RA did decrease early postoperative pain and analgesic requirements in the PACU, confirmed the reduction in PONV, and did not find effects on urinary retention, intraoperative bradycardia or hypotension.9 Importantly, each of these reviews found that the quality of informing trials was overall low with high or very high risk of bias. This may be explained in part by the small number of studies conducted, to date, and also to significant issues of blinding in any study which compares RA to GA.

Data derived from population-based studies may help to overcome some of the methodological and blinding limitations associated with the prospective studies of comparative anesthetic technique. Additionally, the large number of available patient samples should theoretically aid detection of rare events and complications after a relatively safe surgery, such as lumbar decompression/fusion. Unfortunately, as for the clinical studies, population-based data on the topic are limited, and conflicting. The first of two available analyses derived from ACS-NSQIP compared outcomes after lumbar spine surgery among patients who received general or non-GA.10 The authors concluded equivalent outcomes for LOS, readmission and overall complications between the techniques. The second study reported trends in use of GA and non-GA for all types of spine surgery (lumbar, thoracic and/or cervical) between 2005 and 2019.11 The authors found progressive utilization of RA over time for all subtypes of spine surgery as well as shorter LOS and lower readmission and 30-day complications among patients undergoing lumbar surgery. Discrepancies between the results of the former and present studies may be explained by our attempts to control for the retrospective, observational nature of our design by using PS matching.

Despite our efforts to rigorously PS match patients between the two techniques, we noted a small persistent imbalance in surgical duration between the groups. One interpretation of this is that medically and surgically complex patients were more likely to be offered GA, leading to increased operating time in this group. Although possible, this explanation is less likely when considering that we found no differences in other indices of medical complexity (including BMI, ASA class and age) between the groups. Interestingly, we did find a lower incidence of blood transfusion between RA and GA, both before and after PS matching. Although speculative, lower intraoperative blood losses may have facilitated operating conditions, eliminated the time otherwise required to complete intraoperative transfusion and led directly to shorter surgical duration.

Effects of anesthetic technique on blood loss and transfusion after spine surgery have yielded conflicting results. Most3 4 6 8 but not all5 7 10 prior studies concluded that RA was associated with less blood loss compared with GA for spine surgery, possibly due to sympathectomy and consequent intraoperative hypotension. It is possible that other improvements in perioperative blood management could account for differences in the incidence of transfusion between the anesthetic techniques found here, specifically, use of tranexamic acid.15 16 This was not controlled for in the present analysis due to limitations in data recording and extraction from the NSQIP datasets.

There are several unique risks of RA for spine surgery, which should be considered prior to final selection of anesthetic technique. Important considerations include prone positioning in the obese patient and patients with difficult airways; plans for unanticipated lengthy surgical duration, complication, suboptimal spinal anesthetic or patient intolerance of sedation/positioning; risk of postdural puncture headache, epidural–spinal hematoma or neurologic injury/neuropraxia; and what to do in the event of motor weakness/sensory loss in the PACU beyond the expected duration of the spinal anesthetic. Careful patient selection, planning and consensus between anesthetic and surgical teams are likely to be key to successful outcomes using RA. Provocatively, reports from the surgical literature describing ‘awake spine surgery’ are emerging, and, in contrast to the present results, are suggested to offer profound benefits for patient satisfaction and other important outcomes—despite mainly comprizing case series, small cohort studies and/or institutional experiences.17–19 Furthermore, these reports tend to extoll the benefits of spinal anesthesia in combination with regional analgesia techniques—which likewise have not been fully investigated or associated with significant benefits on recovery after lumbar spine surgery.20 21 ‘Awake spine surgery’ reports tend to be from surgeon-led teams and lack important definitions of terms (crucially, what it means to be ‘awake’) details (of sedation, if given, complications and conversion to GA) methods for patient selection22 and assume that patients who receive RA for spine surgery are, defacto, also awake.2 11 These differences may account (at least in part) for differences in outcomes reported in the ‘awake spine surgery’ studies and those reported here.

Strengths and limitations

Our study includes data derived from a range of institutions and includes a large sample of patients, which we carefully matched for analyses. Accordingly, the sample is likely to be representative of the broader population. However, there are several limitations to our study. First, the work is based on a US data set and may not be generalizable to other patient populations or practice settings. Second, as for any study using ACS-NSQIP, coding errors could lead to misclassification of patients, surgeries and anesthetic type. The latter is particularly relevant for the current study, given the rarity with which RA is performed for spine surgery. Allied to this, we included 171 patients in the GA group who had also received RA: whether this represented failed neuraxial anesthesia, a planned combined anesthetic regimen or coding error is unknown. Third, some clinically important data which may guide anesthetic selection and influence the outcomes are not available from ACS-NSQIP. Examples include details of the GA and sedation for RA (if provided); the depth of sedation of patients receiving RA; history of difficult intubation or coagulopathy and surgeon preferences for anesthetic choice.


This analysis of ACS-NSQIP data found few benefits of RA compared with GA on outcomes after lumbar spine surgery. RA was associated with fewer complications, lower incidence of blood transfusion and shorter LOS, but the magnitude of these effects was small and of uncertain clinical significance. Of note, we did not find any benefits of GA over RA for any individual complications or other outcomes of interest, suggesting either technique may be considered in appropriate patients. We caution in interpretation of these results, given the small proportion of patients who received RA, and paucity of prospective data on technique-specific risks and benefits for patients undergoing spine surgery. Subject to these caveats, the results presented here suggest that RA may be a modifiable factor to improve outcomes after lumbar spine surgery.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Hospital for Special Surgery Institutional Review Board, IRB#2021-1099. Informed consent was waived due to the retrospective nature of the work.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @ESoffin

  • Contributors EMS and APH conceived the idea. All authors contributed to the design of the work. KA, JS, MS acquired the data. IO and MS analyzed the data. All authors interpreted the data. EMS, KA, JS drafted the article. All authors revised the work critically for important intellectual content. All authors approved the version for publication. EMS acts as guarantor of the work, had access to the data, and controlled the decision to publish.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.