Abstract
-
Objective
While minimally invasive-transforaminal lumbar interbody fusion (MIS-TLIF) has shown superiority in key clinical metrics over the open approach, evidence regarding patient-reported outcomes remains limited. This study compared postoperative recovery trajectories and symptomatic improvement phases between MIS and open TLIF.
-
Methods
This retrospective review included patients who underwent single-level MIS or open TLIF. Oswestry Disability Index (ODI) and Numerical Rating Scale (NRS) for back and leg pain were collected preoperatively and postoperatively. Segmented regression analysis with mixed-effects modeling, allowing for identification of distinct recovery phases, compared symptomatic trends between approaches.
-
Results
Of 324 patients (268 MIS, 56 open), baseline demographics were similar except for greater preoperative leg pain in the MIS group (NRS: 6.0 vs. 5.0, p = 0.027). A segmented regression model identified 4 ODI recovery phases: postoperative disability phase (PDP, day 0 to 13), early improvement phase (day 13 to 28), late improvement phase (day 28 to 110), and plateau phase (later than day 110). The MIS group exhibited significantly lower disability exacerbation during PDP (β = 0.93 vs. 1.42 points per day, p = 0.008). Additionally, the plateau of NRS back occurred significantly earlier in the MIS group than in the open group (MIS, 26.7 ± 2.6 days vs. open, 51.7 ± 6.6 days, p < 0.001).
-
Conclusion
MIS-TLIF resulted in lower postoperative disability during the first 2 weeks compared to the open approach. Furthermore, low back pain achieved an earlier plateau in back pain by about 4 weeks in the MIS approach.
-
Keywords: Transforaminal lumbar interbody fusion, Minimally invasive spine surgery, Recovery trend, Temporal trend, Oswestry Disability Index, Numerical Rating Scale
INTRODUCTION
Transforaminal lumbar interbody fusion (TLIF) is a widely performed procedure for lumbar degenerative conditions, offering symptom relief through nerve decompression and spinal stabilization [
1,
2]. Over the past decade, the minimally invasive (MIS) approach to TLIF has gained increasing popularity due to its potential benefits over traditional open surgery, including reduced tissue disruption, shorter hospital stays, improved postoperative pain control, and improved cost effectiveness [
3-
5].
Despite these reported advantages, the effectiveness of MIS over open TLIF has often been demonstrated through indirect metrics such as reduced intraoperative blood loss, decreased narcotic use, and faster return to work [
6-
10]. In contrast, studies using patient-reported outcomes and measures (PROMs) frequently indicate similar improvements for both approaches at standard follow-up milestones (e.g., 1, 3, and 6 months) [
3,
9,
11]. A meta-analysis found no significant superiority of MIS-TLIF over open TLIF in terms of the Oswestry Disability Index (ODI). However, it did identify significant improvements in visual analogue scale back pain scores for MIS-TLIF, despite inconsistent results across studies with varying follow-up durations within 6 months [
3,
12]. This discrepancy may be attributed to variations in study designs, patient populations, and follow-up protocols. Relying on predefined intervals may also obscure differences in recovery trajectories, especially during the early postoperative period.
To better capture these nuanced recovery patterns, this study employs segmented regression analysis, a statistical method that allows for the identification of distinct phases of recovery by detecting significant changes in the slope of PROMs over time, thus revealing “turning points” in the postoperative symptomatic trajectory [
13,
14]. Specifically, we aim to (1) characterize TLIF recovery by dividing the postoperative timeline into distinct phases, and (2) compare the recovery trajectories of MIS and open TLIF. By detailing these phase-specific improvements, we aim to investigate differences that may not be fully appreciated in studies limited to discrete follow-up intervals. Ultimately, this approach could help refine clinical decision-making and provide clearer information for patients undergoing TLIF.
MATERIALS AND METHODS
1. Study Design and Patient Population
This retrospective study of a single-center multi-surgeon prospectively collected database was approved by the Institutional Review Board (IRB) of Hospital for Special Surgery (IRB No. 2018-1599). Patients who underwent primary single-level MIS or open TLIF between April 2017 and August 2024 were assessed for eligibility. Informed consent was obtained for database participation.
Inclusion criteria included all patients undergoing primary single-level TLIF. Exclusion criteria were as follows: (1) undergoing other types of interbody fusion, (2) preoperative ODI scores ≤ 20 to exclude patients with minimal functional disability who were less likely to show meaningful postoperative improvement, and (3) revision surgeries. The open approach was defined as procedures performed through a longitudinal midline incision, while MIS was characterized using the modified Wiltse approach, which involves a minimal incision size and avoids midline exposure. As the missing value was assumed to be missing at random, a minimum follow-up period was not set to ensure that all available data, including partial datasets, were included to evaluate trends in symptomatic changes [
15].
Data collection and management were conducted using REDCap (Research Electronic Data Capture), hosted at the Weill Cornell Medicine Clinical and Translational Science Center. This platform is supported by the National Center for Advancing Translational Science of the National Institutes of Health under award number UL1 TR002384 [
16,
17].
Patient demographic data were obtained from the electronic medical record, including age, sex, race, body mass index (BMI), Charlson Comorbidity Index (CCI), smoking status, and the presence of depressed mood.
Pre- and postoperative PROMs were assessed using the ODI, Numerical Rating Scale (NRS) for back pain (NRS back), and NRS for leg pain (NRS leg). PROMs were scheduled to collect prospectively at preoperative, 2 weeks, 6 weeks, 12 weeks, 6 months, and 1-year postoperative time points, along with the corresponding dates of data collection. The postoperative days for each questionnaire were calculated based on the date of questionnaire submission relative to the date of surgery. While data collection at specific times was scheduled via electronic forms or clinic visits, variations in submission dates due to clinic schedules and patient availability were accounted for by including the actual submission dates in the analysis. This approach ensured that the timing of assessments reflected real-world variability, enhancing the robustness and accuracy of the findings.
3. Statistical Analysis
Statistical analyses were performed using R (ver. 4.4.0, R Core Team (2024), Vienna, Austria). Continuous variables were reported as mean± standard deviation, while categorical variables were presented as frequency (percentage, %). Comparisons between the open and MIS group were conducted using unpaired t-tests for continuous variables and chi-square tests for categorical variables.
To analyze the complex and potentially nonlinear recovery trajectories after TLIF, we employed a combination of segmented regression analysis and mixed-effects modeling. Segmented regression allowed us to identify distinct phases of recovery by detecting significant changes in the rate of improvement over time. Improvement trends were quantified as slopes (the rate of improvement or decline in symptoms per day), and breakpoints were detected as postoperative day. Mixed-effects modeling was then used to account for individual patient variability in these recovery trajectories, providing a more accurate and nuanced understanding of postoperative recovery [
18,
19]. Based on prior studies, clinical experience and visual inspection of scatter plot of each PROM [
10,
20], we initially hypothesized 3 phases of recovery:
Postoperative disability phase (PDP): The period after surgery when the impact of surgical stress is most pronounced, lasting until the initial phase of recovery begins.
Improvement phase (IP): Symptom improvement following the peak of pain.
Plateau phase (PP): Minimal or no further symptomatic improvement, with slopes approaching zero.
Preliminary segmented regression analysis, conducted on the overall follow-up data, identified candidate breakpoints. The number of effective breakpoints were assessed with Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), indicating that the smaller is the more appropriate. Based on the suggested number from these metrics, the number and location of breakpoints were determined based on clinical reasoning and visual inspection of regression lines and scatter plots.
Subsequently, mixed-effects models were applied to refine these candidate breakpoints and estimate phase-specific slopes, while accounting for individual patient variability. For example, some patients may demonstrate a rapid initial recovery followed by an early plateau, whereas others might improve more gradually across several months. Without random effects, the analysis would assume uniform trajectories, potentially masking these important individualized patterns in symptomatic improvement.
Segmented regression analysis results were reported as follows: phase-specific slopes with standard errors (β± SE points/ day), representing the rate of change in scores per day; intercepts, representing the estimated mean score at the beginning of each phase; and breakpoints, representing the postoperative day at which a significant change in slope occurred, reported with standard errors where applicable. Coefficient of determination (R²) was reported by marginal (fixed-effects alone) and conditional R² (both fixed and random effects). Statistical significance was defined as a p< 0.05.
RESULTS
1. Patient Demographic Data
A total of 403 patients who underwent MIS or open TLIF were screened for eligibility. Seventy-nine patients were excluded due to preoperative ODI scores below 20, leaving 324 patients (268 MIS and 56 open) included in the analysis. There were no significant differences between groups in age (60.0 years vs. 57.6 years, p= 0.24), male proportion (51.9% vs. 46.4%, p= 0.55), BMI (27.9 kg/m
2 vs. 27.7 kg/m
2, p= 0.83), smoking status (current or previous history, 34.0% vs. 39.3%, p = 0.54), and CCI (p= 0.13). Preoperative PROMs were also comparable in ODI and NRS back; however, the MIS group reported significantly worse leg pain on NRS leg (6.0 ± 2.9 vs. 5.0 ± 3.0, p = 0.027) (
Table 1).
1) Oswestry Disability Index
The assessment of the number of breakpoints indicated that 3 provided the best fit to the model while balancing complexity (AIC: one, 11,685.71; two, 11,686.8; three, 11,672.45; four, 11,675.3; five, 11,675.3. BIC: one, 11,722.3; two, 11,723.3; three, 11,719.5; four, 11,743.2; five, 11,743.2). Preliminary segmented regression analysis for ODI identified three significant breakpoints at days 11.0 ± 2.7, 29.5 ± 4.8, and 76.4 ± 11.7, defining four distinct phases: PDP (up to day 11), early improvement phase (EIP, days 12–29), late improvement phase (LIP, days 30–76), and PP (day 77 and beyond). The overall cohort’s improvement trends (slopes) were 0.72± 0.34 during PDP, -0.89± 0.41 during EIP, -0.22± 0.24 during LIP, and -0.02± 0.06 during PP. This characterized EIP as steep improvement, while LIP was relatively slower improvement (
Fig. 1).
2) NRS back
AIC and BIC suggested one breakpoint is the best number for NRS back (AIC: one, 6,319.2; two, 6,320.4; three, 6,321.2; four, 6,322.9; five, 6,322.9. BIC: one, 6,345.4; two, 6,356.9; three, 6,368.2; four, 6,390.8; five, 6,390.8). The initial segmented regression analysis for NRS back identified a breakpoint at day 27.2± 2.6, defining 2 distinct phases: IP and PP (
Fig. 2).
3) NRS leg
AIC values were: one, 6,533.4; two, 6,519.9; three, 6,524.1; four, 6,523.9; five, 6,527.5. BIC values were: one, 6,559.5; two, 6,556.4; three, 6,571.0; four, 6,581.3; five, 6,595.3. The model with 2 breakpoints indicated β= -0.23 up to day 12.3 ± 2.6, then β= -0.02 until day 79.0± 15.7, defining EIP, LIP, and PP phases (
Fig. 3).
1) PDP to EIP (day 0–30)
A segmented regression model with random effects showed that the open group exhibited a significantly worse trend in ODI (MIS: β= 0.93 vs. open: β= 1.42, p= 0.008) with a breakpoint at 13.4± 3.0 days, whereas preoperative ODI values were comparable between groups (intercept, 41.9 vs. 41.6, p= 0.92) (
Table 2).
2) EIP to LIP (day 10–80)
The open group had significantly worse starting ODI values at day 10 (MIS: 49.3 vs. open: 60.5, p= 0.003). However, the improvement trend (slope) during this phase did not differ significantly between the groups (MIS: β= -1.07 vs. open: β= -1.21, p= 0.08) until the breakpoint at day 28.4± 5.2 (
Table 3).
3) LIP to PP (after day 30)
The model for LIP to PP revealed no significant differences in starting ODI at day 30 or slopes (intercept: MIS, 32.1 vs. open, 36.1, p= 0.11; slope: MIS, β= -0.17 vs. open, β= -0.18, p= 0.35). The model suggested a breakpoint at day 110.2± 8.3. After this breakpoint, the slopes were -0.002 for MIS and -0.014 for the open group, showing plateaued slope and no significant differences between groups (
Table 4). Combined line plots were created from these analyses by extracting fixed-effects slope, intercept, breakpoint, and slope change after breakpoint between LIP and PP (
Fig. 4).
4) NRS back
A regression model with random effects revealed no significant differences in slopes during IP between MIS and open groups (slope difference: -0.001± 0.002, p= 0.53), with a plateau confirmed at day 27.2 ± 2.6. Additional analysis showed that the breakpoint of NRS back occurred earlier in the MIS group (26.7± 2.6 days) compared to the open group (51.7± 6.6 days), indicating that MIS patients reached the PP significantly faster (mean difference: 25.0 days, p< 0.001) (
Fig. 5).
5) NRS leg
A segmented regression model with random effects for leg pain was adjusted for preoperative leg pain given the significant differences preoperatively. Statistically, 2 breakpoints (mixed-effects model: day 12.8± 2.5 and day 94.1± 12.7) appeared optimal based on AIC and BIC, but the slope before second breakpoint had nearly plateaued (β= -0.016). Under this 2-breakpoint model, no significant between-group differences in slopes were observed during EIP (slope difference: 0.010± 0.008, p= 0.21) or during LIP (0.0004± 0.002, p= 0.80). In consideration of relatively small sample size and the need for the direct comparisons, we ultimately adopted a single-breakpoint approach. With this model, the mixed-effects segmented regression initially showed a significant difference (IP to PP: MIS 19.6± 1.7 vs. open 8.9± 4.6, p= 0.031). However, after adjusting for preoperative NRS leg scores in a multivariable model, the difference was no longer statistically significant (IP to PP: MIS 20.3± 1.7 vs. open 9.8± 5.3, p= 0.06) (
Fig. 6).
There were 6 cases (1.9%) requiring reoperation in total (5 MIS and 1 open). In the MIS group, 2 patients underwent redecompression at the index level due to persistent leg pain on postoperative days 63 and 133, while one required adjacent segment decompression on day 420. Additionally, 2 patients in the MIS group (days 154 and 178) and one in the open group (day 337) underwent hardware removal due to associated leg pain.
DISCUSSION
This retrospective cohort study of 324 patients undergoing single-level TLIF compared postoperative recovery trajectories and symptomatic improvement phases between MIS and open approaches using mixed-effects segmented regression analysis. Our analysis identified 4 distinct phases of ODI recovery: a postoperative disability peak around day 13, followed by rapid improvement until day 28 (EIP), gradual improvement from 1 to 3 months (LIP), and a plateau after 3–4 months (PP). A key finding was the significantly less pronounced worsening of disability in the MIS group during the initial postoperative period and a 4-week earlier plateau in back pain compared to the open group. These findings suggest a clear early advantage of MISTLIF for back pain recovery, potentially leading to faster functional recovery and improved patient satisfaction.
This study identified the 4 distinct postoperative phases in ODI recovery: PDP, EIP, LIP, and PP, a level of granularity not typically captured by traditional fixed-time-point analyses. While previous studies utilizing longitudinal PROMs data have reported postoperative worsening of disability following spine surgery, our segmented regression analysis allowed for a more detailed characterization of this early period, defining the PDP and EIP with robustly estimated “turning points” at approximately 2 weeks and 1 month postoperatively. A prospective study comparing open TLIF to MIS-TLIF with 38 patients showed improved ODI in the MIS group at 12 weeks but did not fully characterize early postoperative differences within 1 month [
21]. Similarly, a cohort study of MIS-TLIF demonstrated significant disability worsening at 2 weeks, followed by significant improvement at 6 weeks from 2 weeks [
20], but lacked a comparison with open procedure. Critically, we demonstrated a significantly less pronounced worsening of disability in the MIS group during the first 2 weeks, and a 4-week earlier plateau in back pain compared to open TLIF, highlighting the particular advantages of the MIS approach within the first 1 to 2 postoperative months.
Another key finding of this study is the identification of distinct subphases within the IP of ODI recovery: EIP (days 13.4–28.4) and LIP (days 28.4–110.2). This biphasic pattern of improvement was not typically captured by traditional fixed-time-point analyses [
20]. This novel finding of an EIP, characterized by rapid ODI reduction, likely reflects immediate postoperative pain and inflammation reduction from surgical decompression and stabilization. This may be attributed to decreased mechanical irritation of neural elements and reduced inflammatory responses at the surgical site. The subsequent LIP, with a slower improvement rate, likely represents ongoing tissue healing, bone fusion, and neural adaptation such as neuropathic pain [
22-
24]. While direct pain scores plateaued within 1 to 2 months, the continued improvement in disability observed during the LIP suggest that functional recovery extends beyond symptomatic improvement, persisting until approximately 3 to 4 months postoperatively. Recognizing these distinct recovery phase has important clinical implications for managing patient expectations and tailoring postoperative rehabilitation programs.
Symptomatic plateaus in ODI were observed at approximately 3 to 4 months postoperatively in our study, which contrasts with the 6-month plateau reported in some prior studies [
10,
20]. These discrepancies may arise from differences in study methods, patient populations, and follow-up intervals. Traditional 2-time-point comparisons can miss subtle inflection points due to variable data collection schedules, missing data, and smaller samples. For instance, previous studies using 6-month follow-up may have missed the earlier plateau observed in our study due to the lack of more frequent assessments during 3 to 6 months postoperatively. Our segmented regression with mixed-effects modeling provides a more robust analysis, capturing continuous data and interpatient variability [
25,
26]. Clinical factors also influence recovery trajectories. Multilevel fusions, for instance, commonly require extended healing periods, often plateauing around 6 months or later. A study of open TLIF alone reported a plateau between 6 months and 2 years [
27]. Our focus on single-level MIS and open TLIF likely contributed to earlier plateau times. Although statistical modeling complexities prevented detailed ODI-based comparison of plateau timing, the difference in back pain plateau favoring MIS may account for some advantage observed in ODI recovery patterns. This information offers valuable insights for patients and surgeons when considering the next steps, such as reoperation, for those who do not show significant improvement.
Regarding leg pain, although the study did not reveal significant differences in overall improvement between groups, the open TLIF group reached the PP more quickly. This observed difference in break points could be influenced by the greater baseline leg pain in the MIS group. A multivariable model adjusting for preoperative leg pain reduced the differences, indicating the possible influence of baseline symptoms. Additionally, the relatively small open TLIF sample size may limit definitive conclusions. Improvement trend in leg pain may arise from different decompression techniques between MIS and open TLIF. Open approach provides both direct decompression— through a larger laminectomy and complete facetectomy via midline posterior approach—and indirect decompression through ligamentotaxis from increased disc height. While the efficacy of indirect decompression in severe canal stenosis has been demostrated [
3,
12,
21], the speed of symptom relief associated with this mechanism is not well understood. In MIS-TLIF, if the limited laminectomy and indirect decompression sufficiently relieves neural compression, symptoms should promptly improve. However, when residual stenosis remains, decompression occurs gradually due to the slow shrinkage of the ligamentum flavum following fusion. As a result, symptom improvement may be delayed [
28]. This difference in decompression mechanism, especially the smaller proportion of direct posterior element decompression in MIS-TLIF, may help explain the slower attainment of the PP for leg pain. Further investigation with a larger, well-balanced cohort is needed to confirm these observations.
There are several limitations in this study. First, as a retrospective analysis of prospectively collected PROMs at predefined time frames, data availability was not homogeneous throughout the follow-up term, potentially introducing bias in determining breakpoints. While random effects were employed to mitigate this risk, the precision of breakpoint estimation might still be affected by outliers or gaps in data points. Further studies with varied data collection days from surgery could improve estimations. Additionally, breakpoints for ODI recovery phases were determined based on the full cohort rather than independently within each group to maintain statistical robustness, as estimating multiple breakpoints in the smaller open TLIF group would have led to instability and unreliable comparisons. Second, surgical indications for open and MIS procedures were inconsistent, which can lead selection bias. The study included data from multiple surgeons, some of whom performed both approaches, while others exclusively preferred one. This heterogeneity in practice, combined with challenges in assessing postoperative recovery trends and predicting outcomes based on preoperative “difficulty” (e.g., radiographic findings), may have influenced the observed differences between groups and limit the generalizability of the results. Additionally, the sample size imbalance between the open TLIF (n= 56) and MIS-TLIF (n= 268) groups may have affected the identification of the true turning point in the open TLIF group, potentially impacting the trend analysis. A smaller sample size can increase variability in breakpoint detection and reduce statistical power, potentially limiting the ability to identify subtle recovery differences. However, similar retrospective cohort studies often face group size imbalances, making this a common challenge in retrospective studies. To mitigate this, we utilized mixed-effects segmented regression, which accounts for interindividual variability and enhances the robustness of our findings. Nevertheless, a prospective multicenter study with a larger open TLIF cohort is needed for further validation. A prospective randomized trial or multicenter study would provide more robust conclusions. Third, the study relied solely on PROMs without incorporating objective functional metrics such as gait analysis or physical performance tests, which might offer additional insights into recovery trends. Additionally, psychosocial factors and patient-specific preferences, which could influence recovery trajectories and PROMs, were not accounted for. However, depressive mood was included as a demographic data (
Table 1), and no significant difference was observed between groups (MIS 15.2% vs. open 15.7%). Patient-reported outcomes are also subject to potential reporting bias due to their self-assessed nature. Fourth, the impact of baseline demographic and symptoms were not fully addressed in this study, particularly for leg pain. Fifth, the current statistical package does not support estimating multiple breakpoints within a single regression model, requiring us to determine breakpoints separately for each surgical approach. Finally, while random effects account for individual variability, the lack of long-term follow-up data beyond 1 year limits the evaluation of sustained trends or late complications. As a result, this study cannot fully assess reoperation rates, complications, or fusion status. However, significant symptomatic changes beyond 1 year are clinically uncommon and, when present, often attributable to other factors. Thus, the absence of long-term data likely may have a minimal impact on this study’s conclusions.
CONCLUSION
This study identified 4 distinct phases of symptomatic improvement following TLIF: (1) PDP: an early phase in which disability peaks around 2 weeks, (2) EIP: a period of rapid improvement by 1 month, (3) LIP: a slower but ongoing improvement of disability between 1 and 3 months, and (4) PP: a stable plateau reached after 4 months. In the PDP, the MIS approach was associated with reduced disability from back pain, reaching a plateau of low back pain approximately 4 weeks sooner than the open procedure. These findings highlight that the advantages of minimally invasive surgery in TLIF are most pronounced during the first month postoperatively, resulting in less back pain-related disability and a faster plateau for back pain. By elucidating these distinct recovery trajectories, this study provides valuable insights for both clinical decision-making and patient counseling. However, the limitations of this retrospective study, including potential selection bias and heterogeneity in surgical practice, warrant further investigation through prospective randomized controlled trials.
NOTES
-
Conflict of Interest
Sheeraz Qureshi has the following disclosures: Tissue Differentiation Intelligence: Ownership/Equity/Investment; Stryker K2M: Royalties from Intellectual Property, Designer, Consultant; SpineGuard, Inc.: Consultant; Globus Medical, Inc.: Royalties from Intellectual Property, Speakers’ Bureau, Consultant; Simplify Medical, Inc.: Clinical Event Committee; AMOpportunities: Honoraria; Surgalign: Consultant; Viseon, Inc.: Research Support (either personally or through HSS), Consultant; HS2, LLC: Ownership/Equity/Investment; LifeLink.com Inc.: Medical or Scientific Advisory Board Membership; Spinal Simplicity, LLC: Medical or Scientific Advisory Board Membership; Contemporary Spine Surgery: Editorial Board; North American Spine Society (NASS): Political Engagement Committee member, Payor Policy Review Committee member, SpinePAC Advisory Committee member, CME Committee member; Annals of Translational Medicine (ATM): Editorial Board; Hospital Special Surgery Journal: Editorial Board, Senior Associate Editor (2021–2024); Society of Minimally Invasive Spine Surgery (SMISS): Program Committee member, 2018 Annual Meeting Program Chair, Board of Directors (2021–2024), Professional Society member of Directors/Trustees/Governors/ Managers, Member at Large or Committee member; Lumbar Spine Research Society (LSRS): Website Committee member (2022–2022), Professional Society member of Directors/ Trustees/Governors/Managers, Member at Large or Committee member; Cervical Spine Research Society (CSRS): Publications Committee member (2019–2022), Professional Society member; Minimally Invasive Spine Study Group: Board of Directors - Treasurer; Association of Bone and Joint Surgeons (ABJS): Program Committee member, Professional Society member; International Society for the Advancement of Spine Surgery (ISASS): Education Committee, Program Committee, 2021 Annual Meeting Program Chair, Professional Society member. Sravisht Iyer has the following disclosures: Globus Medical: Paid presenter or speaker; Stryker: Paid presenter or speaker; Vertebral Columns/International Society for the Advancement of Spine Surgery (ISASS): Editorial or governing board. Han Jo Kim has the following disclosures: Zimmer Biomet: Royalties from Intellectual Property; DePuy Synthes: Medical or Scientific Advisory Board Membership; NuVasive: Medical or Scientific Advisory Board Membership; Stryker: Medical or Scientific Advisory Board Membership; Green Sun Medical: Ownership/ Equity/Investment; Hospital for Special Surgery (HSS): Research Support (provided personally or through HSS); Stryker Spine: Royalties from Intellectual Property, Consultant; Thompson Surgical Instruments: Royalties from Intellectual Property; Globus Medical: Speakers’ Bureau; North American Spine Society (NASS): Board of Directors; Scoliosis Research Society (SRS): Board of Directors. James Farmer has the following disclosures: Green Sun Medical: Ownership/Equity/Investment. Harvinder Sandhu has the following disclosures: Medtronic: Consultant; Stryker: Consultant; Spinal Kinetics: Ownership/Equity/Investment; Providence Medical Technology: Ownership/Equity/Investment; Hospital for Special Surgery (HSS): Inventor; Intrinsic Therapeutics: Ownership/Equity/Investment; Orthofix: Consultant; Tissue Differentiation Intelligence: Ownership/Equity/ Investment; Nocimed: Ownership/Equity/Investment. Albert has the following disclosures: Medtronic: Ownership/Equity/ Investment; DePuy Synthes: Consultant; Paradigm Spine: Ownership Interest in a medical device distributor, manufacturer, surgery center, or group-purchasing organization; Stryker Spine: Ownership/Equity/Investment; Thieme: Author, Co-author, or Editor; Jaypee Brothers Medical Publishers: Author, Co-author, or Editor; Elsevier: Author, Co-author, or Editor; Springer: Author, Co-author, or Editor; K2M: Ownership/Equity/Investment; Innovasis: Ownership/Equity/Investment; Paradigm BioDevices: Ownership Interest in a medical device distributor, manufacturer, surgery center, or group-purchasing organization; SLACK Incorporated: Author, Co-author, or Editor; Facet Solutions: Ownership/Equity/Investment; Intrinsic Therapeutics: Ownership/ Equity/Investment; Meditech Spine: Ownership/Equity/ Investment; Facetlink: Ownership Interest in a medical device distributor, manufacturer, surgery center, or group-purchasing organization; Alphatec Spine: Ownership/Equity/Investment; Globus Medical: Ownership/Equity/Investment; Medtronic Sofamor Danek: Royalties from Intellectual Property; Facetlink: Ownership Interest in a medical device distributor, manufacturer, surgery center, or group-purchasing organization; Stryker Spine: Ownership/Equity/Investment; Medtronic: Royalties from Intellectual Property. The other authors have nothing to disclose.
-
Funding/Support
Data collection and management were conducted using REDCap (Research Electronic Data Capture), hosted at the Weill Cornell Medicine Clinical and Translational Science Center. This platform is supported by the National Center for Advancing Translational Science of the National Institutes of Health under award number UL1 TR002384.
-
Author Contribution
Conceptualization: TA, SAQ, SI; Data curation: TA, ERZ, AME, AL, AP, SH, TH, OCT, KA, TJA, JF, RH, HS, HJK, FCL, JED, SI, SAQ; Formal analysis: TA; Investigation: TA, SI, SAQ; Methodology: TA; Project administration: TA, SAQ, SI; Writing – original draft: TA, ERZ, AME, AL, AP, SH; Writing – review & editing: TA, JF, RH, HS, HJK, FCL, JED, SI, SAQ.
Fig. 1.Overall ODI recovery trajectory. The blue line represents the overall ODI recovery trajectory estimated using preliminary segmented regression analysis. This analysis identified 4 distinct phases: postoperative disability phase (PDP, up to day 11), early improvement phase (EIP, days 11–30), late improvement phase (LIP, days 30–76), and plateau phase (PP, day 76 onwards). These phases were used to develop mixed-effects models for between-group comparisons. Multiple R2=0.27; Adjusted R2=0.26. ODI, Oswestry Disability Index; β, regression coefficient (slope, points/day).
Fig. 2.Overall NRS back pain recovery trajectory. The blue line represents the overall NRS back pain recovery trajectory estimated using segmented regression analysis. This analysis identified 2 distinct phases: improvement phase (IP, up to day 27) and plateau phase (PP, day 27 onwards). These phases were used to develop mixed-effects models for between-group comparisons. Multiple R2=0.28; Adjusted R2=0.28. NRS, Numerical Rating Scale; β, regression coefficient (slope, points/ day).
Fig. 3.Overall NRS leg pain recovery trajectory. The blue line represents the overall NRS leg pain recovery trajectory estimated using segmented regression analysis. This analysis identified three distinct phases: early improvement phase (EIP, up to day 12), late improvement phase (LIP, days 12–79) and plateau phase (PP, day 79 onwards). These phases were used to develop mixed-effects models for between-group comparisons. These phases were used to develop mixed-effects models for between-group comparisons. Multiple R2=0.27; Adjusted R2=0.27. NRS, Numerical Rating Scale; β, regression coefficient (slope, points/day).
Fig. 4.Comparison of ODI recovery trajectories between MIS and open TLIF. This figure displays the postoperative ODI recovery trajectories for patients undergoing MIS and open TLIF. The blue and red lines represent the estimated mean ODI scores over time for the MIS and open groups, respectively, derived from segmented regression analysis. Individual patient data points are shown as translucent blue (MIS) and red (open) circles. The vertical dashed lines indicate the identified breakpoints defining the 4 distinct recovery phases: postoperative disability phase (PDP, up to day 13), early improvement phase (EIP, days 13–28), late improvement phase (LIP, days 28–110), and plateau phase (PP, day 110 onwards). The MIS group demonstrated a significantly less pronounced worsening of disability during the PDP and significantly lower ODI scores at the end of the PDP (day 13) compared to the open group. While the open group showed a steeper initial increase in ODI during PDP, there were no significant differences in the slopes of improvement (rate of change) during the EIP, LIP, or PP between the 2 groups. Because separate regression lines were fitted within each phase, the lines are not smoothly connected at the breakpoints. This reflects the distinct rate of change in ODI scores observed across the different recovery phases. ODI, Oswestry Disability Index; MIS, minimally invasive; TLIF, transforaminal lumbar interbody fusion.
Fig. 5.Comparison of back pain recovery trajectories between MIS and open TLIF. A mixed-effects model identified a breakpoint between the improvement phase (IP) and plateau phase (PP) at day 26.7±2.6 for the MIS group (blue line), while the open approach (red line) reached it at day 51.7±6.6, showing that MIS recovery was faster by 25 days (p<0.001). Patient data points are shown as translucent blue (MIS) and red (open) circles. Marginal R2=0.61; Conditional R2=0.78 for overall model. MIS, minimally invasive; TLIF, transforaminal lumbar interbody fusion; NRS, Numerical Rating Scale.
Fig. 6.Comparison of leg pain recovery trajectories between MIS and open TLIF. A mixed-effects model found a breakpoint between the improvement and plateau phases at day 19.6±1.7 for the MIS group (blue line) and day 8.9±4.6 for the open approach (red line), indicating that MIS recovery was 10 days faster (p=0.031). Adjusting for baseline leg pain, there were no significant differences in breakpoint (MIS 20.3±1.7 vs. open 9.8±5.3, p=0.06). Patient data points are shown as translucent blue (MIS) and red (open) circles. Marginal R2=0.66; Conditional R2=0.73 for overall model. The adjusted p-value was calculated using a model that accounts for preoperative NRS leg scores. MIS, minimally invasive; TLIF, transforaminal lumbar interbody fusion; NRS, Numerical Rating Scale.
Table 1.
Table 1.
|
Variable |
MIS (n = 268) |
Open (n = 56) |
p-value |
|
Age (yr) |
60.0 ± 13.1 |
57.6 ± 17.7 |
0.24 |
|
Male sex |
139 (51.9) |
26 (46.4) |
0.55 |
|
Race |
|
|
0.69 |
|
White |
222 (82.8) |
43 (76.8) |
|
|
Black |
16 (6.0) |
5 (8.9) |
|
|
Asian |
6 (2.2) |
1 (1.8) |
|
|
Other |
24 (9.0) |
7 (12.5) |
|
|
BMI (kg/m2) |
27.9 ± 6.0 |
27.7 ± 5.8 |
0.83 |
|
Smoking |
91 (34.0) |
22 (39.3) |
0.54 |
|
Depression mood |
40 (15.2) |
8 (15.7) |
1.00 |
|
CCI |
|
|
|
|
0 |
98 (37.3) |
15 (29.4) |
0.13 |
|
1 |
83 (31.6) |
24 (47.1) |
|
|
2 |
43 (16.3) |
8 (15.7) |
|
|
3 |
22 (8.4) |
4 (7.8) |
|
|
≥4 |
17 (6.5) |
0 (0.0) |
|
|
LOS (hr) |
45.0 ± 31.0 |
66.4 ± 36.3 |
< 0.001*
|
|
Preoperative |
|
|
|
|
ODI |
41.9 ± 15.3 |
41.5 ± 13.7 |
0.85 |
|
NRS back |
6.2 ± 2.6 |
6.1 ± 2.7 |
0.70 |
|
NRS leg |
6.0 ± 2.9 |
5.0 ± 3.0 |
0.027*
|
Table 2.Fixed-effects of mixed-effect segmented regression model from preoperative to day 30
Table 2.
|
Variable |
Estimate (β) |
SE |
t-value |
p-value |
|
Preoperative ODI (intercept) |
|
|
|
|
|
MIS |
41.9 |
0.9 |
44.9 |
< 0.001 |
|
Difference from open |
-0.2 |
2.2 |
-0.1 |
0.92 |
|
Open |
41.6†
|
2.0†
|
20.5†
|
< 0.001†
|
|
Slope before breaking point |
|
|
|
|
|
MIS |
0.93 |
0.54 |
1.72 |
0.09 |
|
Difference from open |
0.48 |
0.18 |
2.70 |
0.008*
|
|
Open |
1.42†
|
0.56†
|
2.53†
|
0.012†
|
|
Breaking point (day) |
13.4 |
3.0 |
|
|
Table 3.Fixed-effects of mixed-effect segmented regression model from day 13 to 80
Table 3.
|
Variable |
Estimate (β) |
SE |
t-value |
p-value |
|
ODI at the starting point (intercept) |
|
|
|
|
|
MIS |
49.3 |
2.5 |
19.9 |
< 0.001 |
|
Difference from open |
11.2 |
3.8 |
3.0 |
0.003*
|
|
Open |
60.5†
|
2.0†
|
29.8†
|
< 0.001 |
|
Slope before breaking point |
|
|
|
|
|
MIS |
-1.07 |
0.39 |
-2.73 |
0.007 |
|
Difference from open |
-0.14 |
0.08 |
-1.76 |
0.08 |
|
Open |
-1.21†
|
0.56†
|
-2.16†
|
0.032 |
|
Breaking point (day) |
28.4 |
5.2 |
|
|
Table 4.Fixed effects of the mixed-effect segmented regression model from day 28 onwards
Table 4.
|
Variable |
Estimate (β) |
SE |
t-value |
p-value |
|
ODI at the starting point (intercept) |
|
|
|
|
|
MIS |
32.1 |
1.2 |
27.2 |
< 0.001 |
|
Difference from open |
4.0 |
2.5 |
1.6 |
0.11 |
|
Open |
36.1†
|
2.0†
|
17.8†
|
< 0.001†
|
|
Slope before breaking point |
|
|
|
|
|
MIS |
-0.17 |
0.02 |
-8.30 |
< 0.001 |
|
Difference from open |
-0.01 |
0.01 |
-0.93 |
0.35 |
|
Open |
-0.18†
|
0.58†
|
-0.32†
|
0.75†
|
|
Breaking point (day) |
110.2 |
8.3 |
|
|
REFERENCES
- 1. Qureshi S. Pearls: improving upon minimally invasive transforaminal lumbar interbody fusion. Clin Orthop Relat Res 2019;477:501-5.
- 2. Carlson BB, Saville P, Dowdell J, et al. Restoration of lumbar lordosis after minimally invasive transforaminal lumbar interbody fusion: a systematic review. Spine J 2019;19:951-8.
- 3. Kim CH, Easley K, Lee JS, et al. Comparison of minimally invasive versus open transforaminal interbody lumbar fusion. Global Spine J 2020;10(2 Suppl):143S-150S.
- 4. Parker SL, Adogwa O, Witham TF, et al. Post-operative infection after minimally invasive versus open transforaminal lumbar interbody fusion (TLIF): literature review and cost analysis. Minim Invasive Neurosurg 2011;54:33-7.
- 5. Tan JH, Liu G, Ng R, et al. Is MIS-TLIF superior to open TLIF in obese patients?: a systematic review and meta-analysis. Eur Spine J 2018;27:1877-86.
- 6. Peng CW, Yue WM, Poh SY, et al. Clinical and radiological outcomes of minimally invasive versus open transforaminal lumbar interbody fusion. Spine (Phila Pa 1976) 2009;34:1385-9.
- 7. Rieger B, Jiang H, Ruess D, et al. First clinical results of minimally invasive vector lumbar interbody fusion (MIS-VLIF) in spondylodiscitis and concomitant osteoporosis: a technical note. Eur Spine J 2017;26:3147-55.
- 8. Wang X, Borgman B, Vertuani S, et al. A systematic literature review of time to return to work and narcotic use after lumbar spinal fusion using minimal invasive and open surgery techniques. BMC Health Serv Res 2017;17:446.
- 9. Bovonratwet P, Samuel AM, Mok JK, et al. Minimally invasive lumbar decompression versus minimally invasive transforaminal lumbar interbody fusion for treatment of low-grade lumbar degenerative spondylolisthesis. Spine (Phila Pa 1976) 2022;47:1505-14.
- 10. Shinn D, Mok JK, Vaishnav AS, et al. Recovery kinetics after commonly performed minimally invasive spine surgery procedures. Spine (Phila Pa 1976) 2022;47:1489-96.
- 11. Wu MH, Dubey NK, Li YY, et al. Comparison of minimally invasive spine surgery using intraoperative computed tomography integrated navigation, fluoroscopy, and conventional open surgery for lumbar spondylolisthesis: a prospective registry-based cohort study. Spine J 2017;17:1082-90.
- 12. Hammad A, Wirries A, Ardeshiri A, et al. Open versus minimally invasive TLIF: literature review and meta-analysis. J Orthop Surg Res 2019;14:229.
- 13. Valsamis EM, Husband H, Burchette D, et al. Modelling the effect of a dedicated hip fracture unit on patient outcomes using segmented robust linear regression techniques. Injury 2021;52 Suppl 5:S3-6.
- 14. Schober P, Vetter TR. Segmented regression in an interrupted time series study design. Anesth Analg 2021;132:696-7.
- 15. Gabrio A, Plumpton C, Banerjee S, et al. Linear mixed models to handle missing at random data in trial-based economic evaluations. Health Econ 2022;31:1276-87.
- 16. Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: building an international community of software platform partners. J Biomed Inform 2019;95:103208.
- 17. Harris PA, Taylor R, Thielke R, et al. Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42:377-81.
- 18. Muggeo VM, Atkins DC, Gallop RJ, et al. Segmented mixed models with random changepoints: a maximum likelihood approach with application to treatment for depression study. Stat Model 2014;14:293-313.
- 19. Muggeo VMR. Estimating regression models with unknown break-points. Stat Med 2003;22:3055-71.
- 20. Shahi P, Subramanian T, Tuma O, et al. Temporal trends of improvement after minimally invasive transforaminal lumbar interbody fusion. Spine (Phila Pa 1976) 2025;50:81-7.
- 21. Hartmann S, Lang A, Lener S, et al. Minimally invasive versus open transforaminal lumbar interbody fusion: a prospective, controlled observational study of short-term outcome. Neurosurg Rev 2022;45:3417-26.
- 22. Woolf CJ. Central sensitization: implications for the diagnosis and treatment of pain. Pain 2011;152(3 Suppl):S2-15.
- 23. Attal N, Cruccu G, Baron R, et al. EFNS guidelines on the pharmacological treatment of neuropathic pain: 2010 revision. Eur J Neurol 2010;17:1113-e88.
- 24. Pilitsis JG, Lucas DR, Rengachary SS. Bone healing and spinal fusion. Neurosurg Focus 2002;13:e1.
- 25. Lohse KR, Shen J, Kozlowski AJ. Modeling longitudinal outcomes: a contrast of two methods. J Mot Learn Dev 2020;8:145-65.
- 26. Wainwright PE, Leatherdale ST, Dubin JA. Advantages of mixed effects models over traditional ANOVA models in developmental studies: a worked example in a mouse model of fetal alcohol syndrome. Dev Psychobiol 2007;49:664-74.
- 27. Poh SY, Yue WM, Chen LT, et al. Two-year outcomes of transforaminal lumbar interbody fusion. J Orthop Surg (Hong Kong) 2011;19:135-40.
- 28. Ohtori S, Orita S, Yamauchi K, et al. Change of lumbar ligamentum flavum after indirect decompression using anterior lumbar interbody fusion. Asian Spine J 2017;11:105-12.
, Eric R. Zhao1,2
