Background

Despite the decline of cigarette smoking over the last 25 years, smoking still remains a health concern to health-care providers all over the world [1]. Smoking continues to be the primary cause of morbidity and mortality causing around one in five deaths each year in the United States [2]. In spine surgery, the negative effects of smoking have been reported several times with respect to bone healing. The following processes play an explanatory role: nicotine is responsible for inhibiting tumor necrosis factor alpha (a cytokine that is crucial to the inflammatory response), angiogenesis, and osteogenesis, processes crucial to bone metabolism; dioxin, is postulated to adversely affect osteogenesis; and polycyclic aryl hydrocarbon compounds bind osteoblast and osteoclast receptors causing significant bone metabolism impairment [3–6].

The negative effects of smoking, however, are not confined to bone health. In fact, research has demonstrated that tobacco not only negatively influences vitamin D and calcium absorption, but also established negative effects on parathyroid hormones, adrenal hormones, gonadal hormones and oxidative stress. All these factors have a detrimental effect on the body’s hemostasis. This suggests that the effect of smoking in relation to spinal surgery is not exclusive to bone healing, radiological outcomes, and surgical site infections; effects that are oftentimes the primary focus of spine surgery research [7–12] but is also a key player in the whole body’s metabolism and thus plays an overall crucial role in physical health and clinical outcomes after surgery. Definitive reports of the direct influence of smoking around the time of surgery vary with some studies observing a negative outcome postoperatively due to preoperative smoking, while other original studies found no measurable influence on the functional outcome of patients following spinal spondylodesis. [13,14].

Although smoking is associated with a high rate of reoperation, current guidelines for spondylodesis do not restrict patients from smoking prior to surgery [15]. At this moment, it is still difficult to answer the question of whether or not a patient will significantly benefit from stopping smoking following instrumented spinal surgery with regard to clinical outcomes. Thus, concrete and cumulative evidence is required to emphasize the risk of smoking on patient outcomes and the possible benefits of smoking cessation prior to spinal fusion surgery.

A previously conducted systematic review suggested that radiological fusion may also depict the effects of smoking on patients undergoing spondylodesis, with findings such as pseudoarthrosis and non-union [16]. However, this review and other meta-analyses on this subject focused on the effects of smoking on radiological parameters [16–19]. While a limited number of systematic reviews [20,21] and meta-analyses [16–19] have investigated the relationship between smoking and fusion, these have largely restricted their focus to radiographic fusion rates and general postoperative outcomes. Such endpoints, though clinically relevant, provide an incomplete picture of patient recovery and functional improvement. Additionally, these studies were conducted several years ago and relied primarily on absolute outcomes without calculating relative mean differences. This limits their ability to calculate the magnitude of risk in a way that is directly comparable across populations and clinically interpretable. Incorporating relative values, as we have done provides a clearer understanding of smoking’s impact on spinal fusion outcomes. Our study also aimed to expand the scope by incorporating patient-reported outcomes while also including the most recently published literature and assessing fusion and clinical outcomes. Furthermore, we aim to provide more granular data by stratifying patients based on the levels operated on and by distinguishing between former and never smokers.

Search strategy and criteria

PubMed, EMBASE and Cochrane were searched up to August 2024. Search terms related to cigarette smoking and spinal fusion were used to identify studies analyzing the effect of smoking versus non-smoking and/or smoking cessation on the clinical and radiological outcome of spine surgery (S1 Appendix). Covidence [23] was utilized for the management of articles and literature for the current study.

Studies were included if they met the following criteria: (1) the patient population consisted of patients who underwent spinal instrumented spondylodesis procedure, (2) smoking was the main exposure in the study and/or patients were required to stop smoking before surgery, (3) duration of follow-up was at least two months.

Articles were excluded if they were not written in English or Dutch, if they were nonhuman studies, non-peer-reviewed, case reports, meta-analyses or systematic reviews. Titles and abstracts were screened independently by two of six authors and a third independent author was consulted in the case of conflicting evaluations. Full texts were independently reviewed by two of six authors. A third author was consulted in the case of conflicting evaluations.

Extraction of data

For each included study, the following information was extracted: study characteristics (study design, year of publication, country where study was conducted, sample size, and study duration); baseline demographics (mean age, smoking history, gender); surgical information (vertebral column level, surgical approach, and numbers of operated levels); clinical outcomes; radiological postoperative outcomes; and post-operative complications. Clinical outcome data that were extracted included: Oswestry disability index (ODI; ranging from 0–100, a higher score indicating worse functionality), Neck Disability Index (NDI; ranging from 0–100, a higher score indicating worse functionality), Visual Analog Scale (VAS) for neck pain, arm pain, back pain, and leg pain (ranging from 0–10; a higher score indicating more pain), EuroQol five-dimension scale (EQ5D; ranging from 0 to 1; a higher score indicating more quality of life), Medical Outcomes Study Questionnaire (SF12; ranging from 0–100, a higher score indicating better health), Japanese Orthopaedics Association score (JOA; ranging from 0 to 17; a higher score indicating better function) and patient satisfaction score measured by the 7 point Likert perceived recovery scale (higher score indicating less satisfaction with outcome). Furthermore, surgical site infections (SSI) and number of reoperations were extracted. The radiological parameters that were extracted concerned bony fusion and pseudoarthrosis rates. Data collection was performed by two of four authors and a third author solved any discrepancies

Quality assessment

Two authors evaluated the quality of each included study using the Newcastle Ottawa Scale for assessing cohort studies and the Joanna Briggs Institute Appraisal checklist for case series [24,25]. The Newcastle Ottawa Scale assesses three domains: subject selection, comparability, and assessment of outcome, with a total of nine possible points. The scores were categorized as follows: 0–3 points as poor quality; 4–6 points as fair quality; and 7–9 points as good quality. For case series, 10 questions from the checklist were addressed of which there were four possible answers, that is, yes (Y), no (N), unclear (U) or not applicable (NA). Values for each answer included: Y = 1, N, U, NA = 0. The quality scores were thus calculated as follows: 1–4 as low, 5–7 as moderate, and 8–10 as good. In case of any disagreements, the authors discussed and resolved them among themselves. If a consensus could not be reached, a third author provided the final judgment.

Statistical analysis

The studies reported raw means along with the difference in means for various continuous outcomes. To make for a fair comparison between smokers and non-smokers and to take into account varying baseline values, we calculated the percent mean change from baseline {[(pre-post)/pre]*100%} for each study. For categorical outcomes, the incidence of events along with the 95% confidence interval was derived from each study. For each of the smokers and non-smokers subgroups, patients were further stratified by region of operation (cervical, thoracic, lumbar) or the number of levels operated on (single, multiple) only for those studies that presented these stratified data. Studies that included patients with single or multiple levels of operation as one category could not be represented in this sub-stratified analysis by level of operation. For the main analysis by smokers vs. non-smokers, because former smokers were included among the non-smokers, we further stratified non-smokers into former smokers and never smokers when data were provided by the original studies. All of the pooled point estimates for the reported clinical outcomes and their respective 95% confidence intervals were calculated using a random effects model via the Dersimonian and Laird method [26]. Notably, a direct comparison between outcomes in smokers vs. non-smokers was not possible with the data provided by the individual articles, as no adjustment for confounding was made to compare these two groups. Providing a p-value for such a comparison would be misleading, as these p-values would be biased. Hence, pooled point estimates were provided for each stratum along with their respective 95% CIs as a measure of variability around the point estimates. Forest plots were used to visualize individual and summary estimates. The Higgins I² index was used to evaluate heterogeneity [27]. An I² value of more than 50% was considered high. Due to the small number of studies (< 10) included per analysis, it was not possible to assess the potential for small study effects through funnel plots. Comprehensive Meta-analysis, version 4 (Biostat, Inc, Englewood, New Jersey) was utilized to perform statistical analysis.

NDI

Six studies evaluated the differences in NDI (cervical spine) pre- and post-operatively in smokers and non-smokers (S1 Table), and in two studies patients were stratified as smokers, former smokers, and never smokers [44,49]. The pooled percent mean change from baseline for smokers was 42.8% (95% CI: 25.4% to 60.1%; I²: 100%; six studies) and for non-smokers was 49.7% (95% CI: 32.4% to 64.0%; I²: 100%; six studies) (Table 2). Further subgrouping provided by two out of the six studies resulted in pooled percent mean changes from baseline of 32.6% (95% CI: 11.1% to 54.1%; I²: 100%) in never-smokers, 22.7% (95% CI: 1.2% to 44.2%; I²: 100%) in former smokers, and 24.2% (95% CI: 2.7% to 45.7%; I²: 100%) in smokers (S2 Table). The results of the subgroup analysis suggested a numerically better outcome in never-smokers, though this may not be clinically important (Table 3). Only two studies distinguished between the number of levels fused (multilevel surgery) and stratification revealed a consistent improvement in NDI amongst non-smokers (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Pooled point estimates of various outcomes (surgical site infection, reoperation, fusion, pseudoarthrosis, NDI, ODI, SF-12, Joa score, VAS back, leg, neck, and arm pain, patient satisfaction) for smokers and non-smokers^±.

https://doi.org/10.1371/journal.pone.0337799.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Pooled point estimates of various outcomes (pseudoarthrosis, NDI, ODI, Joa score, VAS back, leg, neck, and arm pain) stratified by smoking history (smokers, former smokers, and never smokers) undergoing spondylodesis.

https://doi.org/10.1371/journal.pone.0337799.t003

ODI

The change in ODI (lumbar spine) following surgery was determined in smokers and non-smokers in six different studies. This was followed by a subgroup analysis consisting of smokers, former smokers and never smokers in two studies [33,36]. The pooled percent mean change from baseline was 48.6% (95% CI: 34.7% to 62.5%; I²: 100%) in smokers and 56.5% (95% CI: 42.6% to 70.3%; I²: 100%) in non-smokers (Table 2, S3 Table), indicating a greater numerical decrease in ODI in non-smokers. Furthermore, the sub-analysis performed in two [33,36] out of six studies showed incrementally increased percent mean changes from smokers: 30.4% (95% CI: 17.6% to 43.2%; I²: 100%), to former smokers: 39.9% (95% CI: 27.0% to 52.7; I²: 100%), and finally never smokers: 47.7% (95% CI: 34.8% to 60.5%; I²: 100%) (Table 3, S4 Table). Two studies distinguished the number of levels fused with only one study exclusively studying single-level fusion and the other multi-level fusion; stratification revealed similar results, that is, apparent better ODI scores amongst non-smokers (Table 2).

VAS neck pain

Neck pain was recorded using VAS (cervical spine) in six articles [44–46,49–51]. The pooled percent mean change from baseline was 53.3% (95% CI: 39.6% to 66.9%; I²: 100%) for smokers and 49.6% (95% CI: 35.9% to 63.2%; I²: 100%) for non-smokers (Table 2; S5 Table). Non-smokers were divided into former-smokers and never-smokers in three [44,49,51] out of the six studies, yielding 56.8% (95% CI: 35.2%to 78.5%; I²: 100%) for smokers, 54.5% (95% CI: 32.9% to 76.1%; I²: 100%) for former-smokers, and 53.8% (95% CI: 32.1% to 75.4%; I²: 100%) for never-smokers (Table 3, S6 Table). One study limited their data to patients with fusion across multiple levels and upon stratification, demonstrating that, similar to the pooled results, there was a comparable change in the VAS neck pain scores (Table 2).

VAS arm pain

VAS arm pain was assessed in five studies focusing on cervical spine [44,46,49–51]. The pooled percent mean changes were 53.6% (95% CI: 43.6% to 63.5%; I²: 100%) for smokers and 54.7% (95% CI: 44.8% to 64.4%; I²: 100%) for non-smokers (Table 2; S7 Table). On further stratification of non-smokers into former-smokers and never-smokers for three [44,49,51] out of the five studies, a comparable result between groups was obtained: a decrease of 59.8% (95% CI: 41.8% to 77.8%; I²: 99.9%) for smokers, 61.9% (95% CI: 43.8% to 79.9%; I²: 100%) for former-smokers, and 54.8% (95% CI: 36.8% to 72.9%; I²: 100%) for never-smokers (Table 3, S8 Table). Only one study exclusively evaluated multi-level fusions and found that non-smokers showed a larger decrease in arm pain (Table 2).

VAS back pain

Clinical outcomes on back pain scores were assessed in five articles [32–34,36,54] that studied lumbar spine surgeries. Pooled percent mean change from baseline was 55.0% (95% CI: 29.7% to 80.2%; I²: 100%) for smokers, and 60.5% (95% CI: 35.2% to 85.7%; I²: 100%) for non-smokers (Table 2; S9 Table). Further subgrouping for two [33,36] out of the five studies resulted in pooled percent mean changes of 40.3% (95% CI: 29.8% to 50.7%; I²: 100%) for smokers, 42.8% (95% CI: 32.3% to 53.2%; I²: 100%) for former-smokers, and 51.6% (95% CI: 41.1% to 62.1%; I²: 100%) for never-smokers (Table 3; S10 Table). One study evaluated multi-level fusions only, demonstrating similar results. Another study evaluated single-level fusions only and revealed that non-smokers showed more improvement than smokers, without overlapping confidence intervals. (Table 2).

VAS leg pain

VAS leg pain was assessed in three articles (lumbar spine) [32,33,36]. Pooled percent mean change from baseline was 70.8% (95% CI: 42.3% to 99.3%; I²: 100%) for smokers and a comparable 70.6% (95% CI: 42.1% to 99.1%; I²: 100%) for non-smokers (Table 2; S11 Table). Two [33,36] out of the three studies provided data on three categories of smoking history. Further stratification revealed worse values for smokers 38.1% (95% CI: 26.0% to 50.2%; I²: 100%) and former-smokers 54.6% (95% CI: 42.4% to 66.7%; I²: 100%) compared to never-smokers 62.2% (95% CI: 50.0% to 74.3%; I²: 100%) (Table 3; S12 Table). It should be noted that one of the studies did not have enough data available to be utilized in the sub-analysis; this study demonstrated a larger decrease in the outcome (over 80%) compared to the other two studies and had a smaller sample size. This discrepancy should be considered when interpreting the subgroup results. One study also evaluated single-level fusion only and revealed that non-smokers displayed a greater improvement in VAS leg pain, without overlapping confidence intervals. (Table 2).

JOA score

The percent mean change from baseline for JOA scores was derived from four studies (cervical spine) [45,49–51]. The pooled percent mean change was 25.3% (95% CI: 8.4% to 42.2%; I²: 100%) for smokers and 21.4% (95% CI: 4.5% to 38.3%; I²: 100%) for non-smokers Table 2; S13 Table. Further stratification by two [49,51] out of the four studies revealed values of 26.9% (95% CI: −19.9% to 73.7%; I²: 100%) for smokers, 22.6% (95% CI: −24.2% to 69.3%; I²: 100%) for former-smokers, and 27.2% (95% CI: −19.6% to 74.0%; I²: 100%) for never-smokers (Table 3; S14 Table). Stratification based on the number of levels operated on revealed one study that assessed multi-level fusions, demonstrating a trend similar to the overall pooled mean difference (Table 2).

Fusion

Nine articles [28,30–32,34,35,42,51,53] described fusion rates; the pooled incidence was 86.8% (95% CI: 75.9%−93.2%; I²: 77.8%) in smokers vs. 95.1% (95% CI: 91.0%−97.4%; I²: 78.5%) in non-smokers (S6 Appendix). Stratification based on the region, that is, cervical (seven studies), lumbar (two studies), and number of levels operated on, that is, single level (two studies) and multi-level (six studies), displayed the same trend of a higher number of patients demonstrating fusion in non-smokers (Table 2).

Pseudoarthrosis

Likewise, nine articles [10,13,29,32,34,40,42,44,47] described the occurrence of pseudoarthrosis. The pooled incidence of pseudoarthrosis was 17.2% (95% CI: 11.7% to 24.4%; I²: 73.6%) in smokers, 7.3% (95% CI: 4.9% to 10.7%; I²: 64.2%) in non-smokers (S7 Appendix). Subgroup analyses by two [40,44] out of the nine studies yielded a pooled incidence of pseudoarthrosis in 9.5% of participants who were smokers (95% CI: 3.2% to 25.0%; I²: 73.6%), 8.5% of former-smokers (95% CI: 3.5% to 18.8%; I²: 0.00%), and 7.2% of never-smokers (95% CI: 3.3% to 15.0%; I²: 36.2%) (S8 Appendix). Studies focusing exclusively on multi-level fusion demonstrated more pseudoarthrosis in smokers, but a difference was absent in studies focusing on mono-level fusion. Upon stratification based on regions, comparable results were shown for smokers and non-smokers (Table 2).

Patient satisfaction

Patient satisfaction was analyzed in two studies [10,47] as a mean score (continuous outcome) between smokers and non-smokers. The pooled post-surgical mean score for patients’ satisfaction was 2.35 (95% CI: 1.51 to 3.19; I²: 96.3%) for smokers and 2.05 (95% CI: 1.21 to 2.88; I²: 97.6%) for non-smokers (S9 Appendix). Another two studies [32,54] reported patient satisfaction as a binary outcome (1 and 2: good result, 3–7: bad result) for smokers and non-smokers. Surprisingly, the pooled incidence for the good results was 86% (95% CI: 65% to 91%; I²: 71.4%) in smokers and 91% (95% CI: 79% to 97%; I²: 64.1%) in non-smokers (S10 Appendix).

Surgical site infection

Seven articles [13,31,40,43,45,47,50] accounted for surgical site infections (SSI). The pooled incidence of SSI was 4.6% in smokers (95% CI: 2.5%−8.5%; I²: 22.5%) vs. 2.6% (95% CI: 1.6%−4.3%; I²: 6.00%) in non-smokers (S11 Appendix). The pooled incidences of SSI in non-smokers remained relatively lower than in smokers for each of the multi-level groups, cervical spine group, and lumbar spine group (Table 2).

Reoperations

Four studies [28,43,44,50] provided data on reoperations. The pooled incidence of reoperations was 10.4% (95% CI: 4.1%−24.1%; I²: 77.7%) in smokers vs. 7.5% (95% CI: 3.2%−16.7%; I²: 53.46%) in non-smokers (S12 Appendix). One study evaluating reoperations after lumbar spine fusion surgery demonstrated convincingly more reoperations in smokers (Table 2).

SF-12

Amongst two studies [33,47], the change in SF-12 between smokers and non-smokers were comparable. The mean post-operative SF-12 score was 38.07 (95% CI: 35.84 to 40.30; I²: 0.00%) for smokers and 40.31 (95% CI: 39.14 to 41.47; I²: 44.6%) for non-smokers (Table 2; (S13 Appendix). A sub-analysis could not be done for this outcome measure due to a lack of stratified results among studies.

Across outcomes, both smokers and non-smokers demonstrated clinically meaningful improvements in pain and function; however, between smokers and non-smokers, the differences were generally small and often not clinically significant. For VAS back and leg pain, ODI, NDI, and JOA scores, smokers and non-smokers achieved comparable improvements, with subgroup differences showing wide confidence intervals and uncertain clinical importance. In contrast, structural outcomes demonstrated more clinically relevant differences. For SSI and reoperations, the absolute differences were smaller but still potentially meaningful given the morbidity of these events.

Limitations and strengths

Notable limitations were variations evident in baseline characteristics among the groups being compared in the studies (smokers, non-smokers), introducing a confounding element that limited our ability to determine the unbiased association between smoking and adverse clinical outcomes. Such variations can be attributed to poor study designs wherein patient cohorts were not matched, nor were there any attempts to adjust for confounding. There was also a significant heterogeneity across most studies, making it difficult to identify trends and associations between different outcome measures and smokers; however, we purposefully did not conduct a statistical analysis that directly compared smokers to non-smokers because of the lack of adjustment for confounding in the original studies. Moreover, to address this heterogeneity, we adhered to robust statistical techniques and performed sub-analyses by distinguishing between the regions where the fusion surgeries were performed and the number of levels that were operated on. In some cases, this reduced the heterogeneity and allowed us to draw more concrete conclusions. Additional data on the timeline of smoking cessation and pack-years of former/current smokers were not available in any of the studies. This may also have impacted the results.

Furthermore, not all of the included studies undertook the distinction between former-smokers and never-smokers within the non-smoker category, which accounted for a lower number of studies for comparison in the sub-analysis. In addition to this, information about the extent of smoking within both the current smoker and former smoker categories in terms of pack-years and time elapsed between smoking cessation and surgery was unavailable. This information could potentially provide more insight into the disparity in outcomes across included studies. Such information may also hold value in guiding personalized mitigation strategies tailored toward patients based on their individual smoking history. Nevertheless, our sub-analysis revealed a hierarchical association between patients with a history of smoking (whether they be former or current smokers) and clinical outcomes.

Despite these limitations, our meta-analysis had several strengths. We calculated the percent mean change from baseline for several outcomes, and we pooled these results, instead of calculating the absolute mean difference, which does not account for the different baseline values across studies. Additionally, in an attempt to reduce heterogeneity and make the most sense of the relation between smoking status and the different clinical outcomes, we stratified our pooled results in several ways: I) By two smoking categories (smokers vs. non-smokers); II) Within each of these categories, we further stratified by different regions of operation and the number of levels operated on; III) We then further stratified the non-smokers into former smokers and never smokers, when such data were provided; steps that were not taken in previously conducted meta-analyses. These statistical methods helped disentangle some of the associations between smoking status and clinical outcomes.