Relationship between smoking and postoperative complications of cervical spine surgery: a systematic review and meta-analysis

To determine whether smoking has adverse effects on postoperative complications following spine cervical surgery (PROSPERO 2021: CRD42021269648). We searched PubMed, Embase, Cochrane Library, and Web of Science through 13 July 2021 for cohort and case–control studies that investigated the effect of smoking on postoperative complications after cervical spine surgery. Two researchers independently screened the studies and extracted data according to the selection criteria. The meta-analysis included 43 studies, including 27 case–control studies and 16 cohort studies, with 10,020 patients. Pooled estimates showed that smoking was associated with overall postoperative complications (effect estimate [ES] = 1.99, 95% confidence interval [CI]: 1.62–2.44, p < 0.0001), respiratory complications (ES = 2.70, 95% CI: 1.62–4.49, p < 0.0001), reoperation (ES = 2.06, 95% CI: 1.50–2.81, p < 0.0001), dysphagia (ES = 1.49, 95% CI: 1.06–2.10, p = 0.022), wound infection (ES = 3.21, 95% CI: 1.62–6.36, p = 0.001), and axial neck pain (ES = 1.98, 95% CI: 1.25–3.12, p = 0.003). There were no significant differences between the smoking and nonsmoking groups in terms of fusion (ES = 0.97, 95% CI: 0.94–1.00, p = 0.0097), operation time (weighted mean difference [WMD] = 0.08, 95% CI: −5.54 to 5.71, p = 0.977), estimated blood loss (WMD = −5.31, 95% CI: −148.83 to 139.22, p = 0.943), length of hospital stay (WMD = 1.01, 95% CI: −2.17 to 4.20, p = 0.534), Visual Analog Scale-neck pain score (WMD = −0.19, 95% CI: −1.19 to 0.81, p = 0.707), Visual Analog Scale-arm pain score (WMD = −0.50, 95% CI: −1.53 to 0.53, p = 0.343), Neck Disability Index score (WMD = 11.46, 95% CI: −3.83 to 26.76, p = 0.142), or Japanese Orthopedic Association Scores (WMD = −1.75, 95% CI: −5.27 to 1.78, p = 0.332). Compared with nonsmokers, smokers seem to be more significantly associated with overall complications, respiratory complications, reoperation, longer hospital stay, dysphagia, wound infection and axial neck pain after cervical spine surgery. It is essential to provide timely smoking cessation advice and explanation to patients before elective cervical spine surgery.


Data extraction. Data extraction was conducted according to the Preferred Reporting Items for Systematic
Reviews and Meta-Analyses (PRISMA) statement, and the selection of articles and the extraction of data were carried out independently by two reviewers and examined by other authors. Any disagreements were resolved by consensus or discussion with a third reviewer. The following information was extracted from the studies: (1) the general study information (name of the first author, publishing date, country, study design, sample size, age, sex, surgical procedure, follow-up time, and definition of smoking); (2) perioperative parameters, including operative time, estimated blood loss, and length of hospital stay; (3) clinical outcomes, including visual analog scale (VAS) scores of neck pain and arm pain, Neck Disability Index (NDI) score, and Japanese Orthopedic Association Scores (JOA); (4) complications, fusion and reoperation; the complications were defined as primary outcomes in this study, including dysphagia, airway obstruction, nerve palsy, reintubation, axial neck pain, wound infection, deep venous thrombosis, pneumonia, and pseudarthrosis. For continuous outcomes, we extracted the mean and standard deviation, and participant numbers were also extracted. For dichotomous outcomes, we extracted the total numbers and the numbers of events of both groups. The data in other forms was recalculated when possible to enable pooled analysis.
Methodological quality. Reviewers applied the Newcastle-Ottawa Scale (NOS) to evaluate the methodological quality of the included studies 24 . The NOS is a scoring checklist for solving design and implementation issues of a cohort or case-control study, which consisted of participant selection, comparability of cases and controls, exposure, and outcomes. If the study was awarded six or more stars, it was considered a high-quality study and was analyzed.
Statistical analysis. We used STATA version 12.0 (StataCorp, College Station, TX) to generate forest plots to determine whether there was a statistical association between the case and control groups and to assess heterogeneity of the included studies. Dichotomous outcomes were expressed as effect estimates (ESs) with 95% confi- www.nature.com/scientificreports/ dence intervals (CIs); among them, the results of case-control studies are expressed as odds ratios (ORs), and the results of cohort studies are expressed as relative risks (RRs); continuous outcomes are expressed as the weighted mean differences (WMDs). Heterogeneity was quantified using the chi-square based Cochran's Q statistic 25 and the I 2 statistic, which yields results ranged from 0 to 100% (I 2 = 0-25%, no heterogeneity; I 2 = 25-50%, moderate heterogeneity; I 2 = 50-75%, large heterogeneity; and I 2 = 75-100%, extreme heterogeneity) 26 . In cases of substantial heterogeneity, the random-effects model was applied. Otherwise, the fixed-effects model was used. When heterogeneity was present, a 'leave-one-out sensitivity analysis was performed by iteratively removing one study at a time to confirm the source of the heterogeneity. Analysis was then performed without the study to determine if heterogeneity was still present and if so, random-effects modeling was used. Publication bias was assessed using visual inspection of the funnel plot with the Begg 27 and Egger tests 28 . All statistical tests were two-sided, and p-values of < 0.05 were considered statistically significant.

Results
Identification of eligible studies. A flowchart of the search and study selection process is shown in Fig. 1.
The electronic search identified a total of 352 citations (69 from PubMed, 212 from EMBASE, 20 from the Cochrane Library, and 51 from the Web of Science). After screening titles and abstracts and removal of duplicates, 122 were considered of interest; the full text of these 122 studies was retrieved for detailed evaluation; 79 studies were excluded, and 43 studies were ultimately included in the meta-analysis 2-4,7,9-16,18-21,29-55 . Characteristics of included studies. The characteristics of the studies are summarized in Table 1. The 43 independent observational studies included in this meta-analysis were published from 1995 to 2021. These forty-three studies included 10,020 patients, including 3,107 smokers and 6,913 nonsmokers. Twenty-seven studies were conducted in the United States and seven were conducted in China. The other nine were conducted in India, Japan, the Czech Republic, Italy, Korea, Singapore, and Taiwan. Of these, 16 were cohort studies, and 27 were case-controls. All raw data are available in the Supplementary Tables S1 and S2.
Quality of included studies. Because all the included studies were cohort studies or case-control studies, the quality of each study was evaluated using the NOS (maximum of nine stars). Case-control studies were divided into three categories: selection, comparability, and exposure, and cohort studies were divided into three categories: selection, comparability, and outcomes. According to the NOS scale, all included studies were considered to be of high-quality: 12 were awarded eight stars, 10 were awarded seven stars, and 5 were awarded six   Table 2). Nine were awarded eight stars, 3 were awarded seven stars, and 4 were awarded six stars in cohort studies (Table 3). www.nature.com/scientificreports/ patients who had exposure of C4 or above compared with other studies, the heterogeneity was reduced to 31.2% (Fig. 3). Fixed-effects modeling showed that smokers were significantly more associated with respiratory complications than nonsmokers (ES = 2.70, 95% CI: 1.62-4.49, p < 0.0001).   46 , we defined the sum of current and former smokers as the total number of smokers. We then removed Mangan et al., performed a sensitivity analysis, and found that heterogeneity was reduced to 41.4% (Fig. 4). Reanalysis using a fixed-effects model revealed that smokers were significantly more associated with reoperation after cervical spine surgery than nonsmokers (ES = 2.06, 95% CI: 1.50-2.81, p < 0.001).
Operation time. The operation time was provided in two studies 21,48 . No significant heterogeneity was observed, and a fixed-effects model was used (I 2 = 0.0%, p = 0.955). Pooling of the results revealed no significant difference in operation time after cervical spine surgery between smokers and nonsmokers (WMD = 0.08, 95% CI: −5.54 to 5.71, p = 0.955; Supplementary Fig. S1a).
Estimated blood loss. The estimated blood loss was provided in three studies 2,11,48 . Significant heterogeneity was observed, and a random-effects model was used (I 2 = 66.1%, p = 0.053). Pooling of the results revealed no significant difference in estimated blood loss after cervical spine surgery between smokers and nonsmokers (WMD = −5.31, 95% CI: −148.83 to 139.22, p = 0.943; Supplementary Fig. S1b). After performing leave-one-out sensitivity analysis, the heterogeneity did not change substantially and remained significant.
Length of hospital stay. The length of hospital stay was provided in four studies 2,11,21,48 . Significant heterogeneity was observed, and a random-effects model was used (I 2 = 88.3%, p < 0.0001). Pooling of the results revealed no significant difference in the length of hospital stay after cervical spine surgery between smokers and nonsmokers (WMD = 1.01, 95% CI: −2.17 to 4.20, p = 0.534; Supplementary Fig. S1c) . After performing leave-one-out sensitivity analysis, the heterogeneity did not change substantially and remained significant.  VAS: arm pain. VAS-arm pain was reported in two studies 18,48 . No significant heterogeneity was observed, and a fixed-effects model was used (I 2 = 0.0%, p = 1.000). Pooling of the results revealed no significant difference in VAS-arm pain after cervical spine surgery between smokers and nonsmokers (WMD = −0.50, 95% CI: −1.53 to 0.53, p = 0.343; Supplementary Fig. S1e).
NDI. NDI was reported in four studies 18,19,21,48 . Significant heterogeneity was observed, and a random-effects model was used (I 2 = 96.4%, p < 0.0001). Pooling of the results revealed no significant difference in NDI after cervical spine surgery between smokers and nonsmokers (WMD = 11.46, 95% CI: −3.83 to 26.76, p = 0.142; Supplementary Fig. S1f.). After performing leave-one-out sensitivity analysis, the heterogeneity did not change substantially and remained significant.
JOA. JOA was reported in two studies 18,21 . Significant heterogeneity was observed, and a random-effects model was used (I 2 = 89.4%, p = 0.002). Pooling of the results revealed no significant difference in JOA after cervical spine surgery between smokers and nonsmokers (WMD = −1.75, 95% CI: −5.27 to 1.78, p = 0.332; Supplementary Fig. S1g). Each specific result can be found in Table 4.

Subgroup analysis.
For primary outcomes, we conducted subgroup analysis based on the type of study.  Supplementary  Fig. S2A, S2B).

Discussion
The major purpose of the present meta-analysis was to determine whether smoking has adverse effects on surgical outcomes after cervical spine surgery. Our results suggest that smoking is associated with reoperation and postoperative complications, including dysphagia, axial neck pain, and wound infection. Compared with nonsmokers, smokers were more associated with overall postoperative complications and respiratory complications. There were no significant differences between smokers and nonsmokers concerning outcomes, including www.nature.com/scientificreports/ fusion, operation time, estimated blood loss, length of hospital stay, VAS-neck pain score, VAS-arm pain score, NDI score, or JOA score. Our results suggest that smoking might have adverse effects on surgical outcomes in patients who undergo cervical spine surgery.
Complications were the primary outcomes used to evaluate the safety of cervical spine surgery among smoking patients. Siemionow et al. conducted a study of 35 patients undergoing anterior and posterior cervical decompression and fusion and reported that smoking appeared to be the most critical factor related to perioperative complications; the risks for at least one perioperative complication were 50% and 31.6% for smokers and nonsmokers, respectively 7 . Lau et al. studied 160 patients undergoing anterior cervical corpectomy and found that smoking patients had longer hospital stays, more bleeding, a higher rate of pseudarthrosis, and more complications at 30 days than nonsmoking patients 11 . In contrast, Fehlings et al. analyzed data from the AOSpine North America Cervical Spondylotic Myelopathy Study and concluded that perioperative complications were not associated with smoking status 57 . Medvedev et al. reported the complication rates in smoking and nonsmoking patients of 23.5% and 39.8% (p < 0.0001), respectively 22 . Our pooled data showed that smoking was associated with increased postoperative complications, including dysphagia, airway obstruction, nerve palsy, reintubation, axial neck pain, wound infection, deep venous thrombosis and pneumonia.
We assessed perioperative outcomes, including fusion, operation time, estimated blood loss, and length of hospital stay in our meta-analysis and failed to find any significant difference between the smoking and nonsmoking groups. As measured by NDI, JOA, and VAS scores, functional recovery was similar between the two groups. This finding indicates that cervical spine surgery might offer similar functional outcomes in smoking patients. However, only two articles reported VAS-neck pain and JOA scores, one study found that smoking improved both VAS-neck pain and JOA scores, while the other found the opposite. Therefore, more articles can improve the accuracy of the conclusion, and the relatively small sample size limited the generalizability of this conclusion.
After cervical spine surgery, smokers were closely associated with reoperation. In this meta-analysis, given that functional improvement between the groups was similar, it is possible that reoperation was directly related to complications in smoking patients, including wound infection, respiratory complications, and pseudarthrosis. However, due to limited data, we did not perform a subgroup analysis based on the type of surgical procedure.
There are several potential explanations for the observed association between smoking and adverse effects on the surgical outcomes for patients after cervical spine surgery. First, cigarette smoke products have been shown to inhibit prostacyclin production, a potent vasodilator, and an inhibitor of platelet aggregation. This effect can lead to impaired blood flow and increased blood viscosity, which result in impaired blood supply [58][59][60][61][62] , and leads to decreased angiogenesis and epithelialization 63 . Moreover, inhibition of revascularization by nicotine was observed in a rabbit study, and this mechanism may retard cellular metabolism and promote tissue degeneration 64 . Second, at the cellular level, nicotine has been shown to inhibit proliferation, differentiation, and collagen synthesis in osteoblasts 65 , which is the primary determinant of the tensile strength of a surgical wound 66 . Free radicals produced by burning cigarettes have been associated with cell membrane destabilization, impaired osteoblast mitochondrial oxidative function and local tissue hypoxia 58,[67][68][69][70][71] .
Third, it is well-documented that smoking harms bone physiology, which result in decreased bone mineral density, impaired bone metabolism, and accelerated osteoporosis, with produces lower fusion rates 72 . Animal and in vitro studies found that nicotine impaired bone healing, retarded bone formation and growth, and decreased graft biomechanical properties 73,74 .
Finally, cigarette smoke contains many toxic ingredients. Nicotine, tar, and other components irritate mucous membranes of the respiratory tract and cause cilia of bronchial epithelial cells to become shorter and irregular, which can hinder the movement of ciliary bodies, reduce local resistance, and weaken phagocytosis and sterilization functions of alveolar phagocytes, which leads to bronchospasm and increased airway resistance 75 . For these www.nature.com/scientificreports/ reasons, smokers are susceptible to respiratory complications after cervical spine surgery. In addition, carbon monoxide combines with hemoglobin, which reduces the oxygen-carrying capacity of the blood, and hydrogen cyanide inhibits cytochrome c, and leads to inhibition of aerobic metabolism 76 .
To the best of our knowledge, our meta-analysis, on the basis of 16 cohort studies and 27 case-control studies, is the first, also the largest and most comprehensive assessment to investigate the association between smoking and outcomes of cervical spine surgery. The main strength of this systematic review and meta-analysis is the thorough literature search, careful study selection with strict inclusion criteria, and comprehensive assessment of methodological quality of included studies using the NOS, which is, currently, the accepted standard. In addition, we performed subgroup analysis according to the type of study for the primary outcomes. Although we found significant heterogeneity in several outcomes among the included studies, the sensitivity analysis showed no significant change, which suggested that the pooled estimate in our study was stable. Finally, publication bias was quantitatively evaluated using Begg's and Egger's linear regression tests.
This systematic review and meta-analysis have several limitations that are worthy of comment. First, studies included in our review spanned over two decades (1995 to 2021), during which advancements in cervical surgery techniques might have improved outcomes. Despite this, point estimates for earlier and more recent studies were similar. Second, all of the included studies were retrospective observational trials rather than randomized controlled trials. The inherent nature of observational trials may be associated with selective bias, which might have influenced our results. Third, in most studies, the definition of smoking was not standardized, and self-reporting introduces recall bias or response bias because nonsmokers may be current or former smokers. Therefore, the true impact of smoking may be larger than we have reported here. Moreover, the definition of complications was not uniform and might introduce an additional source of bias. Fourth, since most of the information collected was not used to answer specific questions, all characteristics of smoker and nonsmoker cohorts such as age, sex, BMI, ethnic group, indications for surgery, and comorbidities, were not necessarily consistently matched, leaving some possible residual confusion, resulting in high heterogeneity. Moreover, due to the limited number of articles, we did not compare the various types of cervical spine surgeries in detail. In addition, only two studies reported operation time, VAS-neck pain, VAS-arm pain and JOA, and only three studies reported estimated blood loss and axial neck pain. Their results were based on a very small number of studies, which may lack reference value. Finally, we do not know how investigators confirmed that their patients did not smoke before or after surgery or even if they quit smoking before surgery, which may have impacted the evaluated results.
One study analyzed the pack-year history and found that, after lumbar surgery, nicotine exposure was associated with an increased risk of disease, and there was a dose-response trend; however, this trend was not significant 77 . In contrary, another study did not support this view and found that after anterior cervical discectomy and fusion, pack-years were not significantly associated with greater odds of developing any one complication or any major complication 78 . This may be related to differences in the number, characteristics, surgical sites, and follow-up time of the population included in the study. Therefore, there is an urgent need for further high-quality studies that are sufficiently prepared and designed with sufficient detail to adjust for multiple confounders and allow exploration of dose-response relationships.
Some researchers reported that preoperative smoking cessation might improve surgery outcomes and could lower medical costs by decreasing postoperative complications and length of post surgical hospital stay among smokers 11,79 . Sørensen et al. performed a meta-analysis and found that smoking cessation reduced the risk of surgical site infection in plastic and general surgery patients by more than half 80 . Andersen et al. found that quitting smoking significantly increased the rate of fusion after spinal surgery compared to those who continued to smoke, bringing it close to the level of nonsmokers 81 . This may be related to the rapid recovery of local tissue oxygenation and metabolism after smoking cessation 82 . Therefore, it is theoretically necessary to quit smoking before elective surgery.
Nevertheless, the optimal timing to quit smoking remains a matter of considerable debate. A study showed that quitting smoking 1 to 2 months before surgery can significantly reduce the perioperative risk 77 . Some studies indicated that smoking cessation must be at least 4 weeks before surgery to be effective 83,84 . Another study said that smoking cessation should be carried out at least 2 months before coronary artery bypass to maximize the reduction of postoperative respiratory complications 85 . Jung et al. found that preoperative smoking cessation for at least 2 weeks will help to reduce the incidence of postoperative complications in gastric cancer surgery 86 . Thus, exploring the optimal timing to quit smoking before the operation should determine future efforts.

Conclusions
Compared with nonsmokers, smokers seem to be more significantly associated with overall complications, respiratory complications, reoperation, longer hospital stay, dysphagia, wound infection and axial neck pain after cervical spine surgery. Our results suggest that smoking is closely related to adverse consequences after cervical spine surgeries. It is crucial to provide timely smoking cessation advice and explanation to patients before elective cervical surgery.

Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information files.