Wound Healing and Infection in Surgery:  The Clinical Impact of Smoking and Smoking Cessation: A Systematic Review and Meta-analysis FREE

For the past decades, a growing amount of literature has shown that smoking has a negative effect on postoperative outcome. A recent study disclosed that postoperative mortality and morbidity in smokers are substantial.¹

Until now, no general survey on the clinical impact of smoking on postoperative healing has been published, and the literature is dispersed across operations and surgical specialties. The evidence on the impact of smoking cessation on healing complications is sparse, and only a few studies have assessed how long patients must be abstinent from smoking before surgery to reduce the risk. Accordingly, it is not clear if the effort, which is necessary to ensure successful abstinence from smoking, is worthwhile in terms of reducing healing complication rates. Recently published systematic reviews have disclosed that preoperative smoking cessation intervention reduces postoperative complications overall.^2- 3 However, these reviews assessed pooled postoperative outcome and did not address healing complications.

The aims of this systematic review were to describe the association between smoking and healing complications across all surgical specialties and to estimate the impact of perioperative smoking cessation on postoperative healing outcomes.

Computerized searches in the CENTRAL, MEDLINE, and EMBASE databases were performed under supervision from a Cochrane Collaboration information specialist to identify relevant studies (Table 1). In addition, a manual cross-reference search of all potentially eligible articles retrieved for full-text evaluation was undertaken. The searches and study retrieval were performed until May 2010 for cohort studies and January 2011 for randomized controlled trials (RCTs).

Cohort studies with 100 patients or more assessing healing complications in smokers and former smokers were included to ensure that a broad range of surgical procedures and healing complications were addressed. Studies assessing multiple operations or healing outcomes from the same patient cohort were included according to each specified operation or healing outcome.

Randomized controlled trials assessing the effect of perioperative smoking cessation on postoperative healing complications were included. This intervention embraced all types of behavioral or motivational counseling with or without pharmacotherapy. Only RCTs with a minimum of 1 week of preoperative intervention and assessment of healing outcome after specified elective surgical procedures were included. Randomized controlled trials with a dropout rate greater than 40% were excluded.

The outcome measures included all types of adverse healing events after surgical procedures with access through a skin incision. Short-term (necrosis of wound and tissue flaps, healing delay and dehiscence of wounds and sutured tissue, surgical site infections, and nonspecified wound complications) and long-term healing outcomes (hernias and lack of fistula or bone healing) were accessed.

Data from the cohort studies and RCTs were extracted according to the MOOSE (Meta-analysis of Observational Studies in Epidemiology)⁴ and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses)⁵ statements, respectively. The methodological quality of the cohort studies was evaluated in a domain-based evaluation process and by the Newcastle-Ottawa Scale, which is a scoring checklist assigning points (maximum, 9 stars) for patient selection characteristics, exposure ascertainment, comparability, and outcome assessment.⁶ The methodological quality of the RCTs, including risk of bias assessment, was assessed according to Cochrane Collaboration recommendations⁷ and the Jadad score⁸ for consideration of random sequence generation, allocation concealment, blinding procedures, address of incomplete outcome data, and unselective reporting. Publication bias was assessed by inspection of funnel plots calculated from meta-analyses including more than 10 studies.⁷

From each study crude incidence rates or adjusted odds ratios (ORs) were extracted. Based on these data, unadjusted and adjusted estimates were calculated by means of the Mantel-Haenszel and inverse variance methods, respectively. These estimates were included in separate meta-analyses of the cohort studies according to each type of healing complication measure (smokers compared with nonsmokers) and a combined healing complication measure (former smokers compared with patients who never smoked or current smokers). Unadjusted estimates were included in the meta-analysis of RCTs.

Sensitivity analyses were conducted across complication type and included cohort studies with a maximum Newcastle-Ottawa Scale score and more than 1000 patients. Because most of the RCTs assessed outcome by pooling healing complications, sensitivity analyses aimed to estimate the impact of perioperative smoking cessation on different types of healing complications.

The statistical heterogeneity of the studies was reported as an I² value in each meta-analysis. Different methods of analysis were applied to assess the pooled treatment effects. In the meta-analyses of cohort studies, the random-effects model was used irrespective of the I² value. In the analysis of RCTs, the fixed-effects model was applied in case of an I² value of less than 40%.⁷ The statistical analysis was performed with the use of the R program meta-analysis package, version 1.6-0.⁹ In all analyses, a threshold of P ≤ .05 was considered statistically significant.

The search for relevant studies yielded 6306 citations for cohort studies and 979 citations for RCTs (Figure 1).

One hundred forty cohort studies compared smokers and nonsmokers. The total number of patients included was 479 150. The studies originated from countries all over the world and embraced operations from all surgical specialties.

Twenty-six cohort studies assessed healing complications in former smokers. In 18 of these studies, former smokers were compared with current smokers and patients who never smoked. The studies originated from multiple countries and embraced general, thoracic, orthopedic, and plastic and reconstructive surgery. Half the studies defined former smokers as being abstinent from smoking for a median of 4 (range, 2-52) weeks before surgery,^10- 21 whereas the other half did not address the period of preoperative abstinence.^22- 35 In 2 studies, smokers were requested to quit smoking 4 weeks before the operation and then were compared with current smokers and patients who never smoked.^19,21

Four RCTs assessed the impact of perioperative smoking cessation intervention. The trials originated from Denmark and Sweden.^36- 39 The operations were elective orthopedic operations (hip and knee arthroplasty) and general surgical operations (herniotomy, cholecystectomy, and colorectal resection). The studies complied with the similar criteria for inclusion (ie, daily smoking, patients older than 18 years) and exclusion (ie, alcohol or other drug abuse, dementia, and lack of language proficiency). All studies reported the number of eligible patients, accounted for missing data, and discarded data from dropouts from the intention-to-treat analyses.

The intervention periods ranged from 2 to 3 weeks to 6 to 8 weeks before surgery until the day of skin suture removal or 30 days after surgery. Apart from 1 study,³⁷ the intervention was tailored individually and offered by study nurses professionally trained in smoking cessation therapy. The intensity ranged from brief advice with a follow-up telephone or outpatient reminder to multiple sessions of individual face-to-face counseling and unlimited hotline service access. Free-of-charge nicotine replacement drugs were offered by all but 1 study.³⁷ The control interventions ranged from standard advice about smoking and surgical outcome to a request to maintain daily smoking habits during the perioperative period.³⁸

All studies assessed self-reported smoking or abstinence at the day before surgery and at the day of outcome assessment. Biochemical validation was assessed by measurement of cotinine levels in saliva or carbon monoxide levels in expired air. Compliance to abstinence varied from 23% to 64%.

The clinical heterogeneity was considerable in the cohort studies, and a variety of methodological flaws were present across studies. These flaws included retrospective data collection, no report of missing data, detection bias due to inadequate outcome definition, attrition bias due to inadequate postdischarge follow-up reporting, and inadequate confounder control. All RCTs had a low risk of bias, and they achieved a maximum Jadad score.

Inspection of funnel plot symmetry disclosed that the publication bias of the cohort studies was generally low (data not shown). In the cohort studies assessing surgical site infection and wound complications, a discrepancy was found between studies reporting crude incidence rates and adjusted ORs, indicating that some degree of publication bias was present in these studies.

Necrosis of wounds and tissue flaps was assessed in 19 unique studies including 7616 (number of subjects per study, 111-1177) smokers and nonsmokers (eTable 1). Both meta-analyses disclosed a significantly higher incidence of necrosis in smokers (crude OR, 3.61 [95% CI, 2.78-4.68]) and adjusted OR, 3.60 [95% CI, 2.62-4.93]) (Table 2).

Most of the studies were conducted in patients undergoing breast surgery. Wound necrosis after mastectomy was 4-fold more frequent in smokers.^40- 41 Three studies assessing a dose-effect relationship between the intensity of smoking and necrosis found conflicting results.^13,20,42 Two studies reported a dose-effect relationship between lifelong smoking intensity (in pack-years) and necrosis.^17,43

In breast reconstructive surgery ranging from breast reduction to postmastectomy reconstruction, all studies demonstrated a high incidence of necrotic complications.^13,44- 50 Small retrospective studies of flap transposition or free-flap reconstruction after head and neck surgery found conflicting results.^35,51- 53 After lung cancer surgery and pelvic organ prolapse repair, fistulas caused by necrotic suture or mesh erosion were more frequent in smokers.^{17,43,54- 55}

Healing delay and dehiscence of wounds and tissue were assessed in 18 unique studies including 26 297 (number of subjects per study, 111-24 192) smokers and nonsmokers (eTable 2). Both meta-analyses disclosed a significantly higher incidence of healing delay and dehiscence of wounds and tissue in smokers (crude OR, 2.86 [95% CI, 1.49-5.49] and adjusted OR, 2.07 [95% CI, 1.53-2.81]) (Table 2).

Most studies assessing dehiscence of wounds, fascia, and sutured tissue, including anastomotic leakage, found a higher incidence in smokers.^{24,26,56- 63} Postoperative healing delay as an outcome measure was assessed in a few older cohort studies. In orthopedic surgery, the reamputation rate owing to failed healing showed conflicting results^34,64- 66; in breast reconstructive surgery, more recent studies did not find postoperative healing delay to be more frequent in smokers.^16,67

Surgical site infection was assessed in 51 unique studies including 408 428 (number of subjects per study, 100-163 824) smokers and nonsmokers (eTable 3). Both meta-analyses disclosed significantly more surgical site infections in smokers (crude OR, 2.12 [95% CI, 1.56-2.88] and adjusted OR, 1.79 [95% CI, 1.57-2.04]) (Table 2).

In general surgery, most of the studies found a higher surgical site infection in smokers.^{18,27- 28,32,42,60,63,68- 79} In 1 study,⁸⁰ smokers had more surgical site infections after intestinal and colon surgery, but not after gastrectomy. After coronary bypass surgery, sternal wound infection after coronary bypass surgery was more frequent in smokers compared with nonsmokers in most studies.^{10- 12,33,81- 92} In orthopedic and reconstructive surgery, all major studies found surgical site infection to be more frequent in smokers,^{13,16,29,44,46,66,93- 95} contrary to a few small studies.^96- 98 In gynecologic and obstetric surgery, conflicting results were found.^99- 101

Wound complications (nonspecified) were assessed in 31 unique studies including 22 516 (number of subjects per study, 102-6676) smokers and nonsmokers (eTable 4). Both meta-analyses disclosed significantly more wound complications in smokers (crude OR, 2.06 [95% CI, 1.60-2.65] and adjusted OR, 2.27 [95% CI, 1.82-2.84]) (Table 2).

All major studies in breast reconstructive surgery found smoking to predict wound complications.^{13,19- 20,102- 103} A number of smaller studies assessing wound complications after reconstructive surgery showed conflicting results.^{15- 16,19,45,67,102,104- 117} Similar conflicting results were found in larger and smaller cohort studies after orthopedic, obstetric, gastrointestinal tract, and head and neck surgery.^{23,25,94,118- 124}

Incisional or recurrent inguinal hernia was assessed in 9 unique studies including 2296 (number of subjects per study, 114-544) smokers and nonsmokers (eTable 5). The meta-analysis from studies^30,125- 130 reporting adjusted estimates found hernia to be more frequent in smokers (OR, 2.07 [95% CI, 1.23-3.47]), contrary to the meta-analysis based on studies^27,131 reporting crude incidence rates (OR, 2.21 [95% CI, 0.71-6.84]) (Table 2).

In general surgery and urology, most studies found hernia to be more frequent in smokers,^{27,30,127,130} contrary to studies of aortic reconstructive surgery, which showed conflicting results.^{125- 126,128- 129,131}

Lack of fistula and bone healing was assessed in 10 unique studies including 14 293 (number of subjects per study, 105-12 297) smokers and nonsmokers (eTable 6). Both meta-analyses disclosed a significantly higher incidence of lack of fistula and bone healing in smokers (crude OR, 2.21 [95% CI, 1.60-3.05] and adjusted OR, 2.44 [95% CI, 1.66-3.58]) (Table 2).

In a study of open tibial fracture repair, Adams et al¹³² found that smokers' fractures healed slower. All studies assessing long-term outcome after spinal surgery, except one,¹³³ found failed bone union to be more frequent in smokers.^{21,31,134- 136} In addition, unhealed sternocutaneous fistula and anal fistula were more frequent in smokers.^137- 138

Twenty-four unique studies reporting the outcome of 47 764 (number of subjects per study, 177-10 897) former smokers and patients who never smoked were included (eTable 7). Both meta-analyses disclosed significantly more combined healing complications in former smokers than in those who never smoked (crude OR, 1.30 [95% CI, 1.07-1.59] and adjusted OR, 1.31 [95% CI, 1.10-1.56]) (Table 3).

Twenty unique studies reporting the outcome of 40 629 (number of subjects per study, 177-10 897) former smokers and current smokers were included (eTable 8). Both meta-analyses disclosed significantly fewer healing complications in former smokers than in current smokers (crude OR, 0.69 [95% CI, 0.56-0.85]; adjusted OR, 0.28 [95% CI, 0.12-0.72]) (Table 3). Some of these studies assessed the effect of pack-years on healing complications in former smokers, but the results were conflicting.^{13,17,20,23,25,32}

The sensitivity analyses confirmed that smokers had significantly more healing complications than did nonsmokers across complication types (Table 2), contrary to former smokers compared with patients who never smoked, which disclosed a nonsignificant trend (Table 3).

Four RCTs reporting the outcome of 416 patients (number of subjects per study, 57-149) were included and healing complications were found in 15.9% (66 of 416) (Table 4). The meta-analysis disclosed that perioperative smoking cessation did not significantly reduce healing complications combined (Figure 2). In contrast, surgical site infections were significantly reduced by perioperative smoking cessation as shown by a sensitivity analysis (Figure 3).

This systematic review shows that smokers compared with nonsmokers and former smokers compared with those who never smoked have more postoperative healing complications. Former smokers (compared with current smokers) have fewer healing complications. Perioperative smoking cessation reduces surgical site infections, but not other healing complications.

Across cohort studies, necrosis was 4 times more frequent in smokers than nonsmokers, whereas surgical site infection, dehiscence, healing delay, hernia, and lack of fistula and bone healing occurred 2 times more frequently in smokers. The following pathophysiological mechanisms for defective healing in smokers appear to be involved: (1) an acute detrimental vasoactive effect of smoking leads to postoperative necrosis in tissues with fragile blood supply, such as reconstructive tissue flaps and colorectal anastomoses; (2) attenuation of the inflammatory healing response and impairment of oxidative bacterial killing mechanisms lead to surgical site infection; and (3) delay of the proliferative healing response and alteration of collagen metabolism lead to dehiscence, incisional hernia, and lack of fistula or bone healing.^139- 144

Former smokers had a one-third higher incidence of healing complications than did patients who never smoked, although the sensitivity analysis did not confirm the significance of this finding. The difference in complication rate probably reflects a sustained detrimental effect of previous smoking on postoperative healing, implying that former smokers seem to have a lifelong higher risk of healing complications than those who never smoked. The lower incidence of complications in former smokers compared with current smokers suggests that a beneficial effect of abstinence from smoking on healing mechanisms exists. The finding, however, should be interpreted carefully owing to methodological flaws and bias in the cohort studies.

The meta-analysis of the RCTs disclosed that perioperative smoking cessation intervention did not reduce pooled healing complications. This finding contrasts with 2 recent meta-analyses that disclosed that smoking cessation reduced postoperative complications overall.^2- 3 Most likely the reason is methodological because 3 of the 4 RCTs included clinically heterogeneous adverse healing events in a pooled healing complication measure.^36,38- 39 Wound hematomas, seromas, and subfascial collections after hip and knee arthroplasty were included as healing outcome, although none of these complications have been individiually proven to be associated with smoking.^13,145

Perioperative smoking cessation intervention including 4 to 8 weeks of preoperative abstinence from smoking significantly reduced surgical site infections. This finding suggests that the primary impact of smoking cessation on healing is a reduction in infectious healing complications as shown by Møller et al.³⁹ This finding was confirmed by a randomized study of healthy volunteers, which disclosed that 4 weeks of abstinence from smoking significantly reduced incisional wound infection.¹⁴⁶ However, in 2 of the included RCTs, 4 weeks of preoperative abstinence did not reduce surgical site infections significantly, most probably because the RCTs were underpowered.^36- 37 In one of the RCTs, 2 to 3 weeks of preoperative abstinence did not affect healing complications.³⁸

This is the first systematic review to examine the impact of smoking and smoking cessation on healing complications. Strengths of this methodological approach include an extensive search complying with validated search strategies and a systematic scoring of methodological quality and risk of bias assessment. The following limitations are related to methodological issues of the cohort studies: differences in design, inconsistent definitions of smoking, underreporting of smoking habits and lack of biochemical validation, inconsistent definitions of healing outcome, and unclear outcome assessment and follow-up. In addition, the lack of addressing missing data, including former smokers' recall bias for the exact time of smoking cessation, and conflicting clinical confounders to be considered restrict the validity of the cohort studies.^147- 150 Although homogeneous and with a low risk of bias, the included RCTs were small. Apart from 1 study³⁷ that studied healing complications as a secondary outcome measure, the actual patient number included in the other RCTs appeared to be smaller than that specified in the protocol. Consequently, the included RCTs seem to have been underpowered to show a difference in healing complications by smoking cessation.

Smokers have a higher incidence of infectious and noninfectious healing complications after surgery compared with nonsmokers across all surgical specialties. Former smokers appear to have a lifetime higher risk of healing complications compared with patients who never smoked. Smoking cessation for at least 4 weeks before surgery reduces surgical site infections, but not other healing complications. Patients should be encouraged to stop smoking at least 4 weeks before surgery to reduce the risk of surgical site infections.

Further cohort studies are needed to clarify the risk of former smokers for postoperative healing complications. Accordingly, valid data from a detailed smoking history including the period of abstinence from smoking should be included in future clinical database studies on surgical outcome.

Additional RCTs assessing the impact of perioperative smoking cessation on healing outcome are needed for definite confirmation. Because interventions on lifestyle changes afford a number of challenges, multicenter and large-scale RCTs using cluster randomization should be considered.