Mediastinitis in open heart surgery: a systematic review and meta-analysis of risk factors

Abstract

Objective. We aimed to summarize the evidence from observational studies examining the risk factors of the incidence of mediastinitis in open heart surgery. Design. The study was a systematic review and meta-analysis of cohorts and case-control studies. Material and methods. We searched the literature and 74 studies with at least one risk factor were identified. Both fixed and random effects models were used. Heterogeneity between studies was examined by subgroup and meta-regression analysis. Publication bias or small study effects were evaluated and corrected by limit meta-analysis. Results. When correcting for small study effect, presence of obesity as estimated from 43 studies had Odds Ratio OR = 2.26. (95% CI: 2.17–2.36). This risk was increasing with decreasing latitude of study place. Presence of diabetes mellitus from 63 studies carried an OR = 1.90 (95% CI: 1.59–2.27). Presence of Chronic Obstructive Pulmonary Disease (COPD) from 30 studies had an OR = 2.59 (95% CI: 2.22–2.85). Presence of bilateral intramammary graft (BIMA) from 23 studies carried an OR = 2.54 (95% CI: 2.07–3.13). This risk was increasing with increasing frequency of female patients in the study population. Conclusion. Evidence from this study showed the robustness of the risk factors in the pathogenesis of mediastinitis. Preventive measures can be implemented for reducing obesity, especially in lower latitude countries. Furthermore, it is mandatory to monitor perioperative hyperglycemias with continuous insulin infusion. Use of skeletonized BIMA carries higher risk of mediastinitis especially in female patients without evidence of beneficial effect on survival for the time being.

Keywords:

• Mediastinitis
• risk factors
• systematic review
• meta-analysis

Introduction

The reported incidence of mediastinitis after coronary artery bypasses grafting (CABG) is 0.4 to 4%. Mediastinitis is associated with increased morbidity, mortality and cost. The traditional management consists of reoperation with debridement and drainage, followed by prolonged hospitalisation for antibiotic therapy. Early or in- hospital mortality is reported to vary between 10 and 47%, and there is still an excess risk of mortality after 10 years of follow up [1]. Several risk factors have been identified in the literature, however there are some inconsistencies regarding the role of individual factors in the pathogenesis of sternal infection. Most factors include diabetes mellitus, obesity, chronic obstructive pulmonary disease (COPD) and the presence of bilateral intra mammary graft (BIMA). Also the presence of mediastinitis in open heart surgery carries an increasing risk of obstruction of intra mammary graft. The incidence and microbiology of mediastinitis are dependent of the latitude of the country [2,3] and age of the patient.

The aim of this study is to identify and quantify risk factors of mediastinitis, and to evaluate their robustness, and to pinpoint the influence of patient and study related characteristics on the incidence of mediastinitis in the medical literature.

Methods

Literature search

A qualified medical librarian was consulted at the Medical Library, Oslo University Hospital. We searched the Cochran Central Register of controlled trials (1970–2019), Medline/PubMed (1966–2019) and Embase (1980–2019). We used a combination of keywords, in the search procedure. The last update for research was done on January 2^th, 2019. No limitation on language was considered (Figure 1).

Figure 1. Flow chart of the literature searches for the systematic review of studies addressing risk factors of mediastinitis. COPD: Chronic Obstructive Pulmonary Disease; Obesity: BMI > 30; BIMA: Bilateral Intra Mammary grafts.

Studies selection

We included cohort studies, (prospective and retrospective), and case control studies on patients with open heart surgery (CABG and valvular heart replacement) and with at least one risk factor Diabetes, Obesity, COPD or BIMA. Two reviewers independently evaluated reports for eligibility. Disagreements were resolved by discussion. Finally, 74 studies were considered eligible for risk factors study [1,4–76]. Sixty three studies for the association diabetes and incidence of mediastinitis [1,4–14,16,17,19–27,30,31,33–37, 39–42,44–47,49–62,64–71,74,75]. Forty three studies were selected for studying the association between presence of obesity and incidence of mediastinitis, [1,4,5,7–10,12–14, 16–21,23,25,26,30–32,34–36,38,40,44–46,48,50,53,59–61, 66–68,71,73,75]. Also 30 studies for the association between presence of COPD and incidence of mediastinitis [1,5,7,9,12,13,16,17,20,21,23,27,29,30,34,37,38,43,50,52–55, 59–61,64,66,70,73], and 23 for the association between presence of BIMA and incidence of mediastinitis [5,8,9,11,15,17,20,24,28,36,44,52,54,56,60,61,63,67,70,72,74, 76,77].

Quality assessment of the cohort and case-control studies: the component approach

Two of the reviewers independently assessed the studies. For the purpose of critically appraising case-control and cohort studies we developed a check-list based on recommendations by experienced investigators [78].The poor reporting on important methodological details may be due to the fact that most papers were published before the STROBE statements [79]. Sanderson et al [80] have warned for the use of scale. Scales resulting in numerical summary of quality scores can introduce a bias when assessing study quality [81].

Data abstraction

Data regarding publication status, study design, patient-related characteristics, outcome methods, result and funding were extracted in duplicate on a standardized form according to an a priori protocol. Patient-related variables included mean age of the cohort, frequency of female gender, and frequency of CABG operation. The study-level variables included the nature of the cohort, prospective versus retrospective, the use of case-control versus cohort in the risk factors estimation, use of multivariate methods to control for confounders, latitude and longitude of the place of the study.

Endpoints considered

The primary endpoint was the incidence of mediastinitis. The diagnosis of post sternotomy mediastinitis was based upon the criteria established by the Centres for Disease Control and Prevention (CDC).

Risk Factors of mediastinitis as exposition were presence of obesity (BMI > 30), diabetes mellitus, Chronic Obstructive Pulmonary Disease (COPD) and Bilateral Intra Mammary Graft operation (BIMA) as compared to Single Intra Mammary Graft.

Quantitative data synthesis

Statistical pooling

Fixed and random effects model analyses were considered to calculate an overall effect. The DerSimonian Laird estimate [82] is used in the random effects model. The risk factors of mediastinitis (diabetes, obesity, COPD, and BIMA) were quantified by the OR and its 95% confidence intervals. The information required from each study is an appropriate measure of odds ratio, OR, and its variance, V, which can be calculated if the lower and upper confidence limits OR_L and OR_U are given.

Sources of heterogeneity, evaluation and quantification

Statistical heterogeneity among studies was assessed with Cochran’s Q test. The magnitude of heterogeneity was evaluated by the I² statistics which describes the proportion of total variation due to heterogeneity rather than chance. In order to investigate possible sources of heterogeneity, two different methods were used: subgroup analyses and meta-regression.

Evaluation of selection bias or small study effect

Publication bias is known to occur in meta-analyses. In order to assess potential publication bias or small-study effect we used three well established test of small study effect or selection bias; Begg and Mazumdar Rank correlation [83], Egger’s test of asymmetry [84], and Test of moment for asymmetry [85].

Adjusting for small study effect (publication bias) using the limit Meta-analysis method

A popular method for detecting and adjusting publication bias or small study effect is the “trim and fill” method [86]. This method is known to perform poorly in the presence of substantial between-study heterogeneity [87,88]. A regression-based approach is a promising alternative, namely the method of limit meta-analysis [89]. This method is robust against heterogeneity [90]. Power considerations in meta-analysis and meta-regression were following the recommendations of Borenstein M et al [91].

All statistical analyses were performed with R–Package–Meta [92]. We have followed the PRISMA guidelines for meta-analyses and systematic reviews of observational studies [93] in the present study.

Results

Trial flow

After identifying 362 references, many studies were excluded due to irrelevant content and duplicate publications, leaving 63 potentially eligible studies for diabetes, 43 for obesity, 30 for COPD, and 23 for BIMA (Figure 1).

Risk factors of mediastinitis

Obesity

The summarized pooled estimate of the association between obesity and incidence of mediastinitis in the 43 studies using the random effect model was OR = 2.16 95% CI (1.79–2.60) with a substantial heterogeneity 92.0% (Table 1). The funnel plot visually showed evidence of bias and small study effect. This was verified using Begg and Mazumdar Rank and confirmed by the Egger’s test. To control for selection bias and small study effect we considered the limit meta-analysis method because of substantial heterogeneity. The adjusted estimate was OR = 2.26 95% CI (2.17–2.30).One important study-level related variable identified was latitude (Figure 2) (Table 2), with increasing effect of obesity on the incidence of mediastinitis with decreasing latitude of study place. The amount of heterogeneity accounted for latitude was 56.7%

Figure 2. Obesity and risk of mediastinitis with latitude of studies places.

Table 1. Pooled estimate of total or for the association between risk factors and incidence of mediastinitis with stratification on study level variables using the random effect model.

Risk factor	Subdivision	N	OR (95% CI)	Tau^2	I^2	p-Value^a
Obesity; BMI > 30	All studies	43	2.16 (1.79–2.60)	0.2509	92
	Prospective cohort
	Yes	21	1.96 (1.58–2.43)	0.1352	79%	.2863
	No/Unclear	22	2.33 (1.86–2.91)	0.1627	82%
	Case control design
	Yes	6	1.97 (1.16–3.35)	0.2730	73%	.7110
	No	37	2.19 (1.78–2.64)	0.2719	93%
	Control for confounders
	Yes	14	2.17 (1.66–2.84)	0.1483	82%	.8952
	No	29	2.13 (1.77–2.56)	0.1326	77%
Diabetes mellitus	All studies	63	2.31 (2.4–2.49)	0.002	2.3%
	Prospective cohort
	Yes	34	2.48 (2.23–2.76)	0%	0%	.2180
	No/Unclear	29	2.24 (1.97–2.54)	0.0125
	Case control design
	Yes	10	2.32 (1.78–3.02)	0%	0%	.9363
	No	53	2.35 (2.15–2.56)	0.0107	11.5%
	Control for confounders
	Yes	24	2.45 (2.18–2.76)	0%	0%	.2165
	No	39	2.23 (2.02–2.46)	0.004	4.4%
Chronic Obstructive Pulmonary Disease	All studies	30	2.53 (2.11–3.01)	0.0272	35.2%
	Prospective cohort
	Yes	14	2.35 (1.84–3.0)	0.0418	21.3%	.4896
	No/Unclear	16	2.68 (2.03–3.54)	0.1137	44.8%
	Case control design
	Yes	7	2.59 (1.84–3.63)	0%	0%	.9059
	No	27	2.52 (2.01–3.16)	0.1096	47.4%
	Control for confounders
	Yes	6	2.26 (1.76–2.90)	0%	0%	.4197
	No	24	2.60 (2.06–3.27)	0.1150	43.6%
Bilateral Intra Mammary grafts	All studies	23	2.49 (2.05–3.03)	0.0792	39.4%
	Prospective cohort
	Yes	12	2.34 (1.83–3.17)	0.0945	52.3%	.3776
	No/Unclear	11	2.82 (2.04–3.89	0.0500	17.6%
	Case control design
	Yes	4	2.00 (1.50–2.65)	0	0%	.1441
	No	19	2.63 (2.08–3.32)	0.0985	42.6%
	Control for confounders
	Yes	7	2.74 (1.99–3.77)	0.0695	41.8%	.4727
	No	16	2.36 (1.83–3.04)	0.0961	39.8%

Notes: Tau²: inter-study variability.

^a Test for subgroup differences using random effect model between groups with Q statistics df = 1 and p-value.

Table 2. Estimate of the mixed-effect regression model between risk factors and incidence of mediastinitis, and the different study-level and patient-level variables.

Risk factor	Covariates	Level	β-coefficient	Std.err. (β)	Z	p-Value
Obesity; BMI > 30	Latitude of study	Degrees	−0.0121	0.0050	−2.43	.0149
	CABG	Frequency	0.0044	0.0054	0.82	.4147
	Age	Years	−0.0491	0.0397	−1.23	.2163
	Female	Frequency	−0.0024	0.0103	−0.23	.8131
Diabetes mellitus	Latitude of study	Degrees	0.0021	0.0036	0.57	.5632
	CABG	Frequency	0.0038	0.0025	1.49	.1351
	Age	Years	−0.0153	0.0167	−0.91	.3598
	Female	Frequency	−0.0016	0.0041	0.39	.6922
Chronic Obstructive Pulmonary Disease	Latitude of study	Degrees	−0.0062	0.0077	−0.80	.4187
	CABG	Frequency	0.0029	0.0061	0.47	.6337
	Age	Years	0.0007	0.0357	0.02	.9847
	Female	Frequency	−0.0086	0.0101	−0.84	.3975
Bilateral Intra Mammary grafts	Latitude of study	Degrees	−0.0007	0.0102	−0.06	.9469
	CABG	Frequency	−0.0042	0.0087	−0.47	.6325
	Age	Years	−0.0474	0.0344	−1.38	.1675
	Female	Frequency	0.0250	0.009	2.77	.0055

Diabetes mellitus

The summarized pooled estimate of the association between diabetes mellitus and incidence of mediastinitis in the 63 studies using the random effect model was OR = 2.31, 95% CI (2.14–2.49) with a little heterogeneity 2.3% (Table 1). The funnel plot visually showed evidence of bias and small study effect. This was confirmed by the Egger’s test, and the test of method. To control for selection bias and small study effect we considered the limit meta-analysis method because of substantial heterogeneity. The adjusted estimate was 95% CI OR = 1.90, (1.59–2.27). None of the patient level and study level covariates showed significance in the mixed effect model (Table 2).

Chronic obstructive pulmonary disease (COPD)

The summarized pooled estimate of the association between COPD and incidence of mediastinitis in the 30 studies using the random effect model was OR = 2.53, 95% CI (2.11–3.01) with a moderate heterogeneity 35.2% (Table 1). The funnel plot visually did not show evidence of selection bias or small study effect. This was verified using the three different tests of selection bias. No limit meta-analysis method was performed. None of the patient level and study level covariate showed significance using the mixed effect model (Table 2).

Bilateral intra mammary grafting (BIMA)

The summarized pooled estimate of the association between BIMA and incidence of mediastinitis in the 23 studies using the random effects model was OR = 2.49, 95% CI (2.05–3.03) with a substantial heterogeneity 39.4% (Table 1). No small study effect was demonstrated. No adjustment by limit meta-analysis was necessary. One important patient-level related variable identified was female gender (Figure 3) (Table 2) with increasing risk of mediastinitis with increasing frequency of female gender in the study. The amount of heterogeneity accounted for female gender was 58.7%

Figure 3. BIMA and risk of mediastinits versus frequency of female in the studies.

Discussion

Our systematic review and meta-analysis of risk factors of mediastinitis revealed robust risk factors. Obesity as an independent risk factor was decreasing in magnitude with increasing latitude of the study place. Also BIMA as a risk factor of mediastinitis was stronger in female as compared to male. There was substantial heterogeneity that was explained by some patients and study related characteristics and was investigated using metaregression. There was a presence of small study effect/publication bias and this was pinpointed and corrected by limit meta-analysis and not the trim and fill method because of substantial heterogeneity.

Strengths and limitations

Our review was based on a broad literature search, and it seems unlikely that we missed relevant studies [94]. The study selection, data extraction, and data assessment were done by two authors, to minimize bias and transcription errors [95]. The major limitations of our study were the quality of the studies, mainly the retrospective nature of the cohort and the non-random nature of the control group in case-control design, with poor control for confounders and no investigations of effect modification in both cohort and case control studies. There was variability in the diagnosis of obesity and no specification of the nature of diabetes mellitus as diabetes I or II in some studies. There was a possibility of misclassification of the risk factors between patients with mediastinitis and control or non-mediastinitis, but this could have been a non-differential misclassification diluting the magnitude of the risk factors. Small-study effect and selection bias represented a problem in our systematic review. This was corrected using the method of limit meta-analysis, and heterogeneity was investigated and quantified. The majority of studies had no a priori power analysis and some of them were power deficient.

Obesity as an independent risk factor of mediastinitis

The estimation of OR controlling for small study effect was 2.26, 95% CI (2.17–2.30) with substantial heterogeneity. This effect was robust to sensitivity analysis. As underlined by Falagas ME et al [96] the incidence of nosocomial infections in overweight and obese patients is increased compared with regular weight patients, usually because of prolonged duration of stay in hospitals, thus increasing their risk of infections. On the other hand, obesity was found to be associated with staphylococcus aureus nasal carriage, when adjusted for other variables. An aggressive regimen to lower obesity could be implemented before open heart surgery, especially in low latitude countries. On the other hand we pinpointed in our previous systematic review and meta-analysis [3] that latitude of study place is associated with temperature regulated bacterial virulence and is a marker of surgical local practice.

Diabetes mellitus as independent risk factor of mediastinitis

The estimation of OR controlling for small study effect was 1.90, 95% CI (1.59–2.27) with minimal heterogeneity. This result was robust when performing a sensitivity analysis.

This increased infection risk in diabetic patients can be imputed in part to physiologic modifications precipitated by deficient glucose control as underlined by Talbot TR [89]. Over time, chronic hyperglycaemia results in small vessel vasculopathy, contribute to local tissue hypoxia and ischemia, possibly diminished penetration of antibiotics to the surgical field for phagocytosis of invading bacteria. All these components contribute to an increased risk of mediastinitis in diabetic patients.

There is a gradient effect between level of perioperative serum glucose level and incidence of surgical site infection in open heart surgery [97]. Nowadays we have evidence for the benefit of controlling perioperative and postoperative glucose levels in both diabetic patients and hyperglycaemic non diabetic patient by using continuous insulin infusion (CII) according to the Portland protocol [98].

Chronic obstructive pulmonary disease (COPD) as independent risk factor of mediastinitis

The estimation of OR was 2.53, 95% CI (2.11–3.01) with heterogeneity. This result was robust when performing a sensitivity analysis.

As underlined by Wilson and Sethi [99,100] epidemiological cohorts of acute development of COPD are difficult because of the heterogeneous nature of the disease, the part of bacterial infection, and consequently the usage of antibiotics is complex. Some COPD patients are constantly colonized by bacteria between aggravations periods. This could explain the reason why those patients get mediastinitis.

Bilateral ima grafting (BIMA) as independent risk factor of mediastinitis

The estimation of OR was 2.49, 95% CI (2.05–3.03) from 23 studies with heterogeneity. This result was robust when performing a sensitivity analysis and was confirmed by the ART (Arterial Revascularization Trial) [77]. The incidence of Deep sternal wound complication was 3.5% for BIMA as compared to 1.9% in SIMA with OR = 1.9, 95% CI (1.20–3.0) p = .005. The pooled incidence from 22 studies was OR = 2.54 95% CI (2.07–3.01) and was included in the 95% confidence interval of the ART trial. Taggart DP et al [101] in an overview and meta-analysis underlined that BIMA had a better survival than SIMA (Single IMA grafting). . The 10 years ART trial was recently published [102] showing no efficacy on total mortality and composite end point in an intention to treat analysis. Because of methodological problems it could be considered neutral [103]. And further studies are needed to be conclusive on this topic.

Implications for research

The quality of reporting in the included studies of our meta-analysis was poor. Future studies with cohort or case-control design should adhere to the methodological standards that reduce possible bias, including prospective/retrospective cohorts, blinding of exposed/non exposed groups for outcome assessors, measures to reduce the drop-out rate, and an analysis based on all patients recruited. Also, an a priori power analysis is mandatory. Moreover, reports of studies should adhere to generally established standards of reporting [69].

Implications for practice

Prevention and better treatment of mediastinitis complications must be the major goal in assuming the highest quality of cardiovascular care. Prevention needs to handle the major risk factors such as reducing obesity before surgery via an adequate nutritional and physical exercise program, especially in lower latitude countries. Controlling glycaemia, using the Portland protocol, that has shown its efficacy in decreasing incidence of mediastinitis and other complications. It is mandatory to carefully monitor patients with COPD. The procedure of BIMA (bilateral IMA grafting) represents an increasing risk of mediastinitis especially in female patients. Nowadays we cannot confirm its beneficial effect on survival for the patients.

Disclosure

The authors report no conflicts of interest in this work.

Author contribution

All the authors made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data. All participated in drafting of the article or revising it critically for important intellectual content, and gave final approval of the version to be published.

Acknowledgements

We would like to thank our Medical Librarian, Marie Isachsen, for assisting with the literature search.