The prognostic value of perioperative diastolic dysfunction (PDD) in patients undergoing noncardiac surgery remains uncertain, and the current guidelines do not recognize PDD as a perioperative risk factor. This systematic review aimed to investigate whether existing evidence supports PDD as an independent predictor of adverse events after noncardiac surgery.
Ovid MEDLINE, PubMed, EMBASE, the Cochrane Library, and Google search engine were searched for English-language citations in April 2015 investigating PDD as a risk factor for perioperative adverse events in adult patients undergoing noncardiac surgery. Two reviewers independently assessed the study risk of bias. Extracted data were verified. Random-effects model was used for meta-analysis, and reviewers’ certainty was graded.
Seventeen studies met eligibility criteria; however, 13 contributed to evidence synthesis. The entire body of evidence addressing the research question was based on a total of 3,876 patients. PDD was significantly associated with pulmonary edema/congestive heart failure (odds ratio [OR], 3.90; 95% CI, 2.23 to 6.83; 3 studies; 996 patients), myocardial infarction (OR, 1.74; 95% CI, 1.14 to 2.67; 3 studies; 717 patients), and the composite outcome of major adverse cardiovascular events (OR, 2.03; 95% CI, 1.24 to 3.32; 4 studies; 1,814 patients). Evidence addressing other outcomes had low statistical power, but higher long-term cardiovascular mortality was observed in patients undergoing open vascular repair (OR, 3.00; 95% CI, 1.50 to 6.00). Reviewers’ overall certainty of the evidence was moderate.
Evidence of moderate certainty indicates that PDD is an independent risk factor for adverse cardiovascular outcomes after noncardiac surgery.
The authors have performed a random-effects meta-analysis that shows supportive evidence for perioperative diastolic dysfunction as an independent risk factor for adverse cardiovascular events after noncardiac surgery. The work supports the importance of increased awareness of perioperative diastolic dysfunction when considering the cardiac risk factors for noncardiac surgery.
Cardiac morbidity and mortality remain a major source of adverse events in noncardiac surgery
Cardiac diastolic dysfunction is an increasingly recognized form of ventricular dysfunction with high prevalence and association with morbidity
The relationship between perioperative diastolic dysfunction and perioperative cardiac risk in noncardiac surgery is not well understood
The authors have performed a random-effects meta-analysis that shows supportive evidence for perioperative diastolic dysfunction as an independent risk factor for adverse cardiovascular events after noncardiac surgery
The work supports the importance of increased awareness of perioperative diastolic dysfunction when considering the cardiac risk factors for noncardiac surgery
WORLDWIDE about 200 million patients undergo noncardiac surgery annually.1,2 Of these, more than 1 million die within 30 days and 20 million experience major adverse events.2,3 Preoperative risk prediction aims to influence clinical decision and resource planning to avoid or reduce perioperative mortality and morbidity by identifying patients for whom benefits of surgery will outweigh procedure-related harms. Recommended by American College of Cardiology/American Heart Association (ACC/AHA) guidelines, the revised cardiac risk index is a widely used risk prediction tool to stratify candidates for noncardiac surgery.4–6 Limitations of the tool have been noted for specific surgical populations, such as patients undergoing lung resection or vascular surgery.7,8 Modifications of the tool have been proposed to account for changes in clinical practice and subsequent revisions in the definitions of adverse events.9,10 After accounting for predictors that are already included in a risk prediction model, factors that are independently associated with adverse perioperative outcomes are good candidates for model improvement studies.
One such candidate predictor could be diastolic dysfunction. Diastolic dysfunction is characterized by an abnormal relaxation of the ventricles, resulting in high ventricular filling pressure.11 Diastolic dysfunction usually precedes systolic dysfunction.12 The prevalence of diastolic dysfunction in the community is estimated to be 28% in the population 60 yr or older.13 Kuznetsova et al.14 have previously shown that low early diastolic mitral annulus velocity measured by tissue Doppler imaging was an independent predictor (above and beyond the traditional cardiovascular risk factors) of fatal and nonfatal cardiovascular events in the general population. Onset of the dysfunction is asymptomatic and preclinical, progressing over time to symptomatic diastolic heart failure. With aging population demographics, interest in preclinical diastolic dysfunction as a risk factor for cardiovascular outcomes is attracting more attention. While diastolic dysfunction may affect both ventricles, it is mostly the left ventricular diastolic dysfunction that is reported in association with adverse cardiovascular outcomes. In this review, left ventricular diastolic dysfunction will be referred to as perioperative diastolic dysfunction (PDD).
Echocardiography is the imaging modality and the accepted standard used to determine the presence of diastolic dysfunction in clinical practice.15 The severity of diastolic dysfunction is classified as mild or impaired relaxation (grade I), moderate or pseudonormal (grade II), and severe or restrictive (grade III) based on echocardiographic findings. The guidelines recommend a set of echocardiographic measurements15 : the mitral inflow parameters (E/A ratio, isovolumic relaxation time, and deceleration time), tissue Doppler of the mitral annulus (é), pulmonary venous flow (S, D, and A waves), transmitral propagation velocity (Vp), and E/é ratio. The E/é ratio is being widely adopted in clinical research.
Perioperative diastolic dysfunction has been found to be significantly associated with in-hospital mortality, major adverse cardiovascular events (MACE), difficult weaning from cardiopulmonary bypass, and the need for more frequent inotropic/vasoactive pharmacologic support in patients undergoing cardiac surgery.16–18 The relationship between PDD and adverse outcomes after noncardiac surgery, however, is less well understood. The recent ACC/AHA Guidelines on Perioperative Cardiovascular Evaluation for Non-Cardiac Surgery consider heart failure (diastolic or systolic) as a major risk factor without reference to PDD.5,6
We undertook a systematic review of the literature to investigate whether existing evidence supports PDD as an independent predictor of adverse health outcomes in patients undergoing noncardiac surgery. The research question articulated in our a priori protocol was the following: Is perioperative left ventricular diastolic dysfunction an independent predictor of adverse health outcomes within 30 days of noncardiac surgery?
Materials and Methods
We followed a prespecified systematic review protocol. Our review was prospectively registered (Centre for Reviews and Dissemination 42015020173) with the International Prospective Register of Systematic Reviews (PROSPERO).
A systematic search was conducted for studies published from 1946 to April 2015 by searching Ovid MEDLINE, PubMed, EMBASE, and the Cochrane Library. Additional search was conducted using the Google search engine. Keywords and medical subject headings related to diastolic dysfunction, perioperative/intraoperative period, and noncardiac surgery were used. The search results were limited to English language and human studies. The full search strategy is provided in appendix 1.
One reviewer (A.F. or M.A.) screened titles and abstracts for potential relevance, and a second reviewer (H.Y.) verified exclusions at this level. Two independent reviewers (A.F. and H.Y.) assessed the full publication of potentially relevant studies, and discrepancies were resolved by consensus.
We included analytic observational studies on adult patients undergoing noncardiac surgery, comparing echocardiographically established PDD with normal left ventricular diastolic function, lower grade PDD, or both. We accepted all investigator-defined PDD. Eligibility was also restricted to studies that were reported in the English language. We excluded studies that exclusively included patients with symptomatic diastolic dysfunction or congestive heart failure (CHF), patients with low ejection fraction, or patients undergoing cardiac procedures. We also excluded studies that used biomarker proxies of diastolic dysfunction without echocardiographic confirmation to minimize specificity concerns.
Data Extraction and Critical Appraisal
After piloting data extraction forms, a single reviewer (M.T.A.) extracted general study characteristics, including funding source, sample size, study design, eligibility criteria, description of population, exposure definition and measurement details, outcome definition, time point/follow-up duration, measurement tool or scale, cutoffs employed, level of care, type of surgery and anesthesia protocol, quantitative outcome data, statistical test used, and covariate adjustment. Another reviewer (A.F.) independently verified outcome data.
Two reviewers (M.T.A. and A.F.) judged study applicability and risk of bias. For each outcome of interest, we assessed study risk of bias using the Quality In Prognosis Studies tool that covers six domains, namely, study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting.19 Applicability was based on population description, exposure ascertainment (i.e., definition of PDD, echo parameter criteria, and imaging modality), setting, outcome definition, level of care, and anesthesia and surgery protocols. To assess clinical applicability of investigator-defined PDD, we examined whether the characterization of PDD was based on at least one of the following echocardiographic parameters either as individual measurements or in combination with other diastolic parameters.
Mitral E/A ratio (m/s) = early diastolic filling velocity (E-wave) divided by atrial contraction filling velocity (A-wave)
Transmitral flow propagation velocity (Vp)
E/é ratio: the ratio of early diastole E wave mitral inflow to annular velocity é
The overall study risk of bias was categorized as high, moderate, or low. Applicability assessment was rated as no concern and major or minor concerns. To assess confounding in included studies, we considered age, sex, weight, history of cardiovascular disease, diabetes mellitus, renal dysfunction, hypertension, type of surgery, and type of anesthesia as potential confounders.
We employed the directed acyclic graph approach to map the causal relationship of potential confounders with the exposure (i.e., PDD) and outcomes of interest.20,21 This approach is helpful in understanding the structure of biasing pathways. According to the structural theory of epidemiologic bias, a confounder is a “common cause” of both the exposure (i.e., perioperative left ventricular diastolic dysfunction in this case) and the outcome (e.g., cardiovascular death). Confounding bias is controlled in the design of a study (e.g., by matching on a confounder variable) or in its analyses (e.g., statistical adjustment). A “common effect” of exposure and outcome should not be adjusted for in the design or analysis because this will lead to selection bias (for details, readers are referred to the article by Hernán et al.20 ).
As such, we depicted the structure of causal relationships between a covariate, the exposure (left ventricular diastolic dysfunction), and the outcomes to identify potential confounders and assess whether studies adequately controlled confounding bias (appendix 2). The depicted relationships were informed by our understanding of the pathophysiology of diastolic dysfunction. We used online DAGitty software for this purpose.21 The DAGitty software computed the following “minimal sufficient adjustment” sets that studies should optimally control for in the study design or analysis when investigating an unconfounded association between PDD and adverse postsurgical outcomes:
Cardiovascular disease, diabetes mellitus, hypertension, type of anesthesia, and type of surgery, or
Cardiovascular disease, renal dysfunction, type of anesthesia, and type of surgery
Because age and history of chronic diseases are commonly employed proxies for the duration of disease exposure, we examined whether studies adequately controlled for age, history of aforementioned chronic diseases, type of surgery, and anesthesia protocol.
Specifically for studies comparing higher with lower grade PDD, the use of postoperative angiotensin receptor blocker or angiotensin-converting enzyme inhibitor was considered an additional important confounder. For studies that were not conducted exclusively in patients with normal ejection fraction, we also assessed for adequacy of control for this variable in either the design or the analysis of studies while assessing risk of bias.
Data Synthesis and Analysis
The prespecified 30-day outcomes were all-cause mortality, cardiovascular death, pulmonary edema or congestive cardiac failure, length of hospital stay, MACE, myocardial ischemia or infarction, and arrhythmia requiring treatment. Data were quantitatively pooled unless between-study heterogeneity (I2 > 50%) could be explained by study-level clinical or methodologic covariates. Data were pooled in Review Manager 5.3 using random-effects generic inverse variance or Mantel–Haenszel method (Review Manager [RevMan] [Computer Program]; Version 5.3; Copenhagen, Denmark: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). When both adjusted and crude estimates of association were reported, adjusted estimates were selected for meta-analyses. When applicable, we had planned to undertake sensitivity analyses by study risk of bias. We had also planned subgroup analyses for the various grades of PDD, baseline cardiac preoperative risk scores, or other important study-level clinical covariates identified post hoc. Because of the limited number of data-contributing studies, we could not statistically test for publication bias.
Assessment of Reviewers’ Certainty in Estimates of Association
Two reviewers (M.T.A. and A.F.) used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to rate their certainty (or confidence) in estimates of prognostic association using the four tier levels of high, moderate, low, or very low.22,23 For prognostic inference, sound cohort, case-cohort, and nested case-control studies provide the highest quality of evidence.22,24 Reviewers’ certainty in evidence is downgraded when there are important limitations in the validity and generalizability of studies, inconsistency between them, lack of statistical power in the data, or concerns about publication bias. Certain factors (e.g., dose–response relationship and large difference in absolute risk) increase our certainty in estimates. We did not formally grade the certainty of estimates of association that were bounded by wide CIs, precluding meaningful conclusions.
A total of 859 records were identified and screened for eligibility. Seventeen studies met the inclusion criteria for the review (fig. 1). Of those, four studies did not contribute to evidence synthesis either because data for outcomes of interest were not reported or because classification of exposure and controls was unclear.25–28 The remaining 13 studies contributing evidence were of diverse methodologic and clinical characteristics (table 1). Design of the included studies was observational prospective cohort (N = 6), retrospective chart review (N = 6), and case-control (N = 1). PDD was defined as per E/A ratio, E/é ratio, or both parameters with or without consideration of deceleration time. One study, however, defined PDD with transmitral flow propagation velocity (Vp).34 There was heterogeneity in ratio cutoffs for the E/A (0.75, 0.8, and 1.0) and E/é (8, 10, and 15) parameters across the studies with most studies comparing mixed or specific grades of PDD with normal diastolic function (N = 9).30,31,34,36–41 Four studies compared higher (i.e., moderate or severe) grade PDD with the composite of normal diastolic function and lower (i.e., various permutations of mild or moderate) grade PDD.29,32,33,35
Mean age of patients across the 13 studies ranged from 45 to 72 yr. Type of anesthesia administered to patients was not reported in six studies, but the rest reported using general anesthesia protocols.29,33,36–38,40 Of note, Flu et al.30 employed general anesthesia for all open vascular repairs and 35% of endovascular surgeries.
The entire body of evidence addressing the research question was based on a total of 3,876 patients. Because of frequent nonreporting of symptomatic/asymptomatic status of patients or unclear accounting for symptoms of heart failure in statistical analyses across the studies, we documented our corresponding generalizability concerns under assessment of external validity (table 1).
Risk of Bias in Included Studies
Except for the composite outcome of postoperative adverse events reported in the study by Matyal et al.,34 outcome data were judged to be at significant risk of bias across all studies (table 2). In general, the risk of bias concerns were frequently about selection of participants and residual confounding, although studies were also at risk of bias for various other reasons such as subjective, nonblinded classification of diastolic dysfunction or outcome ascertainment and unclear reporting of patient attrition and handling of missing data.
In total, seven studies reported this outcome assessed at variable short- and long-term durations after surgery.29,30,34,37,39–41 We meta-analyzed data from the three studies reporting 30-day or in-hospital mortality, which yielded a nonsignificant odds ratio (OR) with wide CI for PDD (mixed grades) versus normal diastolic function (OR, 1.10; 95% CI, 0.62 to 1.94). Adding the retrospective cohort study by Raevens et al,39 which reported short-term mortality after 3 months of liver transplant surgery, did not change the pooled estimate on early mortality after surgery (fig. 2).
As an exploratory meta-analysis to harness the cumulative power of the evidence base, we also pooled mortality data irrespective of observational study designs, surgical diversity, and outcome measurement time points (range: in-hospital stay to several years postsurgery). Irrespective of severity, PDD was not significantly associated with combined short- and long-term all-cause mortality, but the body of evidence was underpowered to detect a difference (fig. 3). Whether patients had vascular or hepatic surgeries did not yield any statistically significant subgroup differences. Restricting the meta-analysis to studies specifically comparing moderate-to-severe PDD with no or mild PDD also yielded imprecise pooled estimate, but association with higher mortality could not be ruled out (fig. 4).
The 30-day cardiovascular mortality was reported in the study by Flu et al30 on 708 patients undergoing elective open or endovascular repair. There were 11 cardiovascular deaths within 30 days (OR, 2.01; 95% CI, 0.61 to 6.67). For a mean follow-up of 2.2 yr, significantly higher (adjusted) odds of cardiovascular death with PDD were observed in the subgroup of patients undergoing open vascular repair (OR, 3.00; 95% CI, 1.50 to 6.00) as opposed to elective endovascular procedures.
One case-control and two prospective cohort studies reported this outcome during the period of hospitalization after surgery32,34,35 (table 2). A significant association (OR, 3.90; 95% CI, 2.23 to 6.83) between PDD and CHF/pulmonary edema was observed (fig. 5). The study by Cho et al.32 reported estimates of association separately for PDD defined by E/é and E/A ratios. Pooled estimate did not change in sensitivity meta-analyses guided by different definitions of PDD.
Studies were clinically diverse in their echocardiographic approaches, parameters defining PDD, surgery types, mean age of participants, and how pulmonary edema was defined—however, all studies included radiologic evidence in their definition. Diversity was also noted because studies either compared PDD (irrespective of grade) with normal diastolic function or compared moderate to severe PDD with combined mild PDD and normal diastolic function.
Length of Hospital Stay
Three studies reported this outcome in a total of 660 patients undergoing vascular, abdominal, or hepatic surgical procedures34,40,41 (tables 1 and 2). PDD definitions and the analysis of the length of hospital stay data were inconsistent, so meta-analysis was not possible. Matyal et al.34 found significantly longer hospitalization in patients with PDD irrespective of their ejection fraction (median of 7 vs. 5 days). Shounak et al.40 found that a greater proportion of patients with PDD undergoing abdominal surgery were hospitalized longer term, but the findings were not statistically significant. Xu et al.,41 on the other hand, found no association between length of stay and PDD in patients undergoing orthotopic liver transplantation.
Major Adverse Cardiovascular Events
Three studies reported 30-day MACE outcomes data.30,32,33 MACE was variably defined by investigators, but data were far too inadequate to undertake any meta-regression or subgroup analyses sensitive to MACE definitions. All MACE definitions included acute coronary events and mortality. Another study by Matyal et al.34 reported the composite of in-hospital postsurgical adverse events, which we considered a reasonable approximation of 30-day MACE. Observed statistical heterogeneity was not explained by study risk of bias, type of surgery, or definition of PDD. Furthermore, notable overlap of CIs was observed. From a clinical decision-making perspective, the consistency in the direction of estimates of association despite the observed clinical and methodologic diversity across the studies compelled a formal meta-analysis. Pooled estimate of association (OR, 2.03; 95% CI, 1.24 to 3.32) demonstrated a significant risk of MACE with PDD (fig. 6 and table 3). Risk estimate remained unchanged when we excluded the study by Matyal et al34 from the meta-analysis.
Myocardial Ischemia or Infarction
Three studies on patients undergoing various vascular procedures were included in this analysis. PDD-defining parameters and cutoffs were heterogeneous. Follow-up duration ranged from 48 h postsurgery to the entire period of hospitalization. Pooled estimate revealed higher odds (OR, 1.74; 95% CI, 1.14 to 2.67) of myocardial ischemia or infarction with PDD in the short-term period after surgery (fig. 7).
Arrhythmia Requiring Treatment
Crude estimate obtained from a single study on patients undergoing elective aortic or peripheral vascular surgery under general anesthesia failed to reveal any significant association between PDD and arrhythmia requiring treatment for the period of hospitalization (OR, 1.89; 95% CI, 0.66 to 5.42).34 Given wide CI around the point estimate, findings were inconclusive.
Two other studies on patients undergoing lung surgery reported data for atrial fibrillation requiring treatment in the immediate postsurgical period.36,38 Findings were conflicting. Both studies were at high risk of bias for various reasons (table 2). Anile et al.36 found nonsignificant association with PDD, while Nojiri et al.38 demonstrated a relative risk of 1.81 (95% CI, 1.36 to 2.42). To be noted, the latter included an unknown proportion of patients with low ejection fraction and did not adjust for it in their analysis.
Grading Reviewers’ Certainty for Estimates of Association
For outcomes with statistically significant findings, our certainty varied from very low to moderate (table 3). Because evidence was underpowered yielding very wide CIs, an association between PDD and 30-day and longer-term all-cause mortality, 30-day cardiovascular death, length of hospital stay, and arrhythmia requiring treatment could neither be confirmed nor be refuted.
This is the first critical review of the association between PDD and perioperative outcomes in patients undergoing noncardiac surgery. Our systematic review and meta-analysis included more than 3,800 patients undergoing a variety of different noncardiac surgeries. We found scant evidence addressing several key outcomes for the immediate postsurgical period. However, PDD may be an independent predictor of MACE as a composite outcome as well as CHF and myocardial infarction as independent outcomes in the immediate period after surgery. Evidence also demonstrated an association between PDD and cardiovascular death in patients undergoing major surgical procedures, particularly open vascular surgeries. Furthermore, evidence in this analysis does not rule out the possibility that moderate- to severe-grade PDD may also be associated with higher all-cause mortality in patients undergoing noncardiac surgery (hazard ratio, 2.59; 95% CI, 0.85 to 7.93).
The prevalence of PDD in patients undergoing noncardiac surgery is unknown. The reported prevalence of the diastolic dysfunction in the general population, however, varies from 11.1 to 34.7%.35,42,43 As such, with the large population of 200 million patients undergoing noncardiac surgeries annually, at least 20 to 70 million individuals may be at higher risk of MACE due to PDD. Higher prevalence of diastolic dysfunction is observed, particularly in the elderly, patients with coronary artery disease, hypertension, diabetes mellitus, cardiomyopathies, valvular disease, and a variety of other systemic diseases.11,44–52 Preoperative identification and management of this large at-risk surgical group has the potential to improve perioperative outcomes of surgery and utilization of scarce resources.
Several limitations inherent in the body of evidence yielded very low to low certainty for most of the aforementioned outcome-specific estimates of risk as judged using the GRADE approach.22,23 With moderate certainty, however, we can state that 47 more per 1,000 patients (95% CI, from 11 more to 99 more) will experience a MACE within 30 days of noncardiac surgery if they have PDD compared with those without diastolic dysfunction. Notwithstanding, when outcome-specific evidence is viewed in toto, a biologically plausible and coherent account of evidence can be immediately appreciated. If findings were spurious, the direction of association across the cardiovascular outcomes would have been randomly inconsistent and pathophysiologically incoherent. Such is clearly not the case. The higher incidence of MACE, heart failure, myocardial infarction, and cardiovascular death in noncardiac surgical patients with perioperative PDD provides evidence in keeping with the known pathophysiology of cardiovascular disease. While the observed individual estimates of association may be less certain, when viewed collectively, evidence of moderate certainty supports perioperative PDD as a risk predictor of adverse cardiovascular outcomes after noncardiac surgery.
The pathophysiology of diastolic dysfunction provides a biological rationale for an association with myocardial ischemia/infarction, pulmonary edema, and MACE. Left ventricle (LV) end-diastolic pressure is an important factor affecting the oxygen supply to the myocardium. The amount of blood flow entering the coronary circulation during diastole is the result of the pressure gradient between the epicardial coronary artery and the subendocardial segment. Elevation of the LV end-diastolic pressure, as in diastolic dysfunction patients, can reduce this gradient significantly, decreasing the coronary diastolic blood flow and subsequently decreasing myocardial perfusion.42,43 The oxygen cost of “pressure work” is greater than “volume work,” with the area-under-the-curve for LV pressure closely correlating with myocardial oxygen demand.44 During the perioperative period, stress response is well reported, further exacerbating the balance of myocardial oxygen supply and demand, including patients with nonobstructive coronary artery stenoses.45–47 As the disease advances, high ventricular filling pressure leads to high left atrial pressure and pulmonary venous hypertension that results in greater susceptibility to the development of flash pulmonary edema.48,49 In the perioperative period, excessive fluid replacement and hemodynamic instability may trigger pulmonary edema at a lower threshold in patients with PDD.50,51 Furthermore, a catecholamine surge in patients with diastolic dysfunction could potentially alter ventricular–atrial coupling, thereby increasing the risk of pulmonary edema/CHF and hemodynamic instability.52,53 Additionally, occurrence of myocardial ischemia could be further contributing to the aggravation of pulmonary edema or CHF, with the two adverse events accounting for the higher incidence of MACE with PDD that we have observed.
Our claim of the incremental value of PDD as a predictor of adverse surgical outcomes over and above other risk factors such as hypertension, diabetes mellitus, and frailty that are already accounted for in existing cardiac risk prediction models might need to be taken with caution because of the limitations identified in the studies. However, we did appraise whether important confounding was adjusted in the design or analysis of individual studies. While most studies were at high risk of bias for inadequate control of confounding, cofounders were not identical across studies. Therefore, it is more likely that PDD is an independent risk factor in its own right rather than a surrogate for hypertension, diabetes, or frailty. A possible explanation for PDD as an independent risk predictor could be that the composite of age, history of hypertension, and diabetes may not adequately capture the real intensity and duration of exposure to the preexistent cardiovascular stresses that predispose to a higher risk of surgery.
We acknowledge a few limitations in the conduct of our systematic review. We accepted investigators’ classification of PDD as long as at least one of the routine parameters was employed for classification of exposure. Guidelines recommend that diastolic dysfunction should be measured by at least two echo parameters to ensure reproducibility.15 Only 2 of the 13 included studies defined PDD with a single echo parameter, yet neither were found to be outliers in the meta-analyses they were included in, allaying important concerns about systematic error in risk estimation.31,34 We also accepted all investigator-defined parameter cutoff values unless deemed far from cutoff values employed in routine practice. Another limitation is that we did not consider the subjective assessment of PDD as a major limitation of study validity. However, blinded assessment of PDD and outcomes were important considerations in our critical appraisal of studies.
Despite the low level of certainty (as judged using the GRADE approach23 ) for the observed estimates of association, with moderate certainty, we can state that more patients will experience postoperative morbidity and mortality within 30 days of noncardiac surgery if they have PDD compared with those without diastolic dysfunction.
Conclusions and Future Research Recommendations
With moderate degree of certainty, we conclude that PDD is an independent risk factor for adverse cardiovascular outcomes after noncardiac surgery. We propose that future revisions of the ACC/AHA revised cardiac risk index or other risk prediction models in current use should consider PDD as an additional candidate risk predictor in model derivation studies. Such a study should ensure that assessment of PDD is rigorous and blinded with measures incorporated to minimize subjective interpretation of echo parameters. Even better would be to use the established echo parameters as continuous variables in the model as opposed to PDD grade categories. Subsequently, the comparative clinical effectiveness and cost-effectiveness of the earlier and revised versions of the model may be investigated if echo parameters make into the final revised model.
The authors thank Melody Afagh, B.Sc., Schulich School of Medicine and Dentistry, University of Western Ontario, London, Ontario, Canada, for scanning and reviewing citations for eligibility. The authors also thank Alexandra (Sascha) Davis, B.A., M.L.I.S., Library and Learning Centre, Civic Campus, The Ottawa Hospital, Ottawa, Ontario, Canada, for developing the search strategy for this review.
Methodologic consultation, data extraction, and analyses were funded by the first author.
The authors declare no competing interests.
Arrhythmia_Rx = arrhythmia requiring treatment; CHF = congestive heart failure; CVD = cardiovascular disease; LOHS = length of hospital stay; MACE = major adverse cardiovascular events; MI = myocardial infarction/ischemia; PDD = perioperative (left ventricular) diastolic dysfunction; postop = postoperative.
DAGitty minimal sufficient adjustment sets for adequate control of confounding bias:
(Age or history of) cardiovascular disease, diabetes mellitus, hypertension, type of anesthesia, and type of surgery, or
(Age or history of) cardiovascular disease, renal dysfunction, type of anesthesia, and type of surgery