Defining the Relationship and Impact of Left Ventricular Ejection Fraction on the Incidence of Postoperative Adverse Events After Noncardiac Surgery: A Retrospective Cohort Study

Left ventricular ejection fraction (LVEF) is frequently used to quantify systolic function in patients prior to noncardiac surgery. While the perioperative risk of reduced LVEF has been established [1], it is unknown whether values within or above the normal range are associated with increased surgical risk. To refine our understanding of preoperative LVEF values, we examined the relationship between this measurement and adverse events after noncardiac surgery.

The study was approved by the Institutional Review Board. Adult patients who had an echocardiogram prior to noncardiac surgery were eligible for inclusion. Patients receiving anesthesia for dental, radiology, ophthalmology, or medical (e.g. endoscopy) procedures were excluded, as were those with an American Society of Anesthesiologists physical status classification of 6 or missing left ventricular internal diameter at end-diastole (LVIDD) or LVEF values.

The optimal preoperative LVEF was defined by comparing preoperative values with the incidence of acute kidney injury (AKI) within the 30 days (an increase in serum creatinine of at least 0.3 mg/dL or 150% from baseline). This outcome was selected for its high prevalence and relationship with other important outcomes, such as mortality [2]. The relationship between LVEF and LVIDD was modeled using multivariable logistic regression which included baseline characteristics including age, gender, and Elixhauser Comorbidity Index [3]. To detect nonlinear relationships, LVEF and LVIDD were fit using fractional polynomial equations. Discrimination of the models was measured by the c-statistic and DeLong’s 95% confidence interval (CI). Statistical analysis was conducted using R, version 3.5.1.

Between 1/2/2004 and 8/6/2015, 20,758 cases were included in the analysis. The study population had a mean age of 58 years (standard deviation of 14 years) and 54% were females. Preoperative echocardiography revealed a reduced LVEF (45–54%) in 1964 (9.5%) individuals. Of these 664 (3%) and 328 (2%) had moderately (30–44%) or severely (<30%) reduced values, respectively. Higher LVEF values (>65%) were observed in 2,612 (13%) of patients. General (24%), orthopedic (17%), and urologic (10%) were the most common surgical specialties. Kidney injury occurred in 3,727 (18%) patients.

Both high and low LVEF and LVIDD were associated with AKI, and these relationships were best modeled using J-shaped curves (Figure 1). The lowest predicted probability of AKI occurred when LVEF was 58% and LVIDD was 45mm. Overall, the discrimination of this model was fair (c-statistic=0.63, 95% CI 0.62–0.64). This finding is novel because it reveals increased perioperative risk with high LVEF values. Furthermore, it demonstrates that LVIDD, in addition to LVEF, is associated with postoperative adverse events.

Figure 1.

(A) Scatterplot depicting the proportion of individuals with the combined outcome for each value of LVEF. The size of each bubble indicates the number of individuals at each LVEF value, and the blue line indicates the fractional polynomial fit of the proportions by LVEF value. (B) Scatterplot depicting the proportion of individuals with the combined outcome for each value of LVIDD. The size of each bubble indicates the number of individuals at each LVIDD value, and the blue line indicates the fractional polynomial fit of the proportions by LVIDD value. LVEF=left ventricular ejection fraction, LVIDD=left ventricular end-diastolic internal diameter.

While multiple investigations have shown an association between reduced LVEF and postoperative adverse events [1, 4], the relationship between increased LVEF and postoperative adverse events is less well described. Hyperdynamic left ventricles occur in the presence of diastolic dysfunction [5], which represents a possible mechanism for the relationship we observed between increased LVEF values and postoperative adverse events. Conditions associated with decreased afterload such as sepsis or mitral regurgitation may also contribute.

The relationship between left ventricular dimensions and postoperative adverse events after noncardiac surgery is also novel. Both LVEF and LVIDD (chamber dimensions) affect the stroke volume. For patients with reduced LVEF, an increase in LVIDD would correspond to an increased stroke volume, which may help an individual cope with the physiologic stress of surgery. However, our finding of J-shaped relationship for both LVEF and LVIDD does not support this concept. Instead, altered chamber dimensions may more strongly reflect the chronicity and severity of changes in function, which would explain this finding.

Overall, LVEF and LVIDD have a J-shaped relationship with postoperative AKI. These findings argue against the practice of being reassured by LVEF values above a certain level.