Severity and Timing of Elevated Lactate Correlates with Mortality after Cardiac Surgery

Abstract

Objective:

Extreme serum hyperlactatemia is common after cardiac surgery but can portend organ ischemia and demise. Thus, we sought to determine what lactate thresholds should raise concern at various timepoints after cardiac surgery. We hypothesized that post-cardiac surgery lactate correlates with operative mortality and is most worrisome outside of the immediate post-bypass period.

Methods:

All adult patients at a quaternary institution undergoing CABG and/or valve procedures with recorded Society of Thoracic Surgeons Predicted Risk of Mortality (PROM) scores, peak postoperative lactate, and time-to-postoperative peak lactate from 2012-2024 were included for analysis. The correlation between lactate and operative mortality was modeled using log-PROM adjusted logistic regression.

Results:

In total, 7,292 patients (PROM score: 2.4%) met inclusion criteria (median lactate: 2.3 mmol/L). The operative mortality rate was 1.4% (n=103). On log-PROM-adjusted analysis, a postoperative peak lactate > 5 mmol/L (p<0.001) was associated with increased mortality risk (mortality rate: 9.8% [66/672]). Moreover, mortality risk dramatically increased with a postoperative peak lactate ≥ 7 mmol/L (mortality rate: 17.5% [52/297]). Only 6% (n=444) of patients experienced peak hyperlactatemia >24 hours post-operatively. On adjusted analysis, later lactate peaks demonstrated greater model discrimination than peak lactates collected in the first post-operative 24 hours (>24 Hours AUC: 0.94; ≤24 Hours AUC: 0.79).

Conclusion:

Peak lactate values >5 mmol/L correlate with increased mortality risk. Mortality risk increases drastically at lactates above 7 mmol/L. Early hyperlactatemia are less concerning, while after the first 24 hours postoperatively, there is a strong correlation between peak lactate and mortality.

Central Picture

Late hyperlactatemia nmore accurately predicts mortality than early hyperlactatemia.

Central Message:

Hyperlactatemia following cardiac surgery demonstrates improved correlation with mortality after the immediate post-bypass period.

Perspective Statement:

Hyperlactatemia is associated with mortality after cardiac surgery. However, Elevated lactate levels 24 hours after surgery warrant greater concern.

Introduction:

Serum lactate is a heavily utilized resuscitation marker in postoperative care after cardiac surgery. As a byproduct of anaerobic metabolism, elevated lactate may suggest underlying malperfusion and increased mortality risk. However, post cardiac surgery lactate is more nuanced due to the existence of multiple contributing factors such as hypothermia, anemia, or low cardiac output. Prolonged cardiopulmonary bypass can also trigger transient elevations in lactate. Perioperative hyperglycemia and vasopressor usage also increase lactate levels through stimulation of aerobic glycolysis. Hence, while hyperlactatemia is principally recognized as an indicator of tissue ischemia and malperfusion, other factors during cardiac surgery can obscure its specificity ¹. Thus, further contextualization of the association of lactate with mortality is warranted.

The utility of lactate in the treatment of other disease etiologies have been established ^2,3. However, clinically utilized thresholds are often conservative to optimize sensitivity and do not accurately reflect the association between elevated lactate and true mortality risk. Furthermore, due to multiple contributors to hyperlactatemia after bypass, the behavior of lactate after cardiac surgery is unique and requires focused investigation.

Thus, this study aims to characterize the association of peak post-operative lactate with mortality after cardiac surgery. Furthermore, we hypothesize that peak lactates observed beyond the first 24 hours after surgery may be more concerning.

Methods

This study was deemed exempt by the University of Michigan Institutional Review Board (HUM00225159, April 2024). Patient characteristics and outcomes were collected through data extraction from the University of Michigan Society of Thoracic Surgeons Adult Cardiac Surgery Database (STS ACSD).

Population

All adult patients (≥18 years old) with a recorded STS Predicted Risk of Mortality Score (PROM) who underwent isolated CABG, isolated valve surgery, and/or combined CABG + valve procedures from 2012-2024 were included in analysis. Patients without PROM including those who underwent aortic procedures, transplantation, mechanical ventricular assist device placement, concomitant aortic and mitral procedures, and/or extracorporeal mechanical support were excluded. The primary outcome was operative mortality, defined as death within 30 days of surgery or death during the index hospital stay. All patients included in the study received arterial blood gases upon admission to the intensive care unit (ICU). The median time till first ICU admission lactate was 4.7 (2.9, 6.2) hours after skin closure. Subsequent redraws of arterial blood gas were done at the managing team’s discretion. Peak lactate was defined as the highest postoperative lactate drawn. Time-till-peak post-operative lactate (hours) was defined as the difference in time between skin closure and the time of peak lactate collection. ICU admission lactate levels and post-operative peak lactate levels were reported. Categorical variables are represented as count (%) and compared by χ² analysis. Continuous variables are presented as a median (IQR) and compared by Kruskal -Wallis rank test. PROM score was reported as mean±standard deviation.

Statistical Analysis

The relation between peak postoperative lactate and operative mortality was modeled using logistic regression. Risk adjustment was performed using logit-transformed PROM (log-PROM). PROM provides a single estimate based on many pre-procedural variables. While PROM score does not provide exhaustive adjustment, it provides a comprehensive measurement of risk based on nationally sourced data ⁴. By comparison, risk-adjustment using individual factors could inappropriately adjust out the mechanisms through which lactate is raised. PROM scores were log-transformed to maintain their linear relationship with operative mortality within a logistic regression model.

Lactate was first modelled categorically. Lactate categories were formed using Fibonacci thresholds to provide progressively larger bins for less frequent, higher magnitudes of lactate.

Next, lactate was modelled as a continuous variable. Mortality risk estimates were derived using marginal effects analysis of the utilized logistic regression function. Thus, the association between operative mortality risk and peak lactate was modelled nonlinearly following a standard logistic distribution. The lactate value at which the rate of mortality risk surpassed a 1% increase in mortality risk per unit change in lactate was identified as an inflection point. The application of marginal effects analysis in healthcare science has been previously described ^5,6. Receiver operator characteristic (ROC) curves were constructed to evaluate model performance. The difference in the area under curve (ΔAUC) was calculated to identify differences. This was achieved with 1,000 bootstrapped samples with replacement from the original dataset. The association of peak lactate with post-operative complications was also described and modeled using univariate linear regression.

The association of operative mortality risk and post-operative peak lactate was further stratified by measurement time. ROC curves were used to see how peak lactate alone discriminated mortalities from survivors at different time-points. Based on these curves, peak lactates collected ≥ 24 hours post-surgery were modeled separately as a sub-analysis using log-PROM adjusted logistic regression as reported above. Lastly, frequency of post-operative lactate measurements was reported to identify surveillance bias. All variable definitions conform to the standard STS definitions. All analyses were performed in STATA MP/v19 with a significance level of p=0.05.

Results

Patient Characteristics

A total of 7,292 patients met inclusion criteria. Operations included 2,232 (30.6%) isolated CABG, 1,612 (22.1%) isolated SAVR, 2,582 (35.4%) isolated mitral repair/replacements, 545 (7.5%) SAVR + CABG, and 321 (4.4%) mitral surgery + CABG. The mean PROM was 2.4% ± 3.8% (Table 1).

Table 1: Pre-Operative Characteristics:

	Overall:	Isolated CABG:	Isolated SAVR:	Isolated Mitral Surgery	SAVR + CABG:	Mitral Surgery + CABG:	Significance (p<0.05):
Sample Size:	7,292	2,232	1,612	2,582	545	321
PROM (Mean +/−Standard Deviation) (%):	2.4 ± 3.8	1.9 ± 3.4	2.1 ± 2.6	2.4 ± 4.3	3.7 ± 3.4	4.9 ± 5.2	<0.001*
Age (years)	66 (58, 73)	66 (60, 72)	66 (58, 73)	64 (55, 72)	72 (66, 77)	69 (62, 75)	<0.001*
Female Patients (%)	2,472 (34%)	454 (20%)	584 (36%)	1,215 (47%)	126 (23%)	93 (29%)	<0.001*
Diabetes:	4,884 (67%)	1,186 (53%)	461 (29%)	427 (17%)	316 (58%)	216 (67%)	<0.001*
Dyslipidemia:	3,437 (47%)	1,222 (55%)	742 (46%)	1,013 (39%)	291 (53%)	169 (53%)	<0.001*
Dialysis:	192 (3%)	72 (3%)	22 (1%)	69 (3%)	14 (3%)	15 (5%)	0.001*
Hypertension:	5,730 (79%)	1,980 (89%)	1,277 (79%)	1,722 (67%)	485 (89%)	266 (83%)	<0.001*
History of Infective Endocarditis	408 (6%)	1 (<1%)	157 (10%)	222 (9%)	12 (2%)	16 (5%)	<0.001*
Moderate or Greater Chronic Lung Disease:	506 (7%)	154 (7%)	97 (6%)	181 (7%)	43 (8%)	31 (10%)	0.16
Liver Disease:	310 (4%)	82 (4%)	79 (5%)	116 (4%)	20 (4%)	13 (4%)	0.37
Prior Cardiovascular Accident	658 (9%)	218 (10%)	124 (8%)	242 (9%)	46 (8%)	28 (9%)	0.23
Previous Myocardial Infarction:	2,624 (36%)	1,388 (62%)	340 (21%)	554 (21%)	187 (34%)	155 (48%)	<0.001*
NYHA Class 3 or 4	1,379 (19%)	291 (13%)	291 (18%)	604 (23%)	110 (20%)	83 (26%)	<0.001*
Left Ventricular Ejection Fraction (%):	60 (53, 65)	58 (48, 63)	60 (55, 65)	60 (55, 65)	60 (55, 65)	55 (40, 60)	<0.001*

Table 1. Pre-operative characteristics for over-all population and by procedure type. Values reported as median (IQR: 25^th-75^th percentile) or n (%); p-values based on Chi-square test for categorical variables and Kruskal-Wallis test for continuous variables. PROM (Society of Thoracic Surgery- Predicted Risk of Mortality), NYHA (New York Heart Association)

Median patient age was 66 (58, 73) years and 2,472 (34%) patients were female (Table 1). Median cross clamp and cardiopulmonary bypass times were 88 (66, 114) and 110 (84, 144) minutes, respectively (Table 2). Postoperatively, 103 patients experienced operative mortality (1.4%) (Table 3).

Table 2: Intra-Operative Characteristics:

	Overall:	Isolated CABG:	Isolated SAVR:	Isolated Mitral Surgery	SAVR + CABG	Mitral Surgery + CABG	Significance (p<0.05)
Concurrent Tricuspid Repair	673 (9%)	1 (<1%)	0 (0%)	621 (24%)	0 (0%)	51 (16%)	<0.001*
Concurrent Atrial Fibrillation Procedure	942 (13%)	119 (5%)	58 (4%)	673 (26%)	32 (6%)	60 (19%)	<0.001*
Cardiopulmon ary Bypass Time (Minutes)	110 (84, 144)	109 (88, 133)	115 (93, 146)	92 (70, 130)	156 (133, 190)	153 (109, 197)	<0.001*
Cross Clamp Time (minutes)	88 (66, 114)	89 (71, 108)	91 (72, 114)	72 (51, 101)	130 (108, 157)	122 (81, 157)	<0.001*
Re-Operative Surgery	1,187 (16%)	83 (4%)	497 (31%)	530 (21%)	62 (11%)	15 (5%)	<0.001*
Non-Elective Status	1,594 (22%)	932 (42%)	209 (13%)	260 (10%)	113 (21%)	80 (25%)	<0.001*
Sternotomy Approach	6,788 (93%)	2,225 (100%)	1,588 (99%)	2,111 (82%)	545 (100%)	319 (99%)	<0.001*
Blood Transfusion	1,501 (21%)	435 (19%)	323 (20%)	423 (16%)	209 (38%)	111 (35%)	<0.001*
Number of Blood Transfusion for Blood Transfusion Recipients:	1 (0, 2)	1 (1, 2)	1 (0, 2)	1 (0, 2)	2 (1, 3)	2 (1, 3)	<0.001*
ICU Admission Lactate (mmol/L)	1 (0.8, 1.5)	1.1 (0.8, 1.6)	1.1 (0.8, 1.6)	0.9 (0.7, 1.3)	1.2 (0.8, 1.8)	1 (0.8, 1.6)	<0.001*

Table 2. Intra-operative characteristics for over-all population and by procedure type.

Table 3: Post-Operative Outcomes:

	Overall:	Isolated CABG:	Isolated SAVR:	Isolated Mitral Surgery	SAVR + CABG:	Mitral Surgery + CABG:	Significance (p<0.05):
Operative Mortality:	103 (1.4%)	26 (1.2%)	12 (0.7%)	45 (1.7%)	9 (1.6%)	11 (3.4%)	0.001*
Stroke:	111 (1.5%)	20 (0.8%)	26 (1.6%)	48 (1.9%)	9 (1.7%)	8 (2.5%)	0.04*
Atrial Fibrillation:	2,050 (28.1%)	588 (26.3%)	474 (29.4%)	679 (26.3%)	196 (35.9%)	113 (35.2%)	<0.001*
Renal Failure:	162 (2.2%)	49 (2.2%)	21 (1.3%)	61 (2.4%)	15 (2.8%)	16 (5.0%)	0.001*
Prolonged Ventilation (> 24 hours):	477 (6.5%)	124 (5.6%)	87 (5.4%)	184 (7.1%)	42 (7.7%)	40 (12.5%)	<0.001*
Deep Sternal Wound Infection:	26 (0.4%)	22 (1.0%)	1 (0.1%)	1 (<0.1%)	1 (0.2%)	1 (0.3%)	<0.001*
Sepsis:	58 (0.8%)	24 (1.1%)	9 (0.6%)	15 (0.6%)	6 (1.1%)	4 (1.2%)	0.2
Cardiac Arrest:	103 (1.4%)	19 (0.9%)	22 (1.4%)	38 (1.5%)	12 (2.2%)	12 (3.7%)	<0.001*
Frequency of Lactate Measurement *	10 (7, 14)	10 (7, 14)	9 (6, 13)	9 (7, 14)	12 (7, 17)	13 (7, 21)	<0.001*
Peak Post-Operative Lactate (mmol/L)	2.3 (1.7, 3.3)	2.2 (1.7, 3.1)	2.3 (1.7, 3.1)	2.3 (1.7, 3.4)	2.5 (1.9, 3.7)	2.9 (2, 4.5)	<0.001*

Table 3. Post-operative characteristics for over-all population and by procedure type.

Lactate Measurements

The median first ICU admission lactate level was 1 (0.8, 1.5) mmol/L and did not differ between by survivorship (p=0.66). Overall, the median peak post-operative lactate was 2.3 (1.7, 3.3) mmol/L for all patients but was higher in patients who died than that of survivors (Mortality: 7.5 [3.8, 14.2] mmol/L vs. Survivor: 2.3 [1.7, 3.3] mmol/L, p<0.001). When comparing peak lactate by type of surgery performed, values demonstrated statistical, but not clinical significance with the greatest lactate values in CABG + mitral valve operations and the lowest in CABG alone (2.9 [2, 4.5] vs 2.2 [1.7, 3.1], p<0.001) (Table 3). On univariate logistic regression analysis, peak lactate was significantly associated with operative mortality risk (OR: 1.55 ±0.04, 95% CI [1.47-1.63], p<0.001). Lactate values collected at ICU admission also demonstrated an association with mortality, but with smaller effect size (OR: 1.19 ± 0.09, 95% CI [1.03-1.38], p=0.02).

Lactate Measurement Frequency:

Overall, the median number of post-operative lactate measurements was 10 (7, 14) measurements. Operative mortalities received a greater frequency of post-operative lactate measurements compared to survivors (Mortality: 45 [16, 84] times vs. Survivor: 10 [7/14] times, p<0.001).

Next, 0.8% (55/ 7,292) patients only had 1 post-operative lactate measurement upon ICU admission, which by default was their peak lactate. Their median lactate was low at 1.5 (1.1, 2) mmol/L.

Additionally, frequency of post-operative lactate measurements varied by timing of lactate peak. Patients with lactate peaks after after the first 24 hours had almost double the frequency of post-operative lactate measurements compared to those who achieved their lactate peaks in the first post-operative 24 hours (After 24 Hours: 21[13, 41] times vs. Before 24 hours: 10 [6, 13] times, p<0.001).

Peak Lactate Associated with Operative Mortality Risk

On categorical analysis of peak lactate, values above 5 mmol/L demonstrated a significant association with operative mortality after adjusting for log-PROM. When peak lactate values below 2 mmol/L were used as a reference, peak lactates between 3.1-5 mmol/L demonstrated marginal association with mortality risk (OR: 2.09 ± 0.79, 95% CI [0.99-4.37], p=0.06). Next, peak lactates between 5.1-8 mmol/L demonstrated a greater than 5-fold increase in odds of mortality (OR: 5.78 ± 2.21, 95% CI [2.73-12.24], p<0.001). Peak lactate between 8-13 mmol/L demonstrated a greater than 20-fold increase in odds of operative mortality (OR: 20.13 ± 7.65, 95% CI [9.56-42.41], p<0.001). Lastly, peak lactate values above 13 mmol/L had an almost 250-fold increase in mortality risk (OR: 247.66 ± 112.31, 95% CI [101.82-602.35], p<0.001). Values above 5 mmol/L represented the highest 9.2% of all peak post-operative lactate values collected.

On log-PROM-adjusted analysis of peak lactate as a continuous variable, there was a 48% increase in the odds of operative mortality per unit increase in peak lactate (OR: 1.48 ± 0.04, 95% CI [1.40-1.56], p<0.001) (Table 4) (Figure 1). A peak lactate of 5 mmol/L was associated with a 1.4% ± 0.2% operative mortality risk, a lactate of 8 mmol/L was associated with a 4.4% ± 0.5% operative mortality risk, and a lactate of 13 mmol/L reached a 22.2% ± 3.6% operative mortality risk based on marginal means analysis. On analysis of the derivative function, mortality rate increased rapidly when peak lactate exceeded 7 mmol/L. Based on ROC curve analysis, the addition of peak lactate to log-PROM demonstrated superior performance (AUC: 0.89 ± 0.02) compared to log-PROM score alone (AUC: 0.81 ± 0.02) (ΔAUC: 0.07 ± 0.02, 95% CI [0.04-0.10], p<0.001) (Figure 3). Of note, first ICU admission lactate was not associated with operative mortality (p=0.67).

Table 4: Risk-Adjusted Models of Peak Lactate’s Association With Operative Mortality:

Peak Lactate Category	Median Peak Lactate	Sample Size (n)	PROM (%)	Operative Mortality Rate	Odds Ratio	Standard Error	95% CI	Significance:
Lactate (continuous variable): (mmol/L)
All Values	2.3 (1.7, 3.3)	7,292	2.4 ± 3.8	103 (1.4%)	1.48	0.04	1.40-1.56	<0.001*
Values Collected After 24 Hours	3 (2, 4.7)	444	4.6 ± 6.3	57 (12.8%)	1.48	0.07	1.35-1.62	<0.001*
Lactate (categorical stratification) *<2 (mmol/L) used as reference.
<2	1.6 (1.3, 1.8)	2,949	2.1 ± 2.9	13 (0.4 %)	-	-	-	-
2.1 -3	2.4 (2.2, 2.7)	2,188	2.2 ± 3.1	8 (0.4%)	0.80	0.36	0.33-1.93	0.62
3.1-5	3.7 (3.3, 4.2)	1,483	2.6 ± 4.2	16 (1.1%)	2.09	0.79	0.99-4.37	0.06
5.1-8	6.1 (5.5, 6.9)	476	3.2 ± 5.2	16 (3.4%)	5.78	2.21	2.73-12.24	<0.001*
8.1-13	9.7 (8.8, 10.9)	152	4.8 ± 7.6	20 (13.2%)	20.13	7.65	9.56-42.41	<0.001*
>13	16 (14.5, 20)	44	9.9 ± 12.1	30 (68.2%)	247.66	112.31	101.82-602.36	<0.001*

Table 4: Risk adjusted, analysis of peak post-operative lactate versus operative mortality risk. Peak lactate was modelled as a continuous variable for all lactate values and then separately for lactates collected after 24 hours. Peak lactate was then modelled into categories. Lactate values above 5 mmol/L were significantly associated with mortality risk.

Figure 1:

A. Mortality risk estimates and instantaneous change in mortality risk by Log-PROM adjusted, peak lactate with associated standard error standard error. B. Log-PROM adjusted, peak lactate versus mortality risk, graphical analysis C. Log-PROM adjusted, peak lactate versus change in mortality risk.

Figure 3:

Log-PROM adjusted peak lactate demonstrated significantly better model discrimination when compared to log-PROM alone.

Further sensitivity analysis was conducted to refine model performance for when peak lactate was analyzed by a continuous variable. First, survivors and operative mortalities were compared by perioperative risk factors. (Supplementary Table 2) Variables demonstrating significance that were not included in PROM calculation were identified. These variables included bypass time, cross-clamp time, and number of intra-operative blood transfusions. However, both cardiopulmonary bypass time (p=0.11) and cross-clamp time (p=0.21) lost significant association with mortality risk when modelled with postoperative peak lactate. Next, number of intra-operative blood transfusions lost significant association with mortality risk (p=0.48) when modelled with peak lactate and log-PROM. Furthermore, none of these variables demonstrated significant interaction with peak lactate in these models (Supplementary Table 3)

When peak lactate, time till peak lactate, and frequency of lactate measurement were modelled using log-PROM adjusted analysis, time till peak lactate (OR: 1.01, p=0.05) and frequency of late measurement (OR: 1.02, p=0.017) demonstrated marginal association with mortality risk. Additionally, frequency of lactate measurement and time till peak lactate demonstrated significant interaction (p=0.014) suggesting that later lactate peaks were associated with increased surveillance. This model did not demonstrate significantly greater discrimination compared to log-PROM adjusted analysis with peak lactate alone (ΔAUC: −0.01 ± 0.01, 95% CI[-0.03-0.00], p=0.09). (Supplementary Table 3). Thus, the final model for mortality risk was based on log-PROM adjusted peak lactate alone.

Lactate Peak ≥ 24-Hours After Surgery Greater Association with Mortality:

ROC curve analysis of peak lactate at various post-operative time points demonstrated a poor association between operative mortality risk and early peak-lactates (Supplementary Table 4). Peak lactates measured within the first 24 hours post-operatively were associated with poor discriminatory performance (AUC ≤ 0.70). Thus, further sub-analysis of the association between mortality risk and peak lactate was performed for those with lactate peaks before the first 24 post-operative hours versus after. Of note, only 2.9% (3/103) of all operative mortalities occurred within the first 24-hours. Thus, the effect of early patient loss was minimal on time stratified analysis.

In total, 6.1% (n=444) patients experienced their lactate peak ≥24 hours post-surgery. Operative mortality was significantly higher in patients with delayed lactate peaks (≥24 Hours: 12.8% [57/444] vs. <24 Hours: 0.6% [46/6,848], p<0.001). Lactate peak >24 hours demonstrated a significant association with operative mortality (≥ 24 Hours: OR: 1.48±0.07, 95% CI [1.35-1.62], p<0.001) after risk adjustment by log-PROM. (Figure 2) This was a stronger association with mortality risk compared to peak lactate values collected within the first 24 hours during log-PROM adjusted analysis (OR: 1.33 ± 0.06, 95% CI [1.22-1.45], p<0.001). Additionally, log-PROM adjusted peak lactate showed strong discriminatory ability when measured after the first 24 post-operative hours (AUC: 0.94 ± 0.02). (Supplementary Table 3)

Figure 2:

Analysis of log-PROM adjusted peak lactate collected more than 24 hours after surgery. A. Mortality risk estimates and instantaneous change in mortality risk by log-PROM adjusted, peak lactate values with associated standard error. B. Peak lactate versus mortality risk, graphical analysis C. Peak lactate versus change in mortality risk, graphical analysis.

On log-PROM adjusted analysis of peak lactates collected ≥ 24 hours post-surgery, a lactate of 5 mmol/L was associated with a 7.4% ± 1.5% mortality risk, a peak lactate of 8 mmol/L was associated with a 18.4% ± 2.7% mortality risk, and a peak lactate of 13 mmol/L was associated with a 53% ± 6.8% mortality risk. (Figure 2)

Peak Lactate’s Association with Complication Risk

Elevated peak lactate was observed in the setting of several complications. By linear regression, sepsis correlated with a 5.9±0.3 mmol/L increase in peak lactate (n=58), and cardiac arrest was associated with a 5.7±0.2 mmol/L increase in peak lactate (n=103). New onset renal failure was associated with a 3.9±0.2 mmol/L increase in peak lactate (n=162), and patients with prolonged ventilation demonstrated a 3.1±0.1 mmol/L increase in peak lactate (n=477). Acute limb ischemia was associated with a 5.3±0.4 mmol/L increase in peak lactate (n=32). Lastly, mesenteric ischemia was associated with the highest increase in peak lactate at 15.3±0.4 mmol/L (n=11). While temporal association between peak lactate rise and complication diagnosis could not be established, a correlation between elevated hyperlactatemia and complication incidence was observed (Supplementary Table 1)

Despite only marginal clinical difference in lactate level (>24 Hours: 3 (2, 4.7) mmol/L; ≤24 Hours: 2.3 (1.7, 3.2) mmol/L), patients with lactate peaks occurring >24 hours post-surgery demonstrated a greater association with certain complications.

Overall, 69% (40/58) of all sepsis cases and 81.8% (9/11) of all mesenteric ischemia cases demonstrated peak lactates outside the first 24 hours post-operatively. By comparison, those with cardiac arrest (57% [59/103]), post-operative renal failure (51% [82/162]), acute limb ischemia (53.1% [17/32]) did not demonstrate a preponderance toward lactate peaks outside the first 24 hours. Lastly, those with prolonged ventilation more often had peak lactates in the first post operative 24 hours versus after (79% [317/477]).

Discussion:

Based on analysis of 7,292 patients, this study uncovers several key associations. First, elevated postoperative peak lactate demonstrated significant association with mortality risk at values > 5 mmol/L, with increasing correlation at values ≥ 7 mmol/L. This association should help inform care escalation in the acute setting. Next, the timing of lactate elevation also affects correlation as peak lactates collected ≥ 24 hours after surgery demonstrated greater association with mortality risk. In fact, those patients in whom lactate peaked beyond 24-hours postoperatively experienced mortality three times higher than their STS-PROM. A 24-hour grace period after surgery may allow for an improved association between rising lactate and mortality risk.

Hyperlactatemia after cardiac surgery has been previously investigated ^7-11. In one of the largest studies previously reported, Haanschoten and colleagues demonstrated heightened mortality with increasing lactate across 7,916 patients. For peak lactate categories of <5 mmol/L, 5-10 mmol/L, and >10 mmol/L, reported mortality rates were 3.7%, 20.4%, and 62.9%, respectively ¹¹. We similarly found lactate > 5 mmol/L to be associated with increased mortality. However, the time of lactate peak was not reported in their analysis. This is important since factors such as longer bypass times, pressor utilization, and intraoperative hypothermia can cause benign elevations in lactate in the absence of true malperfusion⁹. While early lactate values can still guide resuscitation in the post-operative setting, the etiology of early lactate is heterogeneous and should be interpreted in context.

Peak post-operative lactate and mortality were statistically greater in patients undergoing combined valve + CABG procedures compared to valve or CABG alone (Table 3), though not by clinically significant margin. Procedures with higher peak post-operative lactates and greater mortality also had notably longer CPB and crossclamp times as well as higher STS PROM (Table 2). The association between lactate magnitude and timing may be different in more complex cases not captured in our analysis, such as aortic procedures, transplants, multivalve operations, and ventricular assist devices. More complex procedures may demonstrate drivers of hyperlactatemia and thus warrant separate investigation.

Modeling extreme hyperlactatemia presents its own unique challenges. Higher lactate values were limited in sample size, thus constraining analysis. Additionally, extreme hyperlactatemia after cardiac surgery has been defined with arbitrary thresholds in the past, making interpretation difficult ^10-12. Our analysis identified a peak lactate of 7 mmol/L as an inflection point at which mortality risk increased by one order of magnitude from baseline. However, comprehension of extreme lactate elevation still requires judgement, since hyperlactatemia may take on different meanings based on offending disease etiology. A 2015 Mayo Clinic study identified 121 cardiac surgery patients with extreme hyperlactatemia (defined as ≥10 mmol/L). The mortality rate was 40% overall but differed by disease etiology with the highest mortality seen in those with mesenteric ischemia or compartment syndrome ¹². Thus, while extreme hyperlactatemia should always raise concern, clinical context heavily determines the end outcome.

Time of elevated lactate also influences the prognostic ability of lactate. Early elevated lactate values within 24 hours of surgery should be interpreted with caution. ICU admission lactate values did not demonstrate any association with mortality on risk-adjusted analysis, indicating that initial lactate values do not add value in identifying patients at elevated mortality risk. Causes of early hyperlactatemia are manifold and may not contribute to increased mortality ^9,13,14. Hyperglycemia and vasopressor utilization can drive aerobic glycolysis, leading to increases in lactate production. Cumulatively, these factors contribute to type B hyperlactatemia, which occurs in the absence of an oxygen debt. This is typically driven by transient impairment of hepatic lactate clearance and microvascular dysfunction. This state is transient and generally does not represent worsening patient prognosis. Misinterpretation of hyperlactatemia in these contexts can lead to over-resuscitation and should be avoided ^15,16.

Multiple factors affect interpretation of lactate, including magnitude and trend. Peak lactate and time-till-peak lactate were utilized in our study since they together represent the greatest elevation in lactate during the post-operative period and a period where lactate levels trended upward. Time till lactate clearance has also been studied for its association with mortality risk. A 2025 study from Johns Hopkins following 2,514 cardiac surgery patients showed that lactate clearance times greater than 24 hours post-surgery were associated with elevated operative mortality risk (p<0.001), while earlier clearance was not. This further substantiates our findings that elevated lactates after the first post-operative 24 hours better correspond to mortality risk¹⁷. Overall, interpretation of lactate includes analysis of multiple aspects.

The impact of lactate on mortality is largely dependent on the disease process being treated. Compared to other pathologies (i.e sepsis, shock), early hyperlactatemia after cardiac surgery is relatively safe. The average PROM score of the cohort was 2.4% ± 3.8%, and the observed operative mortality was even lower at 1.4%. Hence, model estimates were adjusted with PROM, a highly validated risk model for routine cardiac surgery operations ⁴. PROM allows for comprehensive risk-adjustment but avoids the pitfall of adjusting out the individual mechanisms by which hyperlactatemia arises.

Limitation:

This study has several limitations. First, the authors believe that lactate should not be used as an independent prognostic marker, but as one of many tools in a practitioner's armamentarium. Higher lactate values represented a smaller portion of the study population, which may affect risk estimates precision. Additionally, this analysis was restricted to patients undergoing CABG, valve surgery, and CABG + valve surgery, who had a reported PROM score. Lactate also has an association with mortality in more complex excluded cases. However, these cases are typically associated with higher mortality risk and introduce new mechanisms through which hyperlactatemia can occur.

Additionally, while peak lactate provides a momentary depiction of hyperlactatemia in the post-operative setting, it does not communicate other parameters such as lactate trend and rate of clearance.

Additionally, serum lactates were measured at the discretion of the clinical team, not at pre-determined time points. While this portrays the use of serum lactate in a real-world context, lactate collection was not standardized across all patients. Thus, while peak post-operative lactate provides an interpretable association with mortality risk, it should be interpreted in the context.

Conclusion:

In summary, peak postoperative lactate following cardiac surgery is associated with operative mortality risk. Among CABG and/or valve cases, postoperative lactate values >5 mmol/L correlated with increased mortality, while the risk of mortality rose rapidly above a peak lactate of 7 mmol/L. Early hyperlactatemia within 24 hours postoperatively is difficult to interpret due to benign contributors to hyperlactatemia after cardiac surgery and warrant less concern. In contrast, elevated lactate levels after the first post-operative 24 hours is suggestive of higher mortality risk at a lower threshold and should prompt careful evaluation for malperfusion or other underlying cause.

Supplementary Material

1

Supplementary Figure 1: The unadjusted association between peak lactate and operative mortality and that of admission lactate and operative mortality (A). Peak lactate versus operative mortality, graphical analysis (B). Admission lactate versus operative mortality, graphical analysis (C).

Supplementary Figure 2. Consort diagram illustrating patients excluded from the study as well as grouping according to peak lactate ranges.

Glossary of Abbreviations:

AUC

Area Under Curve

CABG

coronary artery bypass graft

CI

Confidence Interval

ECMO

Extracorporeal Membrane Oxygenation

ICU

Intensive Care Unit

IQR/ (Q1-Q3)

Interquartile Range

OR

Odds Ratio

SAVR

Surgical Aortic Valve Replacement

ROC Curve

Receiver Operator Characteristic Curve

NYHA

New York Heart Association (functional class)

PROM

STS-Risk Score (predicted risk of mortality)

STS

Society of Thoracic Surgery