Intraoperative peak lactate as a predictor of in-hospital mortality following non-cardiac emergency surgery: a retrospective cohort study Authors: Yibo Fu, Yuelun Zhang, Le Shen, Yuguang Huang Published: 2025 ABSTRACT AbstractBackgroundThe association between intraoperative lactate levels and postoperative prognosis following emergency surgery remains inconclusive. We aimed to investigate the prognostic significance of intraoperative lactate levels in predicting in-hospital mortality.MethodsThis single-center, retrospective, observational study was conducted at Peking Union Medical College Hospital from 2017 to 2023. Intraoperative peak lactate levels were analyzed using a restricted cubic spline (RCS) model, with in-hospital mortality designated as the primary outcome. Secondary outcomes were length of stay, ICU stay, and ventilation duration. Prespecified subgroup analyses (by ASA grade and surgery type) and a sensitivity analysis excluding in-hospital deaths for secondary endpoints were performed.ResultsAmong the 2,452 patients included in the study, 161 died postoperatively. Both univariate and multivariate regression analyses demonstrated that the intraoperative peak lactate level was associated with perioperative mortality (odds ratio [OR] 1.306, 95% CI 1.262–1.352, P < 0.001; OR 1.249, 95% CI 1.185–1.318, P < 0.001). The association was consistent across surgery types and appeared stronger in patients with higher ASA grade. The restricted cubic spline model indicated higher intraoperative peak lactate levels were nonlinearly associated with increased in-hospital mortality. Length of hospital stay (LOS), LOS in the intensive care unit (ICU), and duration of mechanical ventilation exhibited a nonlinear (inverted U-shaped) pattern relationship with intraoperative peak lactate levels, with an initial increasing trend followed by a subsequent decline as peak lactate concentrations rose.ConclusionsIntraoperative peak lactate levels showed a dose-dependent nonlinear association with increased in-hospital mortality following emergency surgery and play a significant role in predicting postoperative mortality in noncardiac emergency surgery patients.Trial registrationNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00423-025-03958-4.Keywords: Emergency surgery, Intraoperative, Peak lactate, Mortality, Outcome Background Emergency surgery patients constitute a substantial proportion of hospital admissions and exhibit relatively high mortality rates [1]. Improving the quality of emergency surgery and refining prognosis predictions for these patients remain crucial research objectives [2]. Lactate, an important indicator of tissue perfusion, oxygen supply and circulatory demand imbalance [3], can be used to assess the severity of critical conditions and the extent of trauma and guide treatment strategies in surgical patients [4, 5]. Recent studies have highlighted lactate as a multifunctional metabolite, participating in energy metabolism, cellular signaling, metabolic adaptation, neuroendocrine regulation, reproductive physiology, and even tissue repair and resuscitation [6]. Most of the time, elevated lactate levels reflect tissue hypoperfusion due to macro- or microcirculation dysfunction, reduced oxygen delivery, increased glycolysis, or metabolic inhibition of pyruvate dehydrogenase [7]. Serum lactate levels are widely regarded as surrogate markers of disease severity in critically ill patients, and monitoring lactate trends is essential for evaluating responses to therapeutic interventions [8]. Several studies have explored the correlation between serum lactate levels and perioperative mortality. Tak et al. demonstrated that elevated perioperative lactate levels are significantly associated with increased 90-day mortality in ICU-admitted postoperative patients [8]. Another study identified early postoperative lactate levels as useful predictors of mortality following liver resection [9]. The postoperative or perioperative peak arterial lactate level could also predict the outcome after cardiac surgery [10] and lung transplantation [11]. The perioperative or intraoperative peak lactate levels are associated with patient outcome following high-risk gastrointestinal surgery [12], especially for patients after emergency abdominal surgery [13]. However, from the aspect of emergency surgery, studies of intraoperative peak lactate levels and perioperative mortality remain limited. Moreover, most studies rely on receiver operating characteristic (ROC) curves to determine lactate cutoff values, yet the clinical utility of these thresholds is debatable. This study aims to explore whether a more clinically meaningful cutoff value exists by employing a restricted cubic spline (RCS) model. Additionally, we sought to determine whether increases in lactate levels below conventional cutoff values have any effect on perioperative mortality. In summary, this study investigated the relationship between intraoperative peak lactate levels and perioperative mortality in emergency surgery patients. Methods Design This single-center, retrospective, observational study was conducted at the Peking Union Medical College Hospital. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. A completed STROBE checklist is provided in the Supplementary Material . This study was approved by the ethics committee of Peking Union Medical College Hospital (IRB No. I-23PJ1666, approved on 2023-10-13). Informed consent was waived because of the retrospective nature of the study and anonymized data collection. Study population Patients who underwent emergency surgery between 2017 and 2023 were included. Only patients who underwent invasive blood pressure measurement and blood gas analysis during surgery were considered eligible. Patients with missing key variables or incomplete perioperative records were excluded. Data source Medical records were reviewed to collect data. We extracted data from two institutional sources: the Clinical Data Repository (CDR) and the PeriOperative Patient Safety (POPS) database. Serum lactate levels, blood gas analysis results, blood pressure, and heart rate data were extracted from the POPS database. Patient baseline data, length of hospital and ICU stay, perioperative mortality, and discharge diagnosis data were retrieved from CDR database. Missing data were handled via listwise deletion, as implemented by default in SPSS. Fewer than 5% of the cases had missing values across key variables, minimizing potential bias. Definition and measurement of intraoperative peak serum lactate level “Intraoperative peak lactate” was defined as the single highest arterial lactate value recorded from skin incision to the end of surgery, irrespective of the measurement timepoint. Lactate levels were measured by standard blood gas analyzers at the discretion of the attending anesthesiologist, typically at induction, major intraoperative events, and the end of surgery. If multiple measurements were performed, the highest value was used for analysis. Covariates The following covariates were selected on the basis of prior literature and clinical experience: sex, age, weight, ASA grade, blood pressure, heart rate, intraoperative blood loss, urine output, fluid infusion, RBC infusion, plasma infusion, PaCO2, plasma calcium, plasma potassium, and plasma glucose concentrations. Outcomes The primary outcome was in-hospital mortality after surgery. Deaths and self-discharge against medical advice were considered in-hospital mortality events. The secondary outcomes included LOS in the hospital, LOS in the ICU, and duration of mechanical ventilation. Statistical analysis Baseline characteristics are presented as counts with percentages. Continuous variables were tested for normality via the Shapiro‒Wilk test. Nonnormally distributed variables are presented as medians (interquartile ranges, IQRs), whereas normally distributed variables are reported as the means ± standard deviations (SDs). To facilitate stratified analyses and enhance clinical interpretability, intraoperative peak lactate levels were categorized into four groups: <2 mmol/L, 2–5 mmol/L, 5–10 mmol/L, and > 10 mmol/L. This categorization was based on a combination of existing literature, clinical rationale and distribution characteristics of the study cohort. The < 2 mmol/L group aligns with the upper limit of normal lactate levels commonly used in sepsis and perioperative guidelines (e.g., Sepsis-3 criteria) [14]. The 2–5 mmol/L range represents mild to moderate hyperlactatemia observed in postoperative and critical care settings [8, 13]. The 5–10 mmol/L and > 10 mmol/L groups reflect increasing severity of lactic acidosis, with > 10 mmol/L often associated with poor perfusion and high mortality risk [11, 12]. This categorization allows for exploration of dose–response trends while preserving clinical relevance. The relationship between peak lactate levels and perioperative mortality was assessed via an RCS model. RCS regression was used to model potential non-linear associations between intraoperative peak lactate and outcomes. This approach offers greater flexibility compared to linear models and facilitates clinical interpretation of continuous predictors [15]. The number and position of knots were prespecified at the 5th, 35th, 65th, and 95th percentiles of the lactate distribution. Consistent with prior methodological recommendations, we compared 4- and 5-knot parameterizations following standard guidance (e.g., Harrell; Desquilbet & Mariotti) [16]. We selected the 4-knot RCS as our primary specification because it provided a more parsimonious model with lower AIC/BIC, retained a statistically significant test of non-linearity, and yielded smoother, more clinically plausible curves with stable boundary behavior in the sparsely populated upper tail (> 15 mmol/L), thereby reducing the risk of overfitting or spurious oscillation. The substantive inference was unchanged when a 5-knot specification (5th, 27.5th, 50th, 72.5th, 95th percentiles) was used in sensitivity analyses, including restriction to survivors. Logistic regression was used for univariate and multivariate analyses to determine whether intraoperative serum lactate levels independently predict postoperative mortality. Sensitivity analyses for secondary outcomes were performed excluding patients who died during hospitalization. All analyses were performed using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria) with the rms package for spline modeling, and IBM SPSS version 24.0 for descriptive statistics. A two-sided p value < 0.05 was considered statistically significant. Results A total of 2,452 emergency surgery cases meeting our inclusion criteria were identified (Fig. 1). Table 1 presents the basic characteristics, intraoperative hemodynamics, intraoperative volume variables, intraoperative arterial blood gas variables, and trends for each segmented lactate level group. Among patients with different lactate levels, there were no significant differences in indicators such as BMI. However, as lactate levels increased, a greater proportion of patients were classified as ASA III–V, the average age gradually increased, and the proportion of male patients slightly increased. No significant trends in hemodynamic changes were observed among patients with different lactate levels. Patients with higher lactate levels tended to have increased blood loss, fluid infusion volume, and blood transfusion volume. No significant trend changes in other intraoperative arterial blood gases were observed among patients with different lactate levels. Flow chart for patient enrollment Table 1Baseline characteristics and intraoperative parameters of patients stratified by intraoperative serum lactate levelsCharacteristicsLac ≤ 2N = 13802< Lac ≤ 5N = 7105< Lac ≤ 10N = 241Lac>10N = 121Overall(2452)Basic characteristics Age, year, median (IQR)60 (44, 72)59 (43, 69)60 (47, 69)62 (47, 70)60 (44,71) Male, n (%)735 (53.3)402 (56.6)155 (64.3)73 (60.3)1365 (55.6) Weight, kg, median (IQR)65 (54, 73)64 (55, 72)64 (55, 75)64 (54, 74)64 (55,73) BMI, kg/m223.12 (20.35, 26.02)23.17 (20.51, 26.13)23.43 (19.96, 26.12)23.66 (21.23, 26.85)23.15 (20.34, 26.12)ASA grade, n (%) I-II372 (27.0)108 (15.2)9 (3.7)1 (0.8)496(20.2) III-V991 (71.8)595 (83.8)230 (95.4)119 (98.3)1956 (79.8)Surgical sites, n (%) Abdominal surgery601 (43.6)381 (53.7)77 (32.0)44 (36.4)1103 (45.0) Thoracic surgery35 (2.5)19 (2.7)25 (10.4)12 (9.9)91 (3.7) Neurosurgery80 (5.8)66 (9.3)13 (5.4)2 (1.7)161 (6.6) Orthopedic surgery134 (9.7)28 (3.0)4 (1.7)1 (0.8)167 (6.8) Vascular surgery145 (10.5)60 (8.5)19 (7.9)21 (17.4)245 (10.0) ENT & HN surgery54 (3.9)33 (4.6)5 (2.1)3 (2.5)95 (3.9) Gynecological surgery75 (5.4)34 (4.8)0 (0)1 (0.8)110 (4.5) Urologic surgery13 (0.9)4 (0.6)2 (0.8)0 (0)19 (0.8)Obstetric surgery73 (6.0)49 (6.9)5 (2.1)2 (1.7)129 (5.3)Intraoperative hemodynamics SBPmax, mmHg, median (IQR)150 (143, 166)145 (131, 164)139 (126, 160)142 (124, 159)147 (132, 165) SBPmin, mmHg, median (IQR)91 (82, 103)88 (79, 101)89 (80, 102)55 (42, 64)91 (81, 103) DBPmax, mmHg, median (IQR)94 (83, 105)93 (83, 107)89 (80, 103)95 (77, 115)93 (83, 106) DBPmin, mmHg, median (IQR)57 (49, 65)54 (46, 64)56 (48, 65)55 (42, 64)56 (48, 65) HRmax, bpm, median (IQR)112 (100, 128)121 (108, 140)137 (116, 158)143 (124, 174)117 (104, 137) HRmin, bpm, median (IQR)61 (52, 73)66 (55, 82)68 (42, 88)52 (25, 90)63 (52, 77)Intraoperative volume variables Blood loss, ml, median (SD)195.53(412.53)343.07(687.64)795.99(1715.68)1041.78(2193.70)338.3 (898.4) Urine volume, ml, median (SD)463.44(587.36)510.27(630.95)611.10(775.32)436.57(528.71)490.1 (619.5) Fluid infusion, ml, median (SD)2242.38 (1277.84)2794.29(1556.05)3257.78(2228.04)2928.21(2206.92)2533.8 (1569.3) Packed RBC infusion, U, median (SD)1.57(24.47)2.89(46.62)3.77(39.04)9.18(74.62)2.8 (40.6) Plasma infusion, ml, median (SD)89.94(199.69)169.76(306.60)380.34(584.01)555.51(765.13)164.1 (355.3)Intraoperative arterial blood gas variables PaCO2, mmHg median (IQR)37.9 (34.4, 41.7)38.6 (34.5, 42.6)40.0 (36.3, 45.5)40.5 (36.3, 45.4)38.4 (34.7, 42.5) Ca2+, mmol/L median (IQR)1.15 (1.10, 1.19)1.14 (1.08, 1.19)1.09 (1.04, 1.15)1.07 (0.97, 1.14)1.14 (1.09, 1.19) K+, mmol/L median (IQR)3.8 (3.5, 4.2)3.9 (3.6, 4.2)3.9 (3.6, 4.4)3.9 (3.6, 4.4)3.82 (3.56, 4.20)Glucose, mmol/L median (IQR)7.3 (6.1, 9.3)8.9 (7.1, 11.2)9.3 (7.1, 12.6)9.1 (6.8, 14.0)7.9 (6.4, 10.2) Baseline characteristics and intraoperative parameters of patients stratified by intraoperative serum lactate levels 195.53 (412.53) 343.07 (687.64) 795.99 (1715.68) 1041.78 (2193.70) 463.44 (587.36) 510.27 (630.95) 611.10 (775.32) 436.57 (528.71) 2794.29 (1556.05) 3257.78 (2228.04) 2928.21 (2206.92) 1.57 (24.47) 2.89 (46.62) 3.77 (39.04) 9.18 (74.62) 89.94 (199.69) 169.76 (306.60) 380.34 (584.01) 555.51 (765.13) A total of 5,161 emergency surgery patients were initially assessed from January 2017 to July 2023. After sequentially excluding patients without intraoperative arterial blood gas analysis (n = 1,605), patients younger than 18 years (n = 105), those with incomplete baseline data (n = 932), and duplicate records (n = 67), a final cohort of 2,452 patients was included in the analysis. Among the 2,452 patients included in the study, 161 patients died during hospitalization, while the remaining patients survived (Table 2). There is a significant association between lactate levels and patient outcomes. Patients with elevated lactate levels (> 2 mmol/L) have a higher mortality rate than those with lower lactate levels (< 2 mmol/L), and as lactate levels increase further, mortality rates also increase. Patients with higher lactate levels require longer periods of mechanical ventilation, longer stays in the ICU, and longer hospital stays. Table 2In-hospital mortality and secondary outcomes stratified by intraoperative serum lactate levelsCharacteristicsLac ≤ 2N = 13802< Lac ≤ 5N = 7105< Lac ≤ 10N = 241Lac>10N = 121OverallN = 2452Primary outcome In-hospital mortality, n (%)10 (0.7)42 (5.9)47 (19.5)62 (51.2)161 (6.6)Secondary outcomes Length of mechanical ventilation, hour, median (IQR)3 (0, 13)10 (0, 47)29 (4, 127)36 (7, 99)6 (0, 30) Length of stay in ICU, day, median (IQR)0.8 (0, 2.0)1.7 (0.1, 4.7)2.7 (0.7, 7.3)2.6 (0.5, 7.6)1.1 (0, 3.4) Length of stay in hospital, day, median (IQR)9 (6, 16)11 (6, 23)15 (9, 27)14 (6, 31)10 (6, 20) In-hospital mortality and secondary outcomes stratified by intraoperative serum lactate levels We subsequently performed univariate and multivariate regression analyses to explore the factors associated with perioperative mortality (Table 3). Univariate and multivariate regression analyses both revealed that the intraoperative peak lactate level was an independent predictor of perioperative mortality (odds ratio [OR] 1.306, 95% CI 1.262 to 1.352; P < 0.001 and OR 1.249, 95% CI 1.185 to 1.318; P < 0.001). Univariate regression analysis also revealed that male sex (OR 1.582, 95% CI 1.131 to 2.214, P = 0.007), older age (OR 1.015, 95% CI 1.005 to 1.024, P = 0.002), lower weight (OR 0.989, 95% CI 0.980 to 0.998, P = 0.019), higher ASA grade (P < 0.001), higher intraoperative RBC infusion (OR 1.003, 95% CI 1.001 to 1.005, P = 0.009), higher intraoperative plasma infusion (OR 1.001, 95% CI 1.001 to 1.001, P < 0.001), lower Ca2+ concentration (OR 0.008, 95% CI 0.002 to 0.038, P < 0.001), and lower blood glucose level (OR 1.021, 95% CI 1.002 to 1.041, P = 0.038) lead to greater perioperative mortality. Multivariate regression analysis revealed that older age (OR 1.024, 95% CI 1.006–1.042, P = 0.008) and higher ASA grade (P < 0.001) were associated with higher perioperative mortality. Table 3Univariate and multivariate logistic regression analyses of factors associated with postoperative outcomesUnivariate logistic regressionMultivariate logistic regressionVariablesOdds ratio95% CIpOdds ratio95% CIpIntraoperative Lactate, mmol/L1.3061.262 to 1.352< 0.0011.2491.185 to 1.318< 0.001Sex, male1.5821.131 to 2.2140.0071.0350.643 to 1.6640.232Age, year1.0151.005 to 1.0240.0021.0241.006 to 1.0420.008BMI, kg/m20.9740.934 to 1.0150.212ASA grade0.070.022 to 0.221< 0.0010.1930.046 to 0.8090.024Intraoperative bleeding, ml1.0001.000 to 1.0000.011.0000.999 to 1.0000.123Intraoperative urine, ml1.0001.000 to 1.0000.305Intraoperative fluid infusion, ml1.0001.000 to 1.0000.041.0001.000 to 1.0000.933Intraoperative RBC infusion, U1.0031.001 to 1.0050.0091.0040.999 to 1.0090.126Intraoperative plasma infusion, ml1.0011.001 to 1.001< 0.0011.0001.000 to 1.0010.239PaCO2, mmHg1.0091.002 to 1.0170.0191.0020.981 to 1.0230.854Ca2+, mmol/L0.0080.002 to 0.038< 0.0010.3450.023 to 5.2730.444K+, mmol/L1.0010.959 to 1.0460.954Glucose, mmol/L1.0211.002 to 1.0410.0321.0040.949 to 1.0630.881 Univariate and multivariate logistic regression analyses of factors associated with postoperative outcomes Intraoperative peak lactate level as a predictor of death On the basis of the results of univariate and multivariate regression analyses, intraoperative serum lactate levels emerged as one of the most significant indicators for assessing patient prognosis. Therefore, the serum lactate level was selected for further investigation. On the basis of the results of the multivariable regression analysis, age and ASA classification were identified as confounding factors and were included in the RCS analysis. The RCS curve revealed that the intraoperative peak serum lactate level was significantly and positively correlated with the postoperative mortality rate (Fig. 2). In the RCS analysis, a reference point was established at a lactate level of 2 mmol/L, with the corresponding odds ratio (OR) fixed at 1. The RCS curve demonstrated a monotonic increase in mortality risk with rising intraoperative lactate levels. Although the test for nonlinearity was statistically significant (P < 0.05), the curve lacked a clear inflection point, suggesting a continuously increasing risk without a definitive threshold. According to the RCS framework described by Gauthier et al., this type of pattern reflects a dose–response relationship that may not be adequately captured by conventional linear or categorical models [15]. Nonlinear association between intraoperative peak lactate and in-hospital mortality Clinically, this implies that even mild elevations in lactate—below commonly accepted thresholds—may be associated with increased mortality, highlighting the importance of interpreting lactate as a continuous rather than dichotomous risk factor. A restricted cubic spline model was used to examine the adjusted odds ratio (OR) for in-hospital mortality, with 2 mmol/L as the reference point. The model was adjusted for age and ASA classification. The curve demonstrates a progressive increase in mortality risk with rising lactate levels.The red line indicates the estimated odds ratio, with the shaded region representing the 95% confidence interval. Intraoperative serum lactate peak levels and secondary outcomes We performed RCS analyses to examine the relationships between intraoperative serum lactate peak levels and three secondary outcomes: LOS in the hospital, LOS in the ICU, and duration of mechanical ventilation. Corresponding curves were generated for each outcome (Fig. 3). A nonlinear association was observed between intraoperative peak lactate levels and perioperative secondary outcomes. The durations of postoperative hospitalization, intensive care unit stay, and mechanical ventilation requirement initially showing progressive prolongation from the prespecified reference of 2 mmol/L to a broad maximum (≈ 5–7 mmol/L), then declined at higher lactate levels, yielding an inverted-U shape (Fig. 3). We performed a prespecified sensitivity analysis excluding in-hospital deaths. When restricted to survivors, the descending limb beyond ~ 8–10 mmol/L was substantially attenuated or flattened, most notably for ICU stay and ventilation time, while the curves below ~ 5 mmol/L were nearly unchanged (Supplementary Fig. 1). Nonlinear associations between intraoperative peak lactate and secondary outcomes. (A) ICU length of stay; (B) hospital length of stay; (C) duration of mechanical ventilation. RCS multivariable OLS models were used with four knots at the 5th, 35th, 65th, and 95th percentiles of the lactate distribution and adjusted for age and ASA physical status. Curves depict predicted values at the cohort’s median age and most frequent ASA level; shaded areas indicate 95% CIs. The vertical dashed line marks lactate = 2 mmol/L, and the horizontal dashed line denotes the model-predicted outcome value at lactate = 2 mmol/L. X-axis: intraoperative peak lactate (mmol/L, truncated to 0–15) Subgroup analysis When stratified by surgery type, the RCS curves showed a broadly monotonic increase in the odds of in-hospital death with rising intraoperative peak lactate in both abdominal and non-abdominal procedures (Fig. 4). The slope appeared steeper among abdominal cases at higher lactate values, but confidence bands overlapped across most of the range. Formal interaction testing in a per-1 mmol/L framework showed no evidence of effect modification by surgery type (interaction p = 0.432; Supplementary Fig. 2). In contrast, ASA grade significantly modified the lactate–mortality association, with a markedly stronger effect among ASA III–V than ASA I–II patients (interaction p < 0.001; Supplementary Fig. 2). Association of intraoperative peak lactate with in-hospital mortality by surgery type (A) Abdominal surgery. (B) Non-abdominal surgery. Multivariable logistic regression using restricted cubic splines (4 knots at the 5th, 35th, 65th, and 95th percentiles of lactate), adjusted for age and ASA grade. Predictions are shown at the cohort median age and the modal ASA grade; x-axis truncated to 0–15 mmol/L. The y-axis shows adjusted odds ratios relative to lactate = 2 mmol/L (vertical dashed line). The horizontal dashed line marks OR = 1. Shaded ribbons denote 95% confidence intervals. See Supplementary Fig. 2 for formal interaction tests. For ICU length of stay, abdominal surgery showed an inverted-U shape peaking around ~ 5–7 mmol/L and then gradually decreasing, whereas non-abdominal cases rose more steadily with a shallower curvature (Fig. 5). For hospital length of stay, both subgroups displayed a convex pattern with a plateau beyond mid-range lactate, with abdominal cases peaking slightly earlier (Fig. 5). For ventilation duration, abdominal surgery exhibited a peak around ~ 6 mmol/L followed by a decline, while non-abdominal cases increased more monotonically across the lactate range (Fig. 5). At the common reference of 2 mmol/L, model-based predictions were numerically lower for abdominal than non-abdominal patients for ICU stay (~ 3.0 vs. ~ 4.8 days), hospital stay (~ 17.5 vs. ~ 19.7 days), and ventilation time (~ 35.5 vs. ~ 66.2 h), but these subgroup contrasts should be interpreted cautiously given model adjustment and uncertainty at distribution tails. Associations of intraoperative peak lactate with secondary outcomes by surgery type (A–B) ICU length of stay (days): abdominal vs. non-abdominal. (C–D) Hospital length of stay (days): abdominal vs. non-abdominal. (E–F) Duration of mechanical ventilation (hours): abdominal vs. non-abdominal. Multivariable ordinary least-squares regression with restricted cubic splines for lactate (4 knots at the 5th, 35th, 65th, and 95th percentiles), adjusted for age and ASA grade. Predictions are presented at the cohort median age and modal ASA grade; x-axis truncated to 0–15 mmol/L. Vertical dashed line marks lactate = 2 mmol/L. Horizontal dashed lines indicate the model-predicted outcome at 2 mmol/L for each panel. Shaded ribbons show 95% confidence intervals. Outcomes are expressed in days for ICU and hospital stay and in hours for ventilation duration. Discussion In this large single-center retrospective cohort study, we found that intraoperative peak lactate levels were independently and nonlinearly associated with in-hospital mortality, with no distinct threshold observed. The risk of death increased progressively with rising lactate levels, even within traditionally considered “normal” ranges. This finding challenges the assumption of fixed lactate thresholds commonly used in clinical practice and suggests that lactate monitoring may be more useful as a continuous risk marker rather than a binary threshold-based approach.Currently, it is generally accepted in clinical practice that the normal range of arterial serum lactate levels is less than 1.5 mmol/L, and the Sepsis-3 guidelines incorporate serum lactate levels of > 2 mmol/L in the definition of septic shock [14]. However, there is currently no precise definition of the normal range for lactate levels during surgery. A retrospective cohort study involving 443 patients admitted to an intensive care unit (ICU) due to severe sepsis or septic shock revealed that a baseline blood lactate level exceeding 2.5 mmol/L presented the highest area under the ROC curve for predicting mortality within 28 days [17]. Sujita et al. reported that the intraoperative lactate level was a strong predictor of in-hospital mortality in patients who underwent emergency abdominal surgery (AUC = 0.762) [13]. The majority of current studies on serum lactate and postoperative mortality derive a cutoff value via receiver operating characteristic (ROC) analysis. The principle behind deriving a cutoff value through ROC analysis is to identify a point that balances sensitivity and specificity. In actual clinical practice, decisions regarding different clinical indicators often involve trade-offs between sensitivity and specificity, depending on which aspect is deemed more critical for the specific clinical context. Therefore, the clinical utility of cutoff values derived solely from ROC analysis, which simply balances sensitivity and specificity, remains open to debate. In contrast, our study’s methodology, employing a restricted cubic spline (RCS) model, provides a nuanced understanding of the relationship between lactate levels and mortality. This approach allows for the identification of non-linear associations and offers a more precise estimation of risk compared to traditional threshold-based models. By integrating lactate measurements into perioperative risk assessment, clinicians can better stratify patients and tailor interventions to improve surgical outcomes. Intraoperative hyperlactatemia is a significant marker of metabolic and circulatory disturbances during surgery [13]. Lactate is primarily produced through anaerobic glycolysis when oxygen delivery to tissues is insufficient [18]. However, elevated lactate levels can also result from increased glycolytic activity in well-oxygenated tissues, such as muscles and the brain, during periods of heightened metabolic demand. This phenomenon underscores the complexity of lactate metabolism and its dual role as both a metabolic byproduct and a potential signaling molecule [19]. Our study’s finding of a continuous, dose-dependent relationship between intraoperative lactate levels and in-hospital mortality aligns with recent research indicating that even modest elevations in lactate can predict adverse outcomes. The pathophysiological mechanisms underlying elevated lactate levels in surgical patients are multifactorial. Intraoperative factors such as blood loss, tissue hypoperfusion, and the inflammatory response to surgery can lead to anaerobic metabolism and lactate accumulation. Additionally, impaired lactate clearance due to hepatic or renal dysfunction can exacerbate hyperlactatemia. Recent studies have highlighted the importance of monitoring lactate levels during surgery, as they can provide early indications of adverse events and guide therapeutic interventions [20]. Furthermore, lactate’s role extends beyond being a mere metabolic byproduct. It has been identified as a signaling molecule involved in various physiological processes, including immune modulation and tissue repair. Elevated lactate levels may reflect a systemic response to stress and injury, influencing outcomes in critically ill patients [21]. The conclusions of this study differ in some respects from existing clinical research findings, which is a key focus of this discussion. Although current guidelines define serum lactate levels of 1.5 mmol/L as normal and some studies suggest a cutoff value of 2.5 mmol/L, we question whether this implies that lactate elevation below 2.5 mmol/L has no effect on postoperative mortality. The findings of this study indicate otherwise. Even within the so-called normal range, mortality continues to increase as lactate levels rise. This serves as a critical warning: even at relatively low levels, increased serum lactate warrants close attention and vigilance. The relationships between several secondary outcomes and lactate levels present intriguing clinical implications. As depicted in the RCS curves, the durations of hospitalization, ICU stay, and ventilator dependence demonstrated a characteristic inverted U-shaped pattern that correlated with intraoperative peak lactate elevations. Notably, this paradoxical phenomenon might be theoretically explained through a dual-pathway mechanism: moderate lactate accumulation likely reflects escalating disease severity and physiological decompensation, necessitating prolonged therapeutic interventions. However, at critically elevated concentrations, the observed decline in care durations may correspond to clinical decision-making shifts associated with deteriorating prognosis, either through physician–family consensus for palliative transition or mortality-driven attrition, potentially truncating active treatment periods. Overall, this study is the first to validate the role of serum lactate levels in postoperative mortality in emergency surgical patients via the RCS model. The findings of this study suggest that intraoperative lactate levels may serve as a valuable marker for perioperative risk stratification in emergency surgical patients. Given the consistent association between elevated lactate and adverse outcomes, integrating lactate measurements into intraoperative monitoring protocols may help identify high-risk patients earlier. However, the retrospective nature and single-center design of our study limit causal inference. Future prospective, multicenter cohort studies are warranted to validate the predictive utility of intraoperative lactate and to determine whether real-time lactate-guided interventions can improve postoperative outcomes. Additionally, integrating lactate dynamics with other physiological indicators may further enhance individualized risk prediction and clinical decision-making. Several limitations should be acknowledged in this study. First, the retrospective and single-center nature of the analysis limits the generalizability of the findings and precludes causal inference. Second, detailed information regarding surgical indications, including the presence or absence of sepsis, was not available in the dataset, which may have influenced both lactate levels and outcomes. Third, there may be a degree of selection bias, as only patients who underwent invasive arterial pressure monitoring and intraoperative blood gas analysis were included. This subgroup likely represents a sicker population, and the findings may not be generalizable to all patients undergoing emergency surgery. Additionally, certain comorbidities and intraoperative interventions may have been incompletely captured, which could confound the observed associations. Conclusions Intraoperative peak lactate levels showed a dose-dependent nonlinear association with increased in-hospital mortality after emergency surgery and play a significant role in predicting postoperative outcomes. This study highlights the prognostic value of lactate monitoring during emergency surgery and suggests potential pathophysiological mechanisms involving tissue hypoperfusion and metabolic stress. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 1 Acknowledgements Not applicable. Author contributions YF collected the data, analyzed and interpreted the patient data, and drafted the manuscript. YZ validated and reviewed the statistics of the study. LS designed and supervised the study, validated the results and drafted the manuscript. YH supervised the study. All the authors read and approved the final manuscript. Funding This work was supported by the Chinese Academy of Medical Sciences Initiative for Innovative Medicine (2024-12 M-TS-001) and the National Major Science and Technology Projects of China (2023ZD0501801). Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This retrospective cohort study was approved (Ethical Committee No. I-23PJ1666), and the need for consent to participate was waived by the Ethical Committee of Peking Union Medical College Hospital, Beijing, China, on October 11, 2023. Our study adhered to the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 1 Supplementary Material 1 Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.