Is postoperative ARDS different from medical ARDS?

Abstract

Background

Postoperative patients have been underrepresented in randomized controlled trials of acute respiratory distress syndrome (ARDS). Whether postoperative ARDS differs from medical ARDS in its clinical trajectory, outcomes, and prognostic determinants remains unclear. We aimed to compare postoperative and medical ARDS with respect to early trajectory, mortality, and risk factors for mortality.

Methods

We conducted a retrospective analysis of prospectively collected data in a tertiary ICU from 2003 to 2023. All consecutive intubated adults fulfilling the ARDS New Global Definition were included. ARDS cases were labeled as postoperative when onset occurred within 15 days after surgery. The primary outcome was 90-day mortality, assessed with multivariable Cox analysis. Early ARDS trajectories were assessed at day 3. Multivariable Cox analyses were used to identify factors independently associated with mortality.

Results

Among 1,077 intubated ARDS patients, 455(42%) had postoperative ARDS. Compared with medical ARDS, postoperative ARDS showed more favorable early trajectories (p = 0.03) and lower 90-day mortality (36% vs. 49%, p < 0.001). Postoperative ARDS remained independently associated with lower 90-day mortality after adjustment (adjusted hazard ratio[aHR] = 0.68, 95%CI:0.56–0.83, p < 0.001). Prognostic determinants differed markedly. In postoperative ARDS, mortality was independently associated with extrapulmonary organ dysfunction (non-respiratory SOFA score: aHR = 1.10, 95%CI:1.05–1.15; bicarbonate: aHR = 0.81 per 5mmol/L, 95%CI:0.69–0.96) and surgical context, (esophageal surgery: aHR = 0.41, 95%CI:0.24–0.70; upper abdominal surgery: aHR = 0.64, 95%CI:0.46–0.91 versus lower abdominal surgery), while no marker of respiratory failure was independently associated with mortality. In medical ARDS, mortality was independently associated with respiratory failure, including PaO₂/FiO₂ ratio (aHR = 0.88 per 50mmHg, 95%CI:0.81–0.96) and driving pressure (aHR = 1.13 per 5cmH₂O, 95%CI = 1.01–1.27), and extrapulmonary organ dysfunction.

Conclusion

Postoperative ARDS differs from medical ARDS in its early clinical trajectory, outcomes, and prognostic determinants. These findings support postoperative ARDS as a distinct clinical subtype, mainly driven by extrapulmonary and surgery-related factors rather than by the lung injury itself, supporting a management approach focused on perioperative prevention and early identification of surgery-related complications.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-026-06112-8.

Background

Acute respiratory distress syndrome (ARDS) remains a leading cause of death in the intensive care unit (ICU), with contemporary mortality estimates ranging from 35% to 50% despite advances in supportive care [1, 2]. The definition of ARDS was revised in 2023 with broader criteria and updated guidelines [1, 3]. ARDS is not a single disease but a clinical syndrome that encompasses heterogeneous etiologies, trajectories, and responses to treatment [4, 5]. Within this spectrum, ARDS occurring in the postoperative setting represents a distinctive and understudied subgroup [6, 7]. Postoperative patients combine surgical stress, anesthesia-related physiological perturbations, fluid shifts, transfusion exposure, and procedure-specific factors (e.g., increased intra-abdominal pressure after abdominal surgery), all of which may shape the onset, severity, and early course of lung injury [8, 9].

Yet, despite the clinical relevance of postoperative ARDS, surgical patients have been consistently underrepresented in major ARDS randomized controlled trials, limiting the generalizability of trial-derived evidence to perioperative populations [6]. Among the ARDS Network and Prevention and Early Treatment of Acute Lung Injury trials, postoperative ARDS accounted for 4.8% of all ARDS [6]. This proportion has decreased, with 2.7% of postoperative ARDS after 2011. In contrast, observational cohorts suggest that postoperative ARDS represents a substantial proportion of cases, accounting for 30% to 50% of all episodes encountered in routine practice [4, 7].

Diagnostic criteria and guidelines on ventilatory management remain similar between postoperative and medical ARDS [1, 9]. However, perioperative respiratory management has significantly improved with intraoperative lung-protective ventilation [8, 10] and postoperative noninvasive respiratory support strategies [11–13]. Meanwhile, surgery has known some important paradigm shifts, with the development of minimally invasive surgery [14, 15]. It is unclear whether these advances have impacted the outcomes of postoperative ARDS over time. In a post-hoc analysis of randomized controlled trials, there was no change in mortality over time in postoperative ARDS [6]. However, the sample size was low, as only 256 postoperative ARDS patients were included over 20 years (13 patients per year) [6].

Accordingly, we aimed to assess whether postoperative ARDS differs from medical ARDS in baseline characteristics and severity at onset, ventilatory management, early clinical trajectories following ARDS diagnosis, and outcomes, including their evolution over a 20-year period, as well as to identify factors independently associated with mortality. We hypothesized that postoperative ARDS would be characterized by a more favorable early trajectory and lower mortality than medical ARDS, with distinct prognostic factors.

Methods

Study design

This study is a retrospective analysis of a prospective cohort of consecutive adult ARDS patients admitted to a 20-bed mixed medical–surgical ICU specialized in abdominal surgery between 2003 and 2023. The study was reported according to STROBE guidelines [16], approved by the Montpellier University Institutional Review Board (agreement number: 2019_IRB-MTP_05–25), and performed in accordance with the Declaration of Helsinki of 1975.

Data collection

Patients were identified through the ICU’s database. ARDS diagnosis was retrospectively adjudicated from the medical records and independently assessed by two physicians (JP and JH) according to the New Global definition [3]: 1/timing: onset within one week; 2/chest imaging including ultrasound: bilateral opacities, not fully explained by effusions, atelectasis, nodules or masses; 3/origin of oedema: respiratory failure not fully explained by cardiac failure or fluid overload; 4/PaO₂/FiO₂ ratio ≤ 300mmHg with PEEP or SpO₂/FiO₂ ratio ≤ 315. Medical records and chest imaging were manually reviewed irrespective of the clinical diagnosis made during routine care. This approach ensured a consistent and standardized application of ARDS criteria across the entire study period, given that clinical diagnoses were initially based on earlier definitions. Patients without sufficient imaging data to confirm bilateral opacities were not classified as ARDS. In case of uncertainty, cases were discussed to reach consensus. Only intubated patients were included in the primary analysis, and non-intubated patients were included in a sensitivity analysis. ARDS severity was assessed at ARDS onset based on the New Global Definition [3]. Rapidly improving ARDS was defined either as a resolved hypoxemia (PaO₂/FiO₂ >300 mmHg or SpO₂/FiO₂ >315) or as successful weaning from invasive mechanical ventilation with unassisted breathing at 48 h, as previously described [17, 18]. Non-rapidly improving ARDS was considered persistent ARDS, as previously described [17, 18].

ARDS cases were considered postoperative when onset occurred within 15 days after surgery [19]. This definition was extracted from an individual patient meta-analysis showing that postoperative lung injury occurred within 15 days in all 3,365 included patients [19]. Other ARDS cases were considered medical. Since surgical patients were admitted to the ICU after abdominal surgery, surgery site was classified in lower abdominal surgery (colorectal surgery and gynecologic surgery), upper abdominal surgery (gastro-duodenal, pancreatic, upper intestinal, hepatobiliary, and urology surgery) and esophageal surgery [20].

All consecutive adult patients admitted to the ICU between 2003 and 2023 who developed ARDS and underwent invasive ventilation were included in the study. Only the first admission with ARDS was analyzed. Collected data included demographics, anthropometrics, clinical and biological data, the Sequential Organ Failure Assessment (SOFA score), the Simplified Acute Physiology Score II (SAPS II), recorded at ARDS onset [21]. Ventilation characteristics [Tidal volume (Vt), Positive End-Expiratory Pressure (PEEP), Plateau Pressure, static Driving Pressure (Plateau Pressure–PEEP), Respiration Rate] were recorded on day one of ARDS onset [21]. Vt and respiratory rate were automatically extracted, while pressure variables were manually recorded at the bedside every 4 h. We used the median over the corresponding time period. Controlled ventilation was performed using volume-assist control modes and assisted ventilation using pressure support ventilation. Plateau pressure was measured under controlled ventilation using a standardized end-inspiratory pause. When deemed insufficient, an additional pause maneuver was performed to confirm plateau pressure. Respiratory system compliance (Crs) was calculated with the formula Vt/ (Plateau Pressure–PEEP) [21]. All patients were treated with a lung-protective ventilation protocol: low Vt targeting 6 ml/kg of predicted body weight (PBW), limited Plateau Pressure, PEEP ≥5cmH₂O [21]. Ventilatory parameters were set to avoid intrinsic PEEP.

Outcomes

The primary outcome was 90-day mortality. Secondary outcomes were early trajectory of ARDS, SAPS II, SOFA score, PaO₂/FiO₂ ratio, and ventilatory parameters at ARDS onset.

Statistical analysis

A descriptive analysis was performed; qualitative variables were expressed as numbers(%) and compared using the uncorrected chi-square test or Fisher test. Quantitative variables were expressed as mean [standard deviation (SD)] or median (interquartile range, 25–75%) and compared using the Student-t test or Wilcoxon test as appropriate [21]. To anticipate the sample size, we estimated a 90-day mortality in medical ARDS of 40% [2, 21, 22]. To detect a 10% absolute 90-day mortality reduction in postoperative ARDS with 40% of postoperative ARDS, a power of 0.90 and a Type I error of 0.05, we calculated a required sample size of 1,045 patients [23].

Firstly, we compared early trajectories of ARDS between postoperative and medical patients, categorized into five ordinal classes on day 3: death, severe, moderate, mild, and Rapidly-Improving. A Cochran-Armitage test was used [21], Sankey plots were designed [4].

Secondly, we plotted Kaplan–Meier survival curves until day 90 and compared them with the log-rank test [24]. We performed a univariable Cox-proportional hazard model to assess whether postoperative ARDS was associated with 90-day mortality compared to medical ARDS. We performed a multivariable Cox-proportional hazard model after adjustment for age, SOFA score, and ARDS severity at baseline [21]. Collinearity between variables was tested. Complete case analysis was used for this analysis, as missing data was ≤ 5% for all variables [25]. We performed two a priori sensitivity analyses: one including ARDS patients treated with noninvasive respiratory support, the second restricting to persistent ARDS (excluding rapidly-improving ARDS) [17, 18]. As a post-hoc sensitivity analysis, postoperative ARDS was redefined using a 7-day window after surgery.

Thirdly, time trends of day-90 mortality in ARDS patients were assessed with the Cochran–Armitage test [26]. The mean change was assessed with Sen’s slope test. Temporal trends were plotted using restricted cubic splines, with 3 knots positioned at Harrell’s recommended percentiles. Models were adjusted for age, SOFA score, and PaO₂/FiO₂ ratio; interaction terms were tested. Time trends in SAPS II and SOFA score were assessed with the Mann-Kendall test [21]. Corresponding figures were plotted using linear regression models, after confirming the validity of the linearity assumption. We performed two sensitivity analyses: one including ARDS patients were treated with noninvasive respiratory support, and the second restricting to persistent ARDS.

Fourthly, risk factors for 90-day mortality were assessed separately in postoperative and medical ARDS. Clinically relevant variables were prespecified and included demographics (age, sex), comorbidities (diabetes, chronic cardiac failure, chronic respiratory failure, cirrhosis, cancer), ARDS origin (pulmonary vs. extrapulmonary), markers of respiratory failure at ARDS onset (PaO₂/FiO₂ ratio, driving pressure, PaCO₂), and markers of extrapulmonary organ failure (non-respiratory SOFA score, serum bicarbonate). In postoperative ARDS, surgery site (lower abdominal vs. upper abdominal and esophageal surgery) was additionally included. Univariable Cox-proportional hazard models were conducted; variables with a p-value < 0.20 were entered into multivariable Cox-proportional hazard models. Missingness reached 32% for driving pressure; therefore, missing values were handled using multiple imputation. Predictive mean matching was used with fifteen imputed datasets, and estimates were pooled according to Rubin’s rules. We used the R software (version 4.2.1) to perform statistical analyses. Tests were two-sided and p < 0.05 were considered significant.

Results

Patients

Between 2003 and 2023, 13,555 patients were admitted to ICU. Among them, 1,331 patients (10%) met the New Global Definition criteria for ARDS. After exclusion of 254 patients (19%) treated only with noninvasive respiratory support (noninvasive ventilation and/or high flow nasal oxygenation), 1,077 (81%) intubated ARDS patients were included. All patients had an available mortality status at day 90. Among these, 455 patients (42%) were classified as postoperative ARDS and 622 patients (58%) as medical ARDS. Over the study period, 83,565 surgical procedures were performed in the institution, corresponding to an incidence of postoperative ARDS of 0.54%. Figure 1 shows the study flow chart. The proportion of postoperative ARDS was stable over the study period (Fig. S1).

Fig. 1.

Study flowchart and 90-day outcomes according to ARDS etiology

Postoperative ARDS patients were older, with higher body mass indexes, and had more frequently a history of cancer (40% vs. 18%, p < 0.001) than medical ARDS patients (Table 1). Postoperative ARDS was also associated with lower SAPS II and SOFA score than medical ARDS. Among postoperative ARDS patients, 85(19%) underwent esophageal surgery, 246(54%) upper abdominal surgery, and 124(27%) lower abdominal surgery. Surgery was urgent for 199(44%) patients. Postoperative ARDS occurred within 7 days after the surgery in 419/455 patients (92%, median: 2 days [interquartile range, 0–4]). Details about the surgery are shown in Table S1 and Fig. S2.

Table 1.

Baseline characteristics in 1077 patients with ARDS

Characteristics	Total (%) (n = 1077)	Postoperative ARDS (n = 455)	Medical ARDS (n = 622)		P value
Age (years)	61 (± 14)	63 (± 13)		60 (± 15)	< 0.001
Sex					0.44
Male	733 (68%)	316 (69%)		417 (67%)
Female	344 (32%)	139 (31%)		205 (33%)
BMI (kg/m2), mean (± SD)	27 (± 7)	28 (± 8)		27 (± 7)	0.01
High blood pressure	345 (32%)	165 (36%)		180 (29%)	0.01
Diabetes mellitus	204 (19%)	97 (21%)		107 (17%)	0.07
Chronic respiratory failure	161 (15%)	71 (16%)		90 (14%)	0.67
Chronic cardiac failure	131 (12%)	55 (12%)		76 (12%)	0.99
Chronic kidney disease	85 (8%)	35 (8%)		50 (8%)	0.92
Cirrhosis	185 (17%)	64 (14%)		121 (20%)	0.02
Cancer	295 (28%)	182 (40%)		113 (18%)	< 0.001
SAPS II, mean (± SD)	53 (± 19)	50 (± 18)		55 (± 19)	< 0.001
SOFA, mean (± SD)	11 (± 4)	10 (± 4)		11 (± 5)	< 0.001
Time between ICU admission and ARDS onset (days), median [IQR]	0 [0–2]	1 [0–3]		0 [0–2]	0.03
Origin of ARDS					< 0.001
Pulmonary	564/986 (57%)	190/432 (44%)		374/554 (68%)
Extrapulmonary	422/986 (43%)	242/432 (56%)		180/554 (32%)
Severity of ARDS					0.27
Mild	180 (17%)	79 (17%)		101 (17%)
Moderate	530 (49%)	211 (47%)		319 (51%)
Severe	367 (34%)	165 (36%)		202 (32%)
Shock	657 (61%)	273 (60%)		384 (62%)	0.66
Renal replacement therapy	96 (9%)	44 (10%)		52 (8%)	0.52
PaO2/FiO2 (mmHg), mean (± SD)	141 (± 80)	143 (± 84)		140 (± 77)	0.64
PaCO2 (mmHg), mean (± SD)	44 (± 5)	43 (± 4)		44 (± 4)	0.84
Lactate, mean (± SD)	2.4 (± 1)	2.1 (± 2)		2.7 (± 1)	0.69
pH, mean (± SD)	7.33 (± 0.13)	7.33 (± 0.13)		7.32 (± 0.12)	0.22
Bicarbonate (mmol/L), mean (± SD)	23 (± 6)	23 (± 6)		23 (± 6)	0.80

SD: standard deviation, BMI: Body Mass Index, SAPS II: Simplified Acute Physiology Score II, SOFA: Sequential Organ Failure Assessment, ARDS: Acute Respiratory Distress Syndrome, PaO₂/FiO₂: PaO₂/FiO₂ ratio

Ventilatory parameters and ARDS management

Data on ventilatory parameters and ARDS management are displayed in Table 2. Postoperative ARDS received significantly higher Vt (6.6 ml/kg of PBW vs. 6.3, p < 0.01) and driving pressure than medical ARDS, while other ventilatory settings were comparable (Figs. S3-S6). After extubation, postoperative ARDS received more frequent prophylactic noninvasive ventilation compared to medical ARDS (40% vs. 33%, p = 0.045).

Table 2.

ARDS management and outcomes in 1077 ARDS patients

Characteristics	Available data	Total (%) (n = 1077)	Postoperative ARDS (n = 455)	Medical ARDS (n = 622)	P value
Tidal volume (ml/kg of PBW), mean (± SD)	1028	6.4 (± 1.3)	6.6 (± 1.2)	6.3 (± 1.3)	< 0.01
PEEP (cmH2O), mean (± SD)	964	10 (± 4)	10 (± 3)	10 (± 4)	0.55
Plateau pressure (cmH2O), mean (± SD)	744	24 (± 5)	24 (± 5)	24 (± 5)	0.29
Driving pressure (cmH2O), mean (± SD)	727	14 (± 5)	14 (± 5)	13 (± 5)	0.048
Compliance (ml/cmH2O), mean (± SD)	727	33 (± 26)	32 (± 20)	34 (± 22)	0.36
Mechanical power (J/min)	723	23.6 (± 7.9)	24.3 (± 7.7)	23.1 (± 8.1)	0.08
Prone positioning	1077	361 (34%)	155 (34%)	206 (33%)	0.74
NMBA	1077	472 (44%)	214 (47%)	258 (41%)	0.08
Inhaled nitric oxide	1019	84 (8%)	26 (6%)	58 (10%)	0.02
Corticosteroids	1015	758 (75%)	320 (73%)	438 (76%)	0.28
ECCO 2 R	1077	39 (4%)	17 (4%)	22 (4%)	0.86
ECMO	1077	16 (1%)	4 (1%)	12 (2%)	0.21
Tracheostomy	996	83 (8%)	30 (7%)	53 (9%)	0.18
Length of invasive MV, mean (± SD)	1077	13 (± 7)	12 (± 6)	13 (± 7)	0.10
Noninvasive ventilation after extubation	911	331 (36%)	161 (40%)	170 (33%)	0.045
Extubation failure during ICU stay	1077	201 (19%)	108 (24%)	93 (15%)	< 0.01
Hospital-acquired pneumonia	1077	224 (21%)	102 (22%)	122 (20%)	0.26
Pneumothorax	1077	44 (4%)	19 (4%)	25 (4%)	0.89
ICU length of stay, mean (± SD)	1077	14 (± 13)	14 (± 12)	15 (± 13)	0.72
Hospital length of stay, mean (± SD)	1077	23 (± 24)	22 (± 23)	23 (± 22)	0.88
ICU mortality	1077	429 (40%)	149 (33%)	280 (45%)	< 0.001
90-day mortality	1077	473 (44%)	166 (36%)	307 (49%)	< 0.001

ARDS: Acute Respiratory Distress Syndrome, PBW: Predicted Body Weight, SD: standard deviation, PEEP: Positive end-expiratory pressure, NMBA: Neuromuscular Blocking Agent, ECCO₂R: extra corporeal carbon dioxide removal, ECMO: Extracorporeal membrane oxygenation, MV: Mechanical Ventilation, ICU: Intensive Care Unit

Evolution over time of the ventilatory parameters are displayed in Figs. S7-S11. Vt decreased significantly over time in both postoperative (-0.07 ml/kg per year, 95% confidence interval [CI]:-0.08;+0.05, p < 0.001) and medical ARDS (-0.08 ml/kg per year, 95%CI:-0.14;-0.05, p < 0.001). PEEP increased significantly over time in postoperative ARDS patients (+ 0.10cmH₂O per year, 95%CI:+0.02;+0.19, p = 0.02), but not in medical ARDS (+ 0.05cmH₂O per year, 95%CI: -0.02;+0.14, p = 0.17).

Early trajectories

Early trajectory at day 3 of ARDS onset was assessed in 902 of the 1,077 included patients (84%, Table S2). At day 3, 23% of postoperative ARDS were classified as Rapidly Improving (89/387), 51% as mild ARDS (197/387), 15% as moderate ARDS (60/387), 4% as severe ARDS (15/387), and 7% had died (26/387). At day 3, 16% of medical ARDS were classified as Rapidly Improving (82/515), 47% as mild ARDS (241/515), 19% as moderate ARDS (97/515), 5% as severe ARDS (28/515), and 13% had died at day 3 (67/515). Postoperative ARDS showed a more favorable early trajectory compared to medical ARDS (Cochran-Armitage test: p < 0.001). Sankey plots are displayed in Fig. 2.

Fig. 2.

Early outcome in postoperative and medical ARDS. Rapidly improving ARDS: resolved hypoxemia or successful weaning from invasive mechanical ventilation with unassisted breathing

90-day mortality

The overall 90-day mortality was 44% (473/1077): 36% (166/455) in postoperative ARDS, compared to 49% (307/622, p < 0.001) in medical ARDS. Kaplan-Meier curves showed that 90-day mortality was significantly lower in postoperative ARDS, compared to medical ARDS (Fig. 3A, p < 0.001 by the log-rank test). In the univariable Cox-proportional hazard model, postoperative ARDS was significantly associated with lower 90-day mortality (unadjusted Hazard Ratio [HR] of 0.66, 95%CI:0.54–0.79, p < 0.001). In the multivariable Cox-proportional hazard model, postoperative ARDS was associated with lower 90-day mortality after adjustment for age, SOFA score, and ARDS severity (adjusted HR [aHR] = 0.68, 95%CI:0.56–0.83, p < 0.001, Table S3). Figure 3B shows crude 90-mortality according to the severity of ARDS in postoperative and medical ARDS.

Fig. 3.

90-day mortality and temporal trends in postoperative and medical ARDS. Panel A shows cumulative 90-day mortality in postoperative and medical ARDS, with numbers at risk displayed below the curves. Panel B displays unadjusted 90-day mortality stratified by ARDS severity at onset (mild, moderate, and severe) for postoperative and medical ARDS.Panel C depicts adjusted temporal trends in 90-day mortality from 2003 to 2023, estimated using a multivariable regression model adjusted on age, SOFA score, and PaO2/FiO2 ratio. Adjusted estimates are shown with 95% confidence intervals, while the overall trend (dashed line) represents the weighted average. P-values for interaction and for temporal trends are indicated

We performed a first sensitivity analysis on all 1,331 ARDS patients, including 254 patients fulfilling the New Global Definition criteria who were not intubated (Fig. S12). 90-day mortality was 39% (524/1331). Postoperative ARDS was associated with lower 90-day mortality (aHR = 0.57, 95%CI:0.47–0.69, p < 0.001, Table S4-S5). We performed a second sensitivity analysis on 906 persistent ARDS, excluding 171 Rapidly Improving ARDS (Fig. S13). 90-day mortality was 47% (423/906). Postoperative ARDS was associated with lower 90-day mortality (aHR = 0.71, 95%CI:0.57–0.87, p < 0.001, Table S6-S7). In a post-hoc sensitivity analysis using a 7-day window to classify postoperative ARDS, the association between postoperative ARDS and lower 90-day mortality remained unchanged (aHR = 0.67, 95%CI:0.54–0.82, p < 0.001; Fig. S14).

Evolution over time

Over the study period, 90-day mortality decreased significantly in postoperative ARDS (− 1.1% per year, 95%CI: −1.9% to − 0.2%; p = 0.03, Fig. 3C) but not in medical ARDS (+ 0.2% per year, 95%CI: −0.7% to + 1.6%; p = 0.34). The temporal trend differed significantly between postoperative and medical ARDS (p-for-interaction = 0.002).

SAPS II increased significantly over time in postoperative ARDS (Fig. S15, + 0.5 point per year, 95%CI: +0.3 to + 0.9, p = 0.004), while it remained stable in medical ARDS (+ 0.3 point per year, 95%CI: -0.1 to + 0.6, p = 0.13, p-for-interaction = 0.059). There was no significant change in SOFA score over time in postoperative ARDS (Fig. S16, + 0.1 point per year, 95%CI: -0.9 to + 1.1, p = 0.83) and in medical ARDS (-0.3 point per year, 95%CI: -1.4 to + 1.2, p = 0.71, p-for-interaction = 0.19). There was no significant change in PaO₂/FiO₂ ratio over time in postoperative ARDS (Fig. S17, -1 mmHg per year, 95%CI: -4 to + 2, p = 0.79) and in medical ARDS (+ 1 mmHg per year, 95%CI: -1 to + 3, p = 0.11, p-for-interaction = 0.41).

We performed sensitivity analyses on all 1,331 ARDS patients (Fig. S18-S21), and on 906 persistent ARDS (Fig. S22-S25), which confirmed the results of the primary analysis.

Independent risk factors for 90-day mortality

Figure 4 presents multivariable Cox regression analyses for 90-day mortality in postoperative and medical ARDS. In postoperative ARDS, increasing age (aHR = 1.29 per 10 years, 95%CI:1.13–1.47; p < 0.001) and chronic respiratory failure (aHR = 1.65, 95%CI:1.12–2.43; p = 0.01) were associated with increased mortality. Markers of extrapulmonary organ failure were independently associated with mortality, including non-respiratory SOFA score (aHR = 1.10 per point, 95%CI:1.05–1.15; p < 0.001) and bicarbonate (aHR = 0.81 per 5 mmol/L increase, 95%CI:0.69–0.96; p = 0.01). Esophageal surgery (aHR = 0.41, 95%CI:0.24–0.70; p = 0.001) and upper abdominal surgery (aHR = 0.64, 95%CI:0.46–0.91; p = 0.01) were associated with lower mortality compared with lower abdominal surgery. No respiratory variable was independently associated with mortality.

Fig. 4.

Independent risk factors for 90-day mortality in postoperative and medical ARDS. Forest plots display adjusted hazard ratios (aHR) with 95% confidence intervals derived from multivariable Cox proportional hazards models, presented separately for postoperative ARDS (Panel A) and medical ARDS (Panel B). Values on the right indicate aHRs with 95% confidence intervals and corresponding p-values

In medical ARDS, increasing age (aHR = 1.20 per 10 years, 95%CI:1.10–1.31; p < 0.001), cancer (aHR = 1.92, 95%CI:1.46–2.52; p < 0.001), and cirrhosis (aHR = 1.45, 95%CI:1.10–1.92; p = 0.009) were independently associated with mortality. Markers of extrapulmonary organ failure were independently associated with mortality, with non-respiratory SOFA score (aHR = 1.05 per point, 95%CI:1.02–1.09; p < 0.001) and bicarbonate (aHR = 0.76 per 5 mmol/L increase, 95%CI:0.67–0.86; p < 0.001). Markers of respiratory failure were independently associated with mortality, including PaO₂/FiO₂ at onset (aHR = 0.88 per 50 mmHg increase, 95%CI:0.81–0.96; p = 0.003), and driving pressure (aHR = 1.13 per 5 cmH₂O increase, 95%CI:1.01–1.27; p = 0.03).

Discussion

In this cohort of 1,077 ARDS patients over 20 years, postoperative ARDS differed meaningfully from medical ARDS in its baseline characteristics, early clinical trajectory, outcomes, and prognostic determinants. Patients developing ARDS in the postoperative setting exhibited a more favorable early evolution marked by a higher proportion of rapidly improving forms, a lower 90-day mortality, and distinct risk factors compared with medical ARDS. Mortality in postoperative ARDS was predominantly associated with markers of extrapulmonary organ dysfunction and surgical context, whereas indices of respiratory failure were independently associated with mortality only in medical ARDS. Collectively, these results suggest that postoperative ARDS should not be viewed as a variant of medical ARDS, but as a distinct subtype with specific pathophysiological drivers and prognostic implications.

Up to 4% of patients undergoing major surgery develop postoperative ARDS [27, 28], and yet very few studies have assessed outcomes and management of these patients. In our cohort, 0.54% of all patients undergoing abdominal surgery (455/83,565) developed ARDS. Differences between postoperative and medical ARDS likely reflect distinct pathophysiological pathways and a unique temporal context. In the postoperative setting, lung injury develops around surgical stress, anesthesia-related physiological perturbations, fluid shifts, transfusion exposure, and procedure-specific constraints, rather than from a sustained primary insult [9]. This constellation may preferentially produce a predominantly indirect form of lung injury [29], driven by transient alveolar instability, atelectasis, or interstitial edema more than by established diffuse alveolar damage [30]. Such a profile is consistent with the more favorable early trajectories observed, including a higher proportion of rapidly improving ARDS. Postoperative ARDS also differs conceptually in that patients are already under close medical care before lung injury develops, creating a concrete window to prevent perioperative insults, both surgical and ventilatory [31]. This emphasis on prevention and early management may reduce the incidence of ARDS, but also mitigate its severity and favor earlier resolution when ARDS nevertheless occurs. Contrary to previous findings, we observed no significant difference in ICU length of stay between groups [7]. This suggests that the duration of mechanical ventilation and hospitalization may be primarily driven by the ARDS itself rather than the surgical context.

Beyond differences in early trajectory, the relevant prognostic factors in postoperative ARDS further support the central role of extrapulmonary and surgery-related drivers. In this group, mortality was independently associated with markers of non-respiratory organ dysfunction, whereas no index of respiratory failure retained prognostic significance. This dissociation suggests that, once postoperative ARDS has developed, outcome is less driven by the severity of lung injury itself than by the presence and evolution of underlying surgical complications, such as anastomotic leakage, intra-abdominal sepsis, bleeding, or ischemic events [9]. From a clinical standpoint, these findings argue for a shift in priorities: in postoperative ARDS, the early diagnostic effort should focus on identifying potentially reversible surgical complications, and timely reoperation should be considered a central component of management. Optimal respiratory management remains necessary but is unlikely to alter prognosis unless the underlying surgical source is rapidly controlled. Unexpectedly, lower abdominal surgery was independently associated with higher mortality compared with upper abdominal and esophageal procedures. This suggests that determinants of mortality in ARDS may not fully overlap with determinants of ARDS development.

The decrease in 90-day mortality observed over 20 years in postoperative ARDS patients is a new finding, especially considering that the SAPS II increased simultaneously. The annual reduction of 1.1% in 90-day mortality is clinically relevant and could be attributed to several factors. Firstly, the widespread adoption of minimally invasive surgical techniques may have played a role in reducing the incidence and severity of postoperative complications, including ARDS [32]. Minimally invasive surgery has been associated with reduced trauma, less postoperative pain, and faster recovery [33], which could decrease the likelihood of severe inflammatory responses leading to ARDS [34]. Secondly, improvements in perioperative respiratory care, with lung-protective ventilation during surgery [35–37] and early postoperative noninvasive ventilation [11, 38, 39], may have mitigated the risk and the severity of postoperative ARDS. However, the main contributing factor highlighted in our study is the progressive improvement in lung-protective ventilation after ARDS onset. Vt was set higher in postoperative ARDS than in medical ARDS in the first years of the study. Vt has decreased linearly over 20 years in all ARDS patients, and the difference of set Vt between postoperative ARDS and medical ARDS has disappeared over time. Concurrently, set PEEP has increased significantly over time in postoperative ARDS in our cohort. Increasing the PEEP might be paramount in managing ARDS after abdominal surgery, as intra-abdominal pressure is often increased, thus increasing pleural pressure and reducing end-expiratory lung volume [40]. Interestingly, the utilization of prone positioning was similar between postoperative and medical ARDS patients.

This study has several strengths. It reports on a large, consecutive cohort of ARDS patients collected over two decades within a single center, ensuring both temporal depth and consistency in diagnosis and management. ARDS diagnosis was reassessed using the New Global Definition, allowing a contemporary, standardized classification across the entire study period. Sensitivity analyses were consistent across different severity and prognosis settings. However, several limitations should be acknowledged. The monocentric design, conducted in a tertiary ICU specialized in abdominal surgery, may limit generalizability to other surgical populations or non-expert centers. The diagnosis of ARDS was retrospectively assessed. However, ARDS was diagnosed with the most up-to-date definition by two independent physicians. Although the time-window for definition of postoperative ARDS may be debated, it was based on individual patient-level evidence and provides a standardized and reproducible framework. Thirdly, changes in surgical techniques, perioperative management, and anesthesia protocols were not specifically assessed. Thus, we cannot speculate on which advancements have most contributed to the improved outcome of postoperative ARDS. Causal relationships cannot be inferred, and the associations should be interpreted as hypothesis-generating. Finally, the definition of postoperative ARDS using a broad 15-day window may be debated. From a clinical perspective, postoperative ARDS may result not only from early perioperative factors, but also from secondary surgery-related complications such as anastomotic leakage or intra-abdominal sepsis, which may occur beyond the immediate postoperative period [41]. In our cohort, 92% of cases (419/455) occurred within 7 days after surgery, and our findings were robust when restricting to this time window.

Conclusions

In conclusion, postoperative ARDS differs from medical ARDS in its early clinical trajectory, outcomes, and prognostic determinants. These differences support the concept of postoperative ARDS as a distinct clinical subtype, in which outcomes are primarily driven by extrapulmonary and surgery-related factors rather than by the severity of lung injury itself. Recognizing its specificity has important clinical implications, underscoring the need to prioritize prevention of perioperative insults, early identification of surgical complications, and timely source control, including early reoperation, alongside optimal respiratory support.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ARDS

Acute Respiratory Distress Syndrome

BMI

Body Mass Index

CI

Confidence Interval

Crs

Compliance of the Respiratory System

FiO₂

Fraction of Inspired Oxygen

HR

Hazard Ratio

ICU

Intensive Care Unit

PBW

Predicted Body Weight

PEEP

Positive End-Expiratory Pressure

SAPS II

Simplified Acute Physiology Score II

SOFA

Sequential Organ Failure Assessment

Vt

Tidal Volume

Author contributions

JP, JH, NM, GC, ADJ and SJ designed the study. JP, JH, YA, CM, IL, MC, and AV collected the data. JP, JH, ADJ, and NM analyzed the data. All authors were involved in the data interpretation. JP, JH, JC, ADJ and SJ wrote the manuscript. All authors read and approved the manuscript.

Funding

The funding source was the Montpellier University Hospital (France). It had no role in the conception, design, or conduct of the trial, nor did their representatives participate in the collection, management, analysis, interpretation, or presentation of the data or in the preparation, review, or approval of the manuscript.

Data availability

Research data and other material (e.g., study protocol and statistical analysis plan) will be made available to the scientific community, immediately on publication, with as few restrictions as possible. All requests should be submitted to the corresponding author who will review with the other investigators for consideration. A data use agreement will be required before the release of participant data and institutional review board approval as appropriate.

Declarations

Ethics approval and consent to participate

We obtained approval from the Montpellier University Hospital ethics committee for the Forever project on May 29, 2019 (Comité Local d’Ethique Recherche, agreement number: 2019_IRB-MTP_05–25). The requirement for informed consent was waived, and the opportunity to opt out of the study was provided. The study has been performed in accordance with the Helsinki declaration of 1975.

Consent for publication

Not applicable.

Competing interests

JP received funding from the Société Française d’Anesthésie-Réanimation, the Philippe Foundation, the Beth Israel Deaconess Medical Center (Boston, MA), and the University Hospital of Montpellier. SJ received consulting fees from Drager, Medtronic, Mindray, Fresenius, Baxter, and Fisher & Paykel. ADJ reports receiving consulting fees from Sanofi, Sedana, Medtronic and Viatris. No potential conflict of interest relevant to this article was reported for other authors.