Ventilation Intensity and Prognosis in Pediatric Oncology: Mechanical Power, Compliance, and Mortality Risk

ABSTRACT

Background

Mechanical power (MP) and dynamic compliance (Cdyn) are key determinants of ventilator‐induced lung injury (VILI) and mortality. This study aimed to identify variables associated with MP and Cdyn and evaluate their impact on mortality and ventilator‐free days (VFD) in critically ill children with cancer.

Methods

We conducted a prospective, observational study of 57 patients. Multivariate logistic regression and competing risk models were applied to assess predictors of mortality and VFD.

Results

The median age was 69 months, with a median of three organ dysfunctions. Mechanical power normalized by dynamic compliance (MP/Cdyn) emerged as an independent predictor of mortality across multivariable models: MP/Cdyn was significantly associated with death alongside the number of organ dysfunctions (OR: 5.44, 95% CI: 1.78–18.6, p = 0.003; OR: 1.62, 95% CI: 1.07–2.47, p = 0.023), platelet count <30,000/μL (OR: 4.23, 95% CI: 1.36–13.1, p = 0.013; OR: 4.92, 95% CI: 1.31–18.4, p = 0.018), and renal and immunologic dysfunctions. ROC analysis of MP/Cdyn for ICU mortality yielded an AUC of 0.76, with a cutoff of 1.11 J·cmH₂O/min·mL (sensitivity 59%, specificity 89%). Kaplan‐Meier analysis showed significantly lower survival for MP/Cdyn >1.11 (p = 0.016). Higher MP was also associated with fewer 28‐day VFD (SHR: 0.17, 95% CI: 0.04–0.8, p = 0.026). VFD also correlated with Cdyn/PBW, PCO₂, B‐lines on lung ultrasound, and oxygen saturation index.

Conclusions

In critically ill children with cancer, elevated mechanical power is strongly linked to reduced ICU survival and fewer VFD. These findings highlight the prognostic importance of ventilation strategies in this vulnerable population.

1. Introduction

Mechanical power (MP) was introduced as an integrative measure of the energy delivered by the ventilator per unit of time, combining tidal volume, airway pressures, respiratory rate, inspiratory flow, and positive end‐expiratory pressure (PEEP) into a single variable [1]. Experimental and clinical studies have shown that higher levels of mechanical power are associated with increased lung injury and worse outcomes, supporting the concept that ventilator‐induced lung injury (VILI) is driven by the energy load applied to the lung. Notably, different combinations of ventilator settings can yield similar mechanical power values and produce comparable degrees of injury, underscoring the importance of this integrated metric over isolated analysis of ventilation parameters [2].

In pediatric populations, applying the concept of mechanical power presents unique challenges due to age‐related differences in respiratory mechanics, lung size, and ventilation strategies. In children with pediatric acute respiratory distress syndrome (PARDS), higher mechanical power values normalized to predicted body weight (PBW) have been linked to fewer ventilator‐free days and longer durations of mechanical ventilation, particularly in younger patients [3]. These findings suggest that mechanical power captures clinically relevant aspects of ventilator‐related risk not fully reflected by traditional lung‐protective metrics. However, the same absolute energy load may have different effects depending on an individual lung's susceptibility to injury. To address this, normalization of mechanical power to physiologically relevant variables has been proposed to better represent the energy density applied to the lung. Mechanical power normalized to pulmonary compliance has demonstrated improved ability to discriminate mortality risk, suggesting that compliance may serve as an indirect marker of functional lung size available to dissipate applied energy [4, 5].

Children with cancer who require mechanical ventilation represent a particularly vulnerable group, often exposed to systemic inflammation and fluid overload, factors that further reduce functional lung size. Evidence on the relationship between compliance, mechanical power, and outcomes in this population remains limited. Therefore, we sought to identify, in critically ill children with cancer on mechanical ventilation, variables associated with mechanical power and dynamic compliance (Cdyn) that may correlate with mortality and ventilator‐free days (VFD), based on a single measurement.

2. Methods

We conducted a prospective, observational cohort study in the oncology intensive care unit of a referral center for children and adolescents with cancer in São Paulo, Brazil. The study protocol was approved by the Research Ethics Committee of the Federal University of Sao Paulo (Certificate of Submission of Ethical Appraisal no. 68803323.9.0000.5505; approval date: July 19, 2023). All procedures complied with Brazilian ethical regulations (Resolution 196/96 of the National Health Council). The study included patients admitted to the ICU between July 2023 and March 2025. No formal sample size calculations were made, but we estimated that we could include at least 50 patients over an approximate 2‐year period, depending on ICU occupancy.

Eligible participants were younger than 18 years, had a diagnosis of oncological disease, were critically ill requiring mechanical ventilation, had a central venous catheter for blood collection, and whose legal guardians provided written informed consent. All patients underwent transthoracic echocardiography and lung ultrasound. Lung ultrasound consisted of 12 B‐line measurements: six on the right side (3rd, 4th, and 5th intercostal spaces in the midaxillary and anterior lines) and six on the left side at corresponding sites. In addition, central venous blood gas analysis was performed in all patients. Previous research in adults has identified a single measurement of mechanical power as a reliable prognostic marker for mortality [6]. Consequently, our study utilized a single‐point measurement rather than tracking data over a specific ventilation duration, but we chose not to include patients on prolonged ventilation. Since there is no single definition of prolonged ventilation, we have established a maximum of 14 days for inclusion. Data collection occurred immediately following the acquisition of informed consent.

Dynamic lung compliance (Cdyn) was calculated as: Cdyn (mL/cmH₂O) = VT/(PIP–PEEP), where VT = tidal volume, PIP = peak inspiratory pressure and PEEP = end‐expiration positive pressure.

Mechanical power (MP) was calculated using the simplified equation proposed by Gattinoni et al. [1] for volume‐controlled ventilation (with PIP replacing plateau pressure in the equation) [3], and the equation proposed by Becher et al. [7] for pressure‐controlled (PCV) and pressure‐regulated volume‐controlled (PRVC) ventilations:

In volume‐controlled (VCV): MP (Joules/min) = 0.098 × VT × RR × (PIP – 0.5 × Delta pressure), where RR = respiratory rate set on the ventilator, and Delta pressure = PIP – PEEP;

In PCV and PRVC: MP (J/min) = 0.098 × VT × RR × (PEEP + Delta pressure).

Predicted weight was used to normalize the Cdyn and VT variables and was calculated using the equations proposed by Braillon [8]. Mechanical power was normalized by dynamic compliance (MP/Cdyn). To define organ dysfunction, we used the PODIUM (Pediatric Organ Dysfunction Information Update Mandate) criteria [9].

2.1. Statistical Analysis

Descriptive statistics were reported as measures of central tendency and dispersion for continuous variables, and as absolute and relative frequencies for categorical variables. Comparisons between survivors and non‐survivors were performed using the Mann–Whitney test for continuous variables and the chi‐square test for categorical variables. Correlations were assessed using Spearman's rank correlation coefficient.

For the outcome of ICU mortality, bivariate and multivariable logistic regression models were constructed, using stepwise selection with bidirectional (forward/backward) elimination, choosing the models based on the lowest values of the Akaike Information Criterion (AIC). The Hosmer–Lemeshow test was used to assess the goodness‐of‐fit. Variables with a p‐value < 0.1 in bivariate analyses were subsequently included in multivariable logistic regression models.

To evaluate associations between 28‐day ventilator‐free days (considering only invasive ventilation) and variables such as compliance, mechanical power, and potential confounders, bivariate and multivariable competing risk regression models were employed, also using stepwise selection with bidirectional (forward/backward) elimination. In these models, the risk of being alive and extubated was analyzed with death as a competing event, and ventilation time was censored at 28 days following ICU admission. This time‐to‐event approach yielded sub‐distribution hazard ratios (SHR), the magnitude of which reflects both the time to extubation and the probability of being alive and successfully extubated at 28 days [10]. Children who died in the ICU were assigned zero VFD. Collinearity among variables was assessed using the variance inflation factor. Receiver operating characteristic (ROC) curve analysis was performed to identify a potential cutoff point (determined by the Youden index) for mechanical power values associated with ICU mortality. A Kaplan–Meier survival curve was generated to estimate survival according to this cutoff.

Statistical analyses were performed using the software jamovi version 2.6 (The Jamovi project, 2025, https://www.jamovi.org) and R (R: A language and environment for statistical computing, https://www.R-project.org). The R package “cmprsk” was used for modeling competing risks, by the Fine–Gray model.

3. Results

We analyzed 57 patients on mechanical ventilation. The median number of days on ventilation at the time of assessment was 2 days (IQR: 0–4). Nine (15.7%) had been on ventilation for more than 6 days, with a maximum of 14 days (one patient). The ICU mortality rate in this cohort was 38.5% (N = 22). 51 patients (89.5%) had two or more organ dysfunctions at the time of assessment. Demographic data and general patient characteristics are shown in Table 1. Eleven patients (19.3%) experienced pulmonary hemorrhage associated with severe thrombocytopenia (<30,000 platelets/μL), evidenced by the presence of blood in the orotracheal tube. Of these, nine patients (81.8%) died. Four cases were clinically consistent with post–hematopoietic stem cell transplantation (HSCT) diffuse alveolar hemorrhage; however, bronchoscopy could not be performed to confirm the diagnosis due to respiratory instability.

Table 1.

Baseline characteristics of patients, including cancer diagnosis.

N = 57
Age (Months, median, IQR)	69	33.5–91
Weight (kg, mean, SD)	21.6	16.7
Predicted weight (kg, mean, SD)	20.3	12
Female sex (N, %)	30	52.6%
Oncological diagnosis (N, %)
Medulloblastoma	9	15.8%
Other nervous system tumors (gliomas, craniopharyngioma, teratoid/rhabdoid, choroid plexus carcinoma)	21	36.8%
Retinoblastoma	3	5.3%
Acute lymphocytic leukemia	3	5.3%
Acute myeloid leukemia	5	8.8%
Burkitt's leukemia	1	1.8%
Non‐Hodgkin's lymphoma	2	3.5%
Neuroblastoma	3	5.2%
Ewing sarcoma	2	3.5%
Other (hepatoblastoma, osteosarcoma, rhabdomyosarcoma, fibrosarcoma, hemangioendothelioma)	8	14%
Hematopoietic Stem Cell Transplantation (HSCT)	9	15.8%
Veno‐occlusive disease post‐HSCT	5	8.7%
Vasoactive drugs (N, %)	16	28%
Reason for admission (N, %)
Septic shock	15	26.3%
Hypovolemic shock	1	1.8%
Seizures	4	7.0%
Post‐operative	12	21.1%
Hyperleukocytosis	2	3.5%
Sepsis without shock	2	3.5%
Decreased level of consciousness	3	5.3%
Respiratory failure	9	15.8%
PARDS	7	12.3%
Pulmonary COVID‐19	1	1.8%
Extrinsic airway compression	1	1.8%
Veno‐occlusive disease	2	3.5%
Acute renal failure requiring continuous renal replacement therapy	1	1.8%
Intracranial hypertension	2	3.5%
Acute abdomen	1	1.8%
Ventriculitis	1	1.8%
Acute chemotherapy complications	2	3.5%
Causes of respiratory failure (N, %)
PARDS	23	40.4%
Status epilepticus	5	8.8%
Aspiration pneumonia	3	5.3%
Invasive pulmonary fungal infections	2	3.5%
Neurological impairment with decreased consciousness or central apnea	12	21.1%
Post‐cardiac arrest/shock	4	7%
Chest restriction due to abdominal mass	1	1.8%
Extrinsic airway compression	1	1.8%
Bronchospasm during anesthesia induction	1	1.8%
Shock with hemodynamic instability	4	7%
Pulmonary COVID‐19		1	1.8%
PODIUM organ dysfunction criteria:
Renal	11	19.3%
Hematologic	24	42.1%
Neurologic	15	26.3%
Hepatic	11	19.3%
Immune	34	59.6%
Endocrine	6	10.5%
Coagulation	14	24.6%
Cardiovascular	24	42.1%
Respiratory	57	100.0%
Number of organ dysfunctions (median, IQR)	3	2‐5
Number of organ dysfunctions (excluding hematologic and immune, median, IQR)	2	2‐3
Mild‐moderate PARDS (PALICC‐2) [11]	17	29.8%
Severe PARDS	6	10.5%
PICU length of stay (median, IQR)	25	51 ‐ 13
Ventilation mode (N, %)
PRVC (Pressure‐Regulated Volume Control)	7	12.3%
VCV (Volume‐controlled)	4	7.0%
PCV (Pressure‐controlled)	46	80.7%

Abbreviations: IQR, interquartile range; PALICC‐2, Second Pediatric Acute Lung Injury Consensus Conference; PARDS, Pediatric Acute Respiratory Distress Syndrome; PODIUM, Pediatric Organ Dysfunction Information Update Mandate; SD, standard deviation.

We evaluated differences between survivors and non‐survivors in ventilation parameters, oxygen saturation index (OSI), dynamic compliance, mechanical power, blood gas and hematologic values, mean number of B‐lines on pulmonary ultrasound, left ventricular ejection fraction, cardiac index assessed by echocardiography, and other clinical variables. The results are presented in Table 2.

Table 2.

Comparative measurements between survivors and non‐survivors, in means and frequencies.

Means	Survivors	N = 35	Non‐survivors	N = 22
	Mean	SD	Mean	SD	p
PIP (cmH2O)	20.5	5.18	26.3	6.94	0.001
PEEP (cmH2O)	7.66	1.89	9.23	2.51	0.013
MAP (cmH2O)	11.8	3.01	15.7	4.52	< 0.001
SpO2 (%)	97.6	2.02	95	3.76	0.007
TV/PBW (mL/kg)	8.25	3.31	7.47	2.05	0.38
FiO2	0.31	0.09	0.54	0.29	0.002
Delta pressure (PIP‐PEEP, cmH2O)	12.8	3.94	17.1	5.6	0.004
Respiratory rate (bpm)	24.2	5.9	26.4	4.44	0.15
OSI	3.89	2	9.78	7.28	< 0.001
Cdyn/PBW (mL/cmH2O. kg)	0.69	0.3	0.49	0.26	0.003
MP/Cdyn (J. cmH20/min. mL)	0.66	0.44	1.34	0.92	0.002
Days of mechanical ventilation	9.53	9.4	21.1	20.4	0.057
Ventilator‐free days	18.6	7.77	0	0	—
Number of organ dysfunctions	3	1.48	4.14	1.7	0.008
Number of organ dysfunctions (excluding hematologic and immune)	2.3	1.07	2.64	1.4	0.4
pH	7.31	0.06	7.33	0.08	0.72
Bicarbonate (mmol/L)	23.6	5.03	25.4	6.7	0.24
PCO2 (mmHg)	46.4	11.7	52.2	10.8	0.025
Central oxygen saturation (%)	73.2	11.2	76.3	14.2	0.3
Lactate (mmol/L)	1.33	1.11	2.27	2.16	0.07
Platelets (cells/μL)	186763	165227	63645	118605	< 0.001
Lymphocytes (cells/μL)	1677	1751	899	2461	0.004
Neutrophils (cells/μL)	7425	7279	3263	782	0.016
Hemoglobin (g/dL)	9.87	1.66	10.2	3.3	0.31
Left ventricular ejection fraction (%)	67.3	9.78	65.1	10.3	0.65
Heart rate (bpm)	112	32.2	123.4	29.2	0.14
Cardiac index (L/min/m2)	3.24	1.07	3.45	1.09	0.58
PRISM IV (Pediatric risk of mortality) raw score	11.6	8.14	13.2	8.20	0.49
Fluid overload percentage (24 h)	0.46%	3.94	‐0.74%	3.77	0.08
Mean number of B‐lines per intercostal space (lung ultrasound)	1.12	0.99	2.05	1.24	0.004

	Survivors	N = 35	Non‐survivors	N = 22
Frequencies	N	%	N	%	p
Platelet count < 30000 cells/μL (N, %)	6	17.1%	12	33.3%	0.001
Lymphocyte count < 1000 cells/μL (N, %)	15	42.8%	19	86.4%	< 0.001
Neutrophil count < 500 cells/μL (N, %)	6	17.1%	11	50%	0.015
PODIUM Renal	2	5.7%	9	40.9%	0.003
Hematologic	10	28.6%	14	63.6%	0.02
Neurologic	12	34.3%	3	13.6%	0.15
Hepatic	6	17.1%	5	22.7%	0.86
Immune	15	42.9%	19	86.4%	0.003
Endocrine	3	8.6%	3	13.6%	0.87
Coagulation	8	22.9%	6	27.3%	0.95
Cardiovascular	14	40.0%	10	45.5%	0.89
Diagnosis of leukemia or lymphoma	4	11.4%	6	27.3%	0.24
Diagnosis of central nervous system tumor	22	62.9%	6	27.3%	0.02
Hematopoietic Stem Cell Transplantation	3	8.6%	6	27.3%	0.13

Note: The p‐values were determined by the Mann–Whitney U test or chi‐square test with Yates’ continuity correction, as appropriate.

Abbreviations: Cdyn/PBW, Dynamic compliance normalized to predicted body weight; J, Joule; MAP, Mean Airway Pressure; MP/Cdyn, Mechanical power normalized to dynamic compliance OSI, Oxygen Saturation Index; PEEP, Positive End‐Expiratory Pressure; PIP, Peak Inspiratory Pressure; TV/PBW, Tidal volume, normalized to predicted body weight.

The variables listed in Table 2, along with organ dysfunctions, were evaluated in bivariate and multivariable logistic regression models for the outcome of ICU mortality, and in competing risk regression models for ventilator‐free days. One of the key assumptions in logistic regression is maintaining an adequate observation‐to‐predictor ratio; a common rule of thumb is to have at least 10 to 15 observations per independent variable. Given that our small sample included only 22 events (deaths), we limited our analysis to simpler models with two independent variables (11 observations per variable) to mitigate the risk of overfitting. Because our primary objective was to evaluate the impact of MP/Cdyn, we sought to identify secondary variables that served as independent predictors alongside MP/Cdyn within these two‐variable models. The final models presented were selected based on their statistical significance, optimal goodness‐of‐fit (evaluated via Akaike's Information Criterion and the Hosmer–Lemeshow test), and clinical relevance.

For ICU mortality, ventilation pressures, dynamic compliance, mechanical power, the number of B‐lines, oxygen saturation index (OSI), and immunological, renal, and hematological dysfunctions, as well as the total number of organ dysfunctions, were significant predictors in bivariate analyses. Mechanical power normalized by compliance (MP/Cdyn) emerged as an independent predictor of death across four multivariable models: (1) in combination with the total number of organ dysfunctions, (2) with platelet count <30000 cells/μL, (3) with renal dysfunction, and (4) with immune dysfunction. These associations remained independent after adjustment for age and other covariates. Table 3 presents the bivariate models with p‐values < 0.1 that were subsequently tested in multivariable analyses. A complete list of tested bivariate models and other multivariate models are in a File S1.

Table 3.

Variables included in bivariate and multivariable models for death in PICU (logistic regression) and bivariate and multivariate competing risk regression models for the outcome “alive and extubated” (controlling for the competing risk of death, with time censored at 28 days of ventilation).

Bivariate models for the outcome of death in the ICU
	Estimate	AIC		Odds ratio (95% CI)	p
PIP (cmH2O)	0.16	68.8		1.17 (1.06–1.29)	0.002
PEEP (cmH2O)	0.33	73.3		1.39 (1.06–1.82)	0.016
MAP (cmH2O)	0.27	66		1.31 (1.11–1.55)	0.002
Delta pressure (PIP‐PEEP, cmH2O)	0.20	69.5		1.22 (1.06–1.40)	0.004
OSI	0.35	60.5		1.42 (1.13–1.78)	0.002
Cdyn/PBW (ml/cmH2O. kg)	‐3.00	73.3		0.05 (0.03–0.76)	0.029
MP/Cdyn (J. cmH20/min. mL)	1.65	67.2		5.2 (1.72–15.7)	0.003
PCO2 (mmHg)	0.045	76.6		1.04 (0.99–1.1)	0.08
Lactate (mmol/L)	0.37	77.5		1.45 (0.99–2.11)	0.05
Number of organ dysfunctions	0.45	73.3		1.57 (1.09–2.26)	0.015
Podium criteria: Immune	2.13	68.7		8.44 (2.1–33.8)	0.003
Podium criteria: Renal	2.43	69.2		11.4 (2.16–60)	0.004
Podium criteria: Hematology	1.48	73.2		4.37 (1.4–13.6)	0.01
Platelet count < 30000 cells/μL	1.94	69.3		6.98 (2.05–23.7)	0.002
Number of organ dysfunctions excluding immunology and hematology	0.24	78.9		1.27 (0.8–1.99)	0.28
Mean number of B‐lines per intercostal space	0.72	71.2		2.07 (1.24–3.45)	0.005
Hematopoietic Stem Cell Transplantation	1.38	76.6		4 (0.88–18)	0.07
Central nervous system tumors	‐1.5	73		−0.22 (0.07–0.7)	0.01

Multivariable models for the outcome of death in the ICU	Estimate	AIC	OR (95% CI)	p	Hosmer p‐value
1. MP/Cdyn	1.70	63.4	5.44 (1.78–16.6)	0.003	0.69
Number of organ dysfunctions	0.49		1.62 (1.07–2.47)	0.023
2. MP/Cdyn	1.44	63.4	4.23 (1.36–13.1)	0.013	0.72
Platelet count < 30,000 cells/μL	1.59		4.92 (1.31–18.4)	0.018
3. MP/Cdyn	1.49	59.5	4.44 (1.53–12.9)	0.006	0.84
Podium criteria: Immune	2.30		10.01 (1.88–53)	0.007
4. MP/Cdyn	1.5	62	4.5 (1.49–13.7)	0.007	0.41
Podium criteria: renal	2.2		9.2 (1.55–54.7)	0.015
5. MP/Cdyn	0.56	61.8	1.74 (0.46–6.58)	0.408	0.23
OSI	0.28		1.33 (1.02–1.74)	0.033

Ventilator‐free days ‐ bivariate models		Estimate		SHR	p
PIP (cmH2O)		−0.12		0.88 (0.8–0.97)	0.013
PEEP (cmH2O)		−0.28		0.75 (0.58–0.99)	0.043
MAP (cmH2O)		−0.14		0.87 (0.76–0.98)	0.025
Delta pressure (cmH2O)		−0.16		0.86 (0.75–0.98)	0.025
OSI		−0.27		0.76 (0.64–0.90)	0.001
Cdyn/PBW (ml/cmH2O. kg)		2.29		9.83 (2.92–33)	<0.001
MP/Cdyn (J. cmH20/min. mL)		‐1.57		0.21 (0.06–0.72)	0.013
PCO2 (mmHg)		‐0.05		0.95 (0.91–0.99)	0.03
Podium criteria: hematology		‐0.85		0.42 (0.16–1.15)	0.07
PARDS (any degree of severity)		‐0.86		0.42 (0.16–1.12)	0.09
Number of B‐lines		‐0.377		0.68 (0.48–0.97)	0.035
Age		0.006		1.01 (0.99–1.01)	0.05

Multivariate models for ventilator‐free days		Estimate		SHR	p
1. Age		0.005		1.005 (0.99–1.01)	0.17
MP/Cdyn		−1.75		0.17 (0.04–0.8)	0.026
PARDS		0.31		1.36 (0.39–4.7)	0.62
2. MP/Cdyn		−1.21		0.29 (0.1–0.9)	0.036
Platelet count < 30000 cells/μL		−1.33		0.26 (0.06–1.04)	0.057
3. MP/Cdyn		−0.74		0.47 (0.13–1.7)	0.26
OSI		−0.18		0.84 (0.70–0.99)	0.05

Note: Hosmer p‐value: p‐value for the Hosmer–Lemeshow test for goodness of fit.

Abbreviations: AIC, Akaike's Information Criterion; Cdyn/PBW, Dynamic compliance normalized to predicted body weight; MAP, Mean Airway Pressure; MP/Cdyn, Mechanical power normalized to dynamic compliance; OSI, Oxygen Saturation Index; PEEP, Positive End‐Expiratory Pressure; PIP, Peak Inspiratory Pressure; SHR, subdistribution hazard ratio.

For VFD, mechanical ventilation pressures were predictive in bivariate models, as were the number of B‐lines, partial pressure of carbon dioxide (PCO₂), and OSI. In multivariable models, higher mechanical power was independently associated with fewer 28‐day VFD after adjustment for age and the presence of PARDS, and also on several other tested models (File S1).

The correlations observed between variables analyzed in the models and VFD are illustrated in Figure 1:

Figure 1.

Correlogram obtained by the Spearman test showing the complex correlations between ventilation intensity (expressed by dynamic compliance and mechanical power), extravascular lung water (represented by the number of B lines), PCO₂ values, oxygenation (OSI) and organic dysfunctions that affect the measurement of VFD. Negative correlations are indicated in shades of red, and positive correlations in shades of blue. Cdyn/PBW, Dynamic compliance normalized to predicted body weight; MP/Cdyn, Mechanical power normalized to dynamic compliance; OSI, Oxygen Saturation Index; VFD, Ventilator‐free days. [Color figure can be viewed at wileyonlinelibrary.com]

An ROC curve analysis of the MP/Cdyn variable for the outcome of ICU mortality demonstrated an area under the curve (AUC) of 0.76 (95% CI: 0.62–0.89). The optimal cutoff point for prediction was MP/Cdyn = 1.11, yielding a sensitivity of 59% and a specificity of 89%. Kaplan–Meier survival analysis further showed that patients with MP/Cdyn values > 1.11 had significantly lower survival, with a log‐rank test p‐value of 0.016 (Figure 2). Seventeen patients had MP/Cdyn > 1.11; of these, 13 died (76.5%). Seven (41.1%) had undergone HSCT, and of these, six died (85.7%).

Figure 2.

Kaplan–Meier curve showing the probabilities of survival at 60 days for patients with MP/Cdyn > 1.1 (dashed red line) or not (blue line). [Color figure can be viewed at wileyonlinelibrary.com]

4. Discussion

In this sample of severely ill children with cancer and predominantly multiple organ dysfunctions, mechanical power normalized by dynamic compliance demonstrated strong discriminatory ability for ICU survival. A cutoff point of 1.11 J·cmH₂O/min·mL was able to stratify survival with reasonable performance. Identifying variables associated with mortality is essential to guide strategies aimed at reducing risk, particularly those that may be modifiable. In this study, all ventilator pressures were predictors of death in bivariate models; however, MP/Cdyn consistently remained a robust, independent predictor of mortality across multiple multivariate models; one exception occurred when it was paired with OSI, which emerged as the sole significant predictor in that model. OSI acted as a dominant immediate predictor and statistically overshadowed mechanical power. This phenomenon is likely due to the immediate, life‐threatening physiological weight of severe hypoxemia, as well as the mathematical structure of OSI itself. Because the OSI formula inherently incorporates mean airway pressure alongside gas‐exchange impairment, it serves as a composite metric of both mechanical stress and parenchymal failure. Consequently, when modeled concurrently, the acute threat of tissue hypoxia captured by OSI carries a higher immediate prognostic weight, absorbing the statistical variance of the driving mechanical stress that precipitates it. The absolute number of organ dysfunctions also emerged as an independent predictor, as expected, as did the criteria for immunological and renal dysfunction.

Platelet count was another independent predictor of mortality, alongside MP/Cdyn. This finding is clinically plausible, as thrombocytopenia constitutes a significant obstacle in the ventilation of pediatric oncology patients. It arises from bone marrow suppression induced by chemotherapy or radiation therapy, direct marrow infiltration by hematologic malignancies such as leukemia or lymphoma, or infectious processes. To mitigate and control bleeding, frequent platelet transfusions are often required. Moreover, thrombocytopenia may predispose to diffuse alveolar hemorrhage and can be a feature of thrombotic microangiopathy in patients undergoing hematopoietic stem cell transplantation [12]. Experimental studies have shown that mechanical power correlates with markers of ventilator‐induced lung injury (VILI), including alveolar rupture and inflammation [13]. The energy load delivered by the ventilator (ergotrauma), beyond overstretching (volutrauma) or pressure (barotrauma), can precipitate intra‐alveolar hemorrhage [14]. The association of high MP/Cdyn with thrombocytopenia therefore plausibly contributes to increased mortality. Moreover, platelets are increasingly recognized as mediators of inflammation; stored platelets undergo functional changes (platelet storage lesions) that promote the release of cytokines such as IL‐1, IL‐6, and RANTES, potentially exacerbating transfusion reactions and systemic inflammation [15].

Immune dysfunction was highly prevalent in this cohort (56%), with profound neutropenia in 50% of non‐survivors and lymphopenia in 86%. Neutropenia is associated with the development of acute respiratory distress syndrome and poor prognosis, while neutrophil recovery may worsen pre‐existing lung injury [16, 17]. Lymphopenia is a recognized risk factor for multiple organ dysfunction and mortality, particularly in septic patients [18]. These findings underscore the heightened vulnerability of this population to infections—particularly among individuals with cancers that compromise immune function (such as hematological malignancies, including leukemias and lymphomas) and those undergoing intensive chemotherapy or radiotherapy, where cumulative effects lead to profound short‐term suppression and lasting alterations in immune response. Renal dysfunction also emerged as an independent predictor when combined with MP/Cdyn, consistent with the frequent association of respiratory failure and fluid overload in critically ill children with cancer. Mechanical ventilation itself is an independent predictor of mortality in patients requiring continuous renal replacement therapy [19].

Normalization of mechanical power by dynamic compliance was chosen in this study. Prior work by Khorasanee et al. demonstrated that this approach provided modest but statistically significant superiority in discriminating ICU mortality compared with normalization by predicted body weight [5]. In our cohort, MP/Cdyn showed acceptable predictive performance in ROC analysis, with a cutoff point capable of defining survival probabilities. Conceptually, MP/Cdyn reflects the delicate balance between delivering sufficient energy to achieve oxygenation and minimizing injury in lungs with variable distensibility. Higher MP/Cdyn values indicate that greater resistance and energy load must be overcome to achieve adequate ventilation and oxygenation [20].

High MP/Cdyn was also associated with fewer 28‐day ventilator‐free days. VFD is a widely used measure of organ failure‐free outcome in intensive care, accounting for the competing risk of death [10]. Our sample size did not permit subgroup analysis by age, and age was not an independent predictor of VFD. In our cohort, VFD correlated moderately with PCO₂ and compliance normalized by predicted weight, and strongly with OSI and the number of pulmonary B‐lines. The number and morphology of B‐lines reflect interstitial fluid and loss of aeration, with lower scores associated with successful extubation [21, 22]. We observed a strong correlation between B‐lines and MP/Cdyn, underscoring the relationship between extravascular lung water and the increased energy required to ventilate poorly aerated lungs. When evaluated alongside OSI, MP/Cdyn did not prove to be an independent predictor of ventilator‐free days. Instead, OSI remained significantly associated with fewer ventilator‐free days, with its hazard ratio confidence interval approaching but not crossing the upper boundary of significance (SHR range: 0.70–0.99). This finding further demonstrates the dominant prognostic power of OSI for this outcome as well.

The main limitation of this study is its small sample size, which restricted the ability to conduct meaningful subgroup analyses based on oncological diagnosis or age. In pediatric oncology patients, critical illnesses can manifest in various phenotypes and prognoses, influenced by the underlying disease and additional risk factors [23]. The small number of HSCT patients also did not allow for adequate assessment of complications such as veno‐occlusive disease and thrombotic microangiopathy. As a single‐center investigation, the findings may have limited generalizability. Moreover, assessments were conducted at a single time point, which in some cases did not align with peak disease severity. Although most patients were evaluated within 48 h of ventilation initiation, heterogeneity in timing was present. As with any regression analysis, residual confounding cannot be excluded. The observational nature of this study precludes causal inference; we cannot determine if intensive ventilation was applied inappropriately or was simply a reflection of the patient's clinical status. While both factors likely influence the results, the data highlight an opportunity for optimization. By understanding the relationship between mechanical power and compliance, clinicians may be able to manage ventilation more effectively and enhance protective strategies.

5. Conclusions

In children with cancer, mechanical power—particularly when normalized by dynamic compliance—served as a robust independent predictor of ICU mortality and fewer ventilator‐free days. This association was most pronounced in patients with concurrent thrombocytopenia, immune dysfunction, or renal failure. Our results suggest that integrating mechanical power into protective ventilation protocols may offer superior prognostic value and guide more precise management for vulnerable pediatric patients.

Author Contributions

Bruno Sanchez Camargo: conceptualization, investigation, writing – original draft, methodology, writing – review and editing, visualization, data curation. Gabriela Maria Virgílio Dias Santos: conceptualization, investigation, writing – original draft, methodology, visualization, writing – review and editing, data curation. Orlei Ribeiro de Araujo: data curation, supervision, formal analysis, project administration, writing – review and editing, visualization, methodology, conceptualization, investigation, writing – original draft. Carolina Oberg Ribeiro: conceptualization, investigation, writing – original draft, methodology, visualization, writing – review and editing, data curation. Dafne Cardoso Bourguignon da Silva: conceptualization, investigation, writing – original draft, methodology, writing – review and editing, formal analysis, supervision, project administration, visualization.

Funding

The authors have nothing to report.

Ethics Statement

The study protocol was approved by the Research Ethics Committee of the Federal University of Sao Paulo (Certificate of Submission of Ethical Appraisal no. 68803323.9.0000.5505; approval date: July 19, 2023). All procedures complied with Brazilian ethical regulations.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File

1.

STROBE Statement—Checklist of items for cohort studies

	Item No	Recommendation	Page No
Title and abstract	1	(a) Indicate the study's design with a commonly used term in the title or the abstract	2
		(b) Provide in the abstract an informative and balanced summary of what was done and what was found
Introduction
Background/rationale	2	Explain the scientific background and rationale for the investigation being reported	3
Objectives	3	State specific objectives, including any prespecified hypotheses	3
Methods
Study design	4	Present key elements of study design early in the paper	4
Setting	5	Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow‐up, and data collection	4
Participants	6	(a) Give the eligibility criteria and the sources and methods of selection of participants. Describe methods of follow‐up	4
		(b) For matched studies, give matching criteria and number of exposed and unexposed	NA
Variables	7	Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable	4–5
Data sources/measurement	8*	For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group	4–5
Bias	9	Describe any efforts to address potential sources of bias	4–5
Study size	10	Explain how the study size was arrived at	4
Quantitative variables	11	Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why	5
Statistical methods	12	(a) Describe all statistical methods, including those used to control for confounding	5
		(b) Describe any methods used to examine subgroups and interactions
		(c) Explain how missing data were addressed
		(d) If applicable, explain how loss to follow‐up was addressed
		( e ) Describe any sensitivity analyses
Results
Participants	13*	(a) Report numbers of individuals at each stage of study—eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow‐up, and analysed	5
		(b) Give reasons for non‐participation at each stage
		(c) Consider use of a flow diagram
Descriptive data	14*	(a) Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders	5–6
		(b) Indicate number of participants with missing data for each variable of interest
		(c) Summarise follow‐up time (eg, average and total amount)
Outcome data	15*	Report numbers of outcome events or summary measures over time	5–6
Main results	16	(a) Give unadjusted estimates and, if applicable, confounder‐adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included	7–8
		(b) Report category boundaries when continuous variables were categorized
		(c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period
Other analyses	17	Report other analyses done—eg analyses of subgroups and interactions, and sensitivity analyses	7–10
Discussion
Key results	18	Summarise key results with reference to study objectives	11–12
Limitations	19	Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias	13
Interpretation	20	Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence	13
Generalisability	21	Discuss the generalisability (external validity) of the study results	13
Other information
Funding	22	Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based	1

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.