Body composition-guided protein dosing in critically ill patients: a before-and-after study (PROGRESS-ICU)

Abstract

Background

To compare whether fat-free mass (FFM)-adjusted protein dosing is associated with different clinical and metabolic outcomes compared with total body weight (TBW)-based dosing in critically ill patients in the intensive care unit (ICU).

Methods

Adult ICU patients (April 2020-October 2025) with serial bioelectrical impedance analysis were included in the PROGRESS-ICU observational cohort study. Patients received either TBW-based protein dosing (1.5 g/kg/day; April 2020-June 2023) or FFM-based dosing after protocol revision (1.85 g/kg/day dry FFM, corrected for fluid overload; July 2023 onward). The primary outcome was 90-day all-cause mortality. Secondary outcomes included 30-day mortality, ICU and hospital length of stay, invasive mechanical ventilation (IMV) duration, and longitudinal changes in FFM and urea-to-creatinine ratio (UCR). Survival was analysed using multivariable Cox regression and inverse probability of treatment weighting; longitudinal outcomes were assessed using mixed-effects models.

Results

Among 620 patients (310 per cohort), median age was 69 years, and ~ 60% were male. Notable baseline differences included a higher proportion of COVID-19 admissions in the TBW cohort (40% vs. 1%), which may have influenced outcomes. Ninety-day mortality did not differ significantly between groups (adjusted hazard ratio [HR] 0.78, 95% CI 0.56–1.09). FFM-based dosing was associated with shorter ICU stay (HR 0.79, 95% CI 0.69–0.91) and fewer IMV days (HR 0.79, 95% CI 0.65–0.96); hospital length of stay was similar (HR 0.86, 95% CI 0.67–1.10). TBW-based dosing was associated with higher UCR trajectories and greater FFM loss. These associations should be interpreted cautiously due to the observational design, substantial baseline imbalances, and potential residual confounding.

Conclusions

In this observational before–after study, FFM-based protein dosing was not associated with differences in mortality compared with TBW-based dosing. Associations with ICU stay, IMV duration, and metabolic markers were observed but may reflect residual confounding and temporal changes. The study was underpowered to detect modest differences in mortality, and confidence intervals include the possibility of clinically relevant benefit or harm. These findings are exploratory and require confirmation in adequately powered randomised controlled trials.

Trial registration

The study was approved by the local ethics committee (protocol number 2403–017; 11 April 2024) and registered prior to data analysis on ClinicalTrials.gov (NCT07312708; 5 December 2025), the Open Science Framework (https://doi.org/10.17605/OSF.IO/3SU49; 4 December 2025), and the Dutch Central Committee on Research Involving Human Subjects (NL-010953; 26 October 2025). Patients admitted before ethics approval were retrospectively included under a waiver of informed consent, and from April 2024 an opt-out consent procedure was implemented for all new admissions. Data were collected as part of routine clinical care. The study protocol, including prespecified analyses, was finalised and registered before the research team accessed the data.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13054-026-06107-5.

Background

Skeletal muscle wasting is a common and severe complication of critical illness, with 10–20% of muscle mass lost within the first week of Intensive Care Unit (ICU) admission [1]. To mitigate this loss, protein provision is a central component of nutritional strategies, as protein is the primary macronutrient for preserving fat-free mass (FFM) [2]. The American Society for Parenteral and Enteral Nutrition (ASPEN) recommends 1.2–2.0 g/kg/day of protein during critical illness [2, 3], while the European Society for Clinical Nutrition and Metabolism (ESPEN) advises 1.3 g/kg/day [4]. These recommendations are mostly based on total body weight (TBW) and usually do not consider individual body composition. Adjusting for fat-free mass (FFM) is recommended only in certain groups, such as patients with obesity [4].

Emerging evidence suggests that higher protein provision during critical illness may not improve outcomes and may even be harmful [5–7]. In the PRECISe trial, high enteral protein provision was associated with predominantly harmful effects in female patients [6], who generally have a lower FFM-to-TBW ratio than males. Importantly, FFM also varies between critically ill patients with similar TBW, independent of sex [8]. Together, these findings highlight the limitations of TBW-based protein dosing and underscore the potential relevance of individual body composition in nutritional decision-making [9].

A potential approach to protein dosing in critical care is to adjust intake based on FFM rather than TBW [10]. It aligns with precision nutrition principles and may be relevant given sex-related differences in body composition [8]. However, the impact of FFM-based protein dosing on clinically relevant outcomes, including mortality, ICU and hospital length of stay, FFM preservation, and urea-to-creatinine ratio (UCR), remains unclear. Therefore, this single-centre observational before–after study aimed to investigate whether FFM-based protein dosing is associated with clinical and metabolic outcomes compared with standard TBW-based dosing in critically ill patients. The primary outcome was 90-day all-cause mortality. Secondary outcomes included 30-day mortality, ICU and hospital length of stay, duration of invasive mechanical ventilation (IMV), longitudinal changes in FFM and UCR, and protein and energy intake during the first 14 ICU days.

Methods

Study setting

The PROGRESS-ICU cohort study (PROtein Guided by REal body compoSition in the ICU) was conducted in a 17-bed mixed ICU at Gelderse Vallei Hospital (Netherlands). From April 2020, bioelectrical impedance analysis (BIA) was performed at ICU admission and every three days thereafter as part of routine care, except when contraindicated (e.g., pregnancy, wounds, or amputations). From April 2020 to June 2023 (TBW cohort), protein dosing targeted 1.5 g/kg TBW. After protocol revision in July 2023, protein dosing was based on 1.85 g/kg dry FFM (FFM cohort). The FFM cohort was initiated one week after protocol revision to reduce potential implementation bias.

Study design and participants

All adult ICU patients admitted between April 2020 and October 2025 were screened for inclusion. This single-centre observational before–after cohort study was embedded in routine clinical practice. Initially, the study followed a retrospective design, as protein dosing protocols were based on TBW prior to the formal study. From July 2023, the protein dosing protocol was revised as part of routine practice, transitioning to FFM-adjusted dosing. Starting in April 2024, after ethics committee approval, the study adopted a prospective observational design.

All data were collected as part of routine clinical care, ensuring consistency across patients and cohorts. The research team accessed the data only after the study protocol and prespecified analyses had been submitted and the study had been registered. All consecutive eligible patients were included according to prespecified criteria, which were identical across both cohorts, minimising potential selection bias. Inclusion required a BIA measurement within 48 h of ICU admission and at least one follow-up BIA during the ICU stay. Patients needed at least two BIA measurements to allow analysis of longitudinal changes in FFM. While this criterion enabled evaluation of patients with longer ICU stays, it introduces potential survivorship bias by excluding patients who died early.

Patients admitted to the ICU within two weeks prior to the current admission were excluded to avoid bias from potential muscle wasting, which could have altered baseline FFM and compromised the reliability of protein dosing in the FFM-based cohort. No formal sample size calculation was performed as recruitment continued until the FFM-based cohort matched the TBW-based cohort, resulting in two equally sized cohorts. Patients were followed for the entire duration of their hospital stay, with 90-day mortality data available for all patients.

Ethical approval

The study was approved by the local ethics committee (protocol number 2403–017; April 11, 2024) and was registered before data analysis on ClinicalTrials.gov (NCT07312708; December 5, 2025), the Open Science Framework (doi.org/https://doi.org/10.17605/OSF.IO/3SU49; December 4, 2025), and the Dutch Central Committee on Research Involving Human Subjects (NL-010953; October 26, 2025). For patients admitted before ethics approval (April 2024), a waiver of informed consent was granted, as obtaining retrospective consent was infeasible and excluding these patients would have biased mortality estimates. From April 2024, an opt-out consent procedure was introduced, where all admitted patients were informed about the study via a leaflet and had the option to opt out. No patients declined consent. The study protocol was defined and registered prior to data analysis, and data collection was embedded in routine clinical practice before formal study registration.

Nutritional assessment and nutrition targets

In the TBW cohort, protein dosing was set at 1.5 g/kg TBW for BMI < 27 kg/m², 1.5 g/kg adjusted body weight for BMI 27–30 kg/m², and at ideal body weight for BMI > 30 kg/m² (Supplementary Appendix 1) [4]. In the FFM cohort, protein dosing was based on 1.85 g/kg dry FFM determined by BIA measurement at ICU admission, assuming FFM makes up 70% of TBW [11]. Both dosing strategies followed international guideline targets [2–4]. Subsequent BIA measurements did not alter protein targets, ensuring that targets remained constant despite expected FFM loss. Energy targets were similar between cohorts and calculated using predictive equations, with adjustments by indirect calorimetry performed every three days for patients receiving invasive mechanical ventilation (Supplementary Appendix 1). Nutrition was advanced by 25% per day to reach targets by day 4. Targets and step-up schedules were embedded in the patient data management system (MetaVision, iMDsoft, Tel Aviv, Israel), enabling protocolised delivery. Serum phosphate was measured daily, and a refeeding protocol was initiated if phosphate decreased from > 0.70 mmol/L at admission to < 0.65 mmol/L within 72 h without an alternative explanation. Daily protein and energy intake from all sources (oral, enteral, parenteral, and non-nutritional) was recorded.

Outcomes & data collection

The primary outcome was all-cause 90-day mortality. Data were available for all patients and obtained from Xcare® (Nexus, Vianen, the Netherlands), an electronic health record system linked to the Dutch population registry. Secondary outcomes included 30-day mortality, ICU and post-ICU length of stay, duration of IMV, changes in FFM, UCR trajectories, and nutritional intake during the first 14 ICU days. The relationship between protein targets, intake, and mortality was also examined. Clinical, laboratory, nutritional, and body composition data were extracted from the patient data management system. Longitudinal FFM changes were assessed using repeated multifrequency BIA (InBody S10, InBody Co., Seoul, Korea), performed by trained ICU nurses as part of routine care. To account for possible fluid overload, which can overestimate skeletal muscle area [12], a correction was applied in the study. Fluid overload was calculated using a formula previously employed in ICU research: extracellular water (ECW) − (0.38 × intracellular water (ICW) / 0.62), and dry FFM was derived by subtracting fluid overload from total FFM [11, 13]. UCR measurements were collected during ICU admission, excluding values after continuous venovenous hemofiltration or active gastrointestinal bleeding in trajectory analyses. Elevated UCR may reflect increased muscle protein breakdown or protein intake exceeding the body’s capacity for synthesis [14]. Nutritional intake was calculated from ICU admission until the last complete ICU day, accounting for admission time, and expressed relative to TBW and FFM; in patients with BMI ≥ 27 kg/m², body weight was adjusted to a BMI of 27 kg/m².

Statistical analysis

Analyses were conducted using R (version 4.5.1). Continuous variables are presented as mean ± SD or median [IQR], and categorical variables are expressed as counts and percentages. Baseline characteristics were compared between groups (TBW and FFM cohort) using the appropriate tests: t-tests or Mann–Whitney U tests for continuous variables and χ²/Fisher’s tests for categorical variables. Analyses were performed in the full cohort and repeated excluding COVID-19 patients as a pre-specified sensitivity analysis to test outcome robustness.

Both 90-day (primary outcome) and 30-day mortality (secondary outcome) were analysed using Kaplan–Meier survival curves, with log-rank testing to compare survival distributions between the groups. Multivariable Cox proportional hazards models were applied to examine 30-day and 90-day mortality, adjusting for age, sex, APACHE IV score, COVID-19 status, UCR, and dry FFM, as specified in the study protocol. Mechanical ventilation at ICU admission was included as an additional covariate because of baseline imbalances and its known association with mortality [15]. Variance inflation factors (VIFs) were checked to assess multicollinearity among the covariates, with no significant issues detected. Proportional hazards assumptions were evaluated using Schoenfeld residuals. Results are reported as hazard ratios (HRs) with 95% confidence intervals. Missing data were limited to non-key variables and were not imputed. As part of the sensitivity analysis for the primary outcome, inverse probability treatment weighting (IPTW) was used to account for baseline imbalances. Propensity scores were calculated using logistic regression, incorporating the same covariates as the multivariable cox proportional hazard models. Stabilised weights were truncated at the 1st and 99th percentiles, and covariate balance was assessed using standardised mean differences (SMD), with a threshold of SMD < 0.1 considered acceptable. Finally, weighted cox models with robust standard errors were used to evaluate mortality outcomes.

Secondary outcomes, including ICU and hospital length of stay and duration of IMV, were analysed using adjusted linear regression models. The same covariates as those used in the primary outcome analysis (age, sex, APACHE IV score, COVID-19 status, UCR, dry FFM, and mechanical ventilation at ICU admission) were included in these models. Longitudinal outcomes, such as changes in dry FFM, UCR, and resting energy expenditure over time, were analysed using non-linear mixed-effects models, adjusted for age and sex, and incorporated patient-level random intercepts and group-by-time interactions.

In exploratory post-hoc analyses, dose–response relationships between protein target and 90-day mortality were assessed using sex-stratified Cox models with restricted cubic splines in the overall and TBW cohorts. In patients with an ICU stay of ≥ 7 days, similar spline-based cox models were used to evaluate the associations of achieved protein intake with 90-day mortality. Separate models were estimated for men and women, adjusting for age, APACHE IV score, COVID-19 status, and UCR.

Results

Of 2,177 and 1,496 patients screened in the TBW and FFM cohorts, 620 were included (310 per cohort) (Fig. 1). Baseline demographics were similar, with a median age of 69 years (p = 0.549) and 61% male in the TBW cohort versus 60% in the FFM cohort (p = 0.742) (Table 1). Median FFM was similar between cohorts (63.0 [54.7–71.9] kg TBW cohort vs. 63.2 [52.4–73.2] kg FFM cohort; p = 0.988), and pre-existing comorbidities did not differ (p = 0.090). The TBW cohort had more COVID-19 admissions (40% vs. 1%, p < 0.001), while the FFM cohort had higher severity scores (APACHE-II 22 vs. 20, APACHE-IV 80 vs. 72, SOFA 8 vs. 6; all p < 0.01). UCR and CRP were higher in the TBW cohort (UCR 98 vs. 81, p < 0.001; CRP 111 vs. 87 mg/L, p = 0.010).

Fig. 1.

Study flowchart. Study flowchart depicting patient screening, exclusions, and inclusion of ICU patients. This study included 620 adult ICU patients (310 per dosing strategy) to compare the outcomes of fat-free mass (FFM)-adjusted protein dosing (1.85 g/kg/day dry FFM) versus total body weight (TBW)-based dosing (1.5 g/kg/day)

Table 1.

Baseline characteristics of ICU patients with nutritional protein dosing based on total body weight equations (TBW cohort) and based on bioelectrical impedance assessed fat-free-mass equations (FFM cohort)

Variable	TBW cohort (n = 310)	FFM cohort (n = 310)	p value
Demographic variables
Age, years	69 [61–74]	69 [60–76]	0.549
Males, no. (%)	190 (61%)	185 (60%)	0.742
Total Body Weight, kg	85 [75–98]	84 [70–98]	0.152
Height, cm	174 [167–180]	173 [165–180]	0.481
Body Mass Index, kg/m2	28 [25–32]	28 [24–33]	0.095
Pre-existing comorbidities, no. (%)			0.090
Chronic heart failure	16 (5.2%)	32 (10%)	0.024
Chronic dialysis	1 (0.3%)	2 (0.6%)	> 0.999
Chronic Obstructive Pulmonary Disease	57 (18%)	65 (21%)	0.479
Chronic respiratory insufficiency	9 (2.9%)	25 (8.1%)	0.008
Cirrhosis	5 (1.6%)	4 (1.3%)	> 0.999
Diabetes	74 (24%)	75 (24%)	0.925
Haematological Malignancy	6 (1.9%)	6 (1.9%)	> 0.999
Immunological Insufficiency	34 (11%)	39 (13%)	0.618
Kidney Insufficiency	16 (5.2%)	19 (6.1%)	0.728
Metastatic Neoplasm	8 (2.6%)	13 (4.2%)	0.375
ICU admission diagnosis, no. (%)
Medical	258 (83%)	262 (85%)	0.743
Surgery elective	12 (3.9%)	10 (3.2%)	0.828
Surgery emergency	40 (13%)	37 (12%)	0.808
Covid-19 at admission	125 (40%)	3 (1.0%)	< 0.001
Proven infection	206 (66%)	168 (54%)	0.002
Upper gastro-intestinal bleeding	1 (0.3%)	9 (2.9%)	0.026
Interventions and clinical scores at ICU admission
Cardiopulmonary resuscitation	10 (3.2%)	17 (5.5%)	0.238
Mechanical ventilation at ICU admission	145 (47%)	131 (42%)	0.293
APACHE-II Score	20 [16–26]	22 [18–28]	< 0.001
APACHE-IV Score	72 [60–88]	80 [63–100]	< 0.001
SOFA Score	6 [4–9]	8 [5–10]	0.003
NUTRIC Score	4 [3–6]	5 [3–6]	0.012
Barthel Index	20.0 [20.0–20.0]	20.0 [18.0–20.0]	0.009
Laboratory Results
Blood Urea Nitrogen, mmol/L	8.0 [6.0–12.4]	8.0 [5.7–13.3]	0.782
Creatinine, µmol/L	81 [62–125]	94 [65–154]	0.008
Urea-to-creatinine ratio	98 [70–127]	81 [61–112]	< 0.001
C-reactive Protein, mg/L	111 [47–209]	87 [15–216]	0.010
Albumin, g/L	27 [23–30]	27 [22–33]	0.143
BIA measurements
Fat Mass, kg	25.4 [16.2–35.3]	21.9 [12.4–32.6]	0.006
Fat-Free Mass (FFM), kg	63.0 [54.7–71.9]	63.2 [52.4–73.2]	0.988
Dry Fat-Free Mass (FFM), kg	61.5 [53.8–70.3]	61.7 [51.2–71.3]	0.768
Males	66.4 [59.5–73.3]	66.4 [59.9–75.6]	0.643
Females	52.0 [46.8–60.6]	52.3 [46.2–61.2]	0.634
Percentage Body Fat, %	28.5 [20.8–36.2]	25.7 [17.0–34.4]	0.009
Visceral Fat Area, cm2	120.8 [85.5–176.2]	102.0 [65.0–162.4]	< 0.001
Skeletal Muscle Mass, kg	34.7 [29.9–39.6]	34.6 [28.6–40.4]	0.795
Skeletal Muscle Mass Index, kg/m2	8.8 [7.9–9.8]	9.2 [7.9–10.3]	0.071
Soft Lean Mass, kg	59.9 [51.6–68.4]	59.8 [49.4–69.2]	0.855
Total Body Water, ℓ	47 [40–53]	47 [39–54]	0.999
Extracellular Water (ECW), ℓ	18.5 [15.9–21.3]	18.9 [15.5–21.7]	0.704
Intracellular Water (ICW), ℓ	28.2 [24.5–31.9]	28.0 [23.4–32.5]	0.712
ECW/Total body water, ratio	0.397 [0.387–0.408]	0.402 [0.391–0.415]	< 0.001
50-kHz Phase Angle, °	4.5 [3.6–5.3]	4.1 [3.3–4.7]	< 0.001
Males	4.7 [3.6–5.5]	4.3 [3.5–4.9]	0.012
Females	4.2 [3.4–5.0]	3.6 [3.0–4.3]	0.002

p values < 0.05 were considered statistically significant and are presented in bold

Thisstudy included 620 adult ICU patients (310 per dosing strategy) to compare the outcomes of fat-free mass (FFM)-adjusted protein dosing (1.85 g/kg/day dry FFM) versus total body weight (TBW)-based dosing (1.5 g/kg/day). Abbreviations: APACHE, Acute Physiology and Chronic Health Evaluation; Barthel Index, a measure of the ability to perform basic activities of daily living; NUTRIC, Nutrition Risk in Critically Ill; SOFA, Sequential Organ Failure Assessment. Dry fat-free mass (dry FFM) was calculated as fat-free mass corrected for extracellular fluid overload using the following formula: dry FFM = FFM − [ECW − (0.38 / 0.62 × ICW)]

In the non-COVID subcohort, baseline characteristics were largely similar, though the TBW cohort had higher CRP (110 vs. 86 mg/L, p = 0.034) and more patients required mechanical ventilation at ICU admission (61% vs. 42%, p < 0.001) (Supplementary Table 1). These differences indicate potential residual confounding despite excluding COVID-19 patients.

Primary outcome: mortality

Clinical outcomes are summarised in Table 2. Overall 90-day mortality was 28% in the TBW cohort versus 27% in the FFM cohort (p = 0.858) (Table 2, Supplementary Fig. 1). In the non-COVID subcohort, 90-day mortality was 36% in the TBW cohort versus 27% in the FFM cohort (p = 0.066) (Table 3). No significant differences were observed for ICU, hospital, or 30-day mortality (Table 2–3, Supplementary Fig. 2).

Table 2.

Clinical outcomes, nutritional targets and achieved intakes of ICU patients with nutritional protein dosing based on total body weight equations (TBW cohort) and based on bioelectrical impedance assessed fat-free-mass equations (FFM cohort)

Variable	TBW cohort (n = 310)	FFM cohort (n = 310)	p value
Mortality Rates, no. (%)
ICU mortality	52 (17%)	48 (15%)	0.743
Hospital mortality	74 (24%)	66 (21%)	0.501
30-day mortality	69 (22%)	72 (23%)	0.848
90-day mortality	88 (28%)	85 (27%)	0.858
Length of Stay, days
ICU stay	10 [6–18]	7 [5–12]	< 0.001
Hospital stay	19 [13–32]	17 [11–26]	0.002
Post-ICU hospital stay	7 [3–11]	7 [1–11]	0.876
Respiratory outcomes
Mechanical ventilation within 24 h	224 (72%)	217 (70%)	0.595
IMV, no. (%)	231 (75%)	198 (64%)	0.005
Invasive Ventilation Duration, days	8.0 [4.5–14.9]	5.4 [2.6–9.9]	< 0.001
Non-invasive mechanical ventilation, no. (%)	111 (36%)	116 (37%)	0.739
Non-invasive mechanical ventilation, days	0.6 [0.2–2.4]	0.9 [0.2–2.8]	0.234
Nutritional targets
Vasoactive medication < 24 h	188 (61%)	220 (71%)	0.009
Acute renal failure < 24 h	90 (29%)	46 (15%)	< 0.001
CVVH, no. (%)	18 (5.8%)	34 (11%)	0.030
Refeeding syndrome, no. (%)	63 (20%)	67 (22%)	0.767
Nutritional targets
Protein target, g/day	121 [108–134]	114 [95–132]	< 0.001
Protein target, g/kg TBW/day	1.50 [1.50–1.56]	1.50 [1.36–1.68]	< 0.001
Protein target, g/kg FFM/day	1.98 [1.82–2.14]	1.85 [1.85–1.85]	< 0.001
Energy target, kcal/day	1,920 [1,727–2,120]	1,926 [1,745–2,205]	0.169
Energy target, kcal/kg TBW/day	25 [24–26]	26 [25–28]	< 0.001
Energy target, kcal/kg FFM/day	31 [28–35]	32 [29–36]	0.047
Nutritional intake during ICU (all days)
Daily protein intake, g/day	75 [51–93]	51 [29–71]	< 0.001
Daily protein intake, g/kg TBW/day	0.98 [0.71–1.19]	0.68 [0.41–0.95]	< 0.001
Daily protein intake, g/kg FFM/day	1.21 [0.89–1.56]	0.86 [0.49–1.21]	< 0.001
Daily energy intake, kcal/day	1,383 [1,056–1,673]	1,055 [682–1,392]	< 0.001
Daily energy intake, kcal/kg TBW/day	18 [15–22]	15 [9–18]	< 0.001
Daily energy intake, kcal/kg FFM/day	22 [18–29]	18 [11–23]	< 0.001
Nutritional intake during ICU (days with 100% target)
Daily protein intake, g/day	98 [81–123]	85 [64–106]	< 0.001
Daily protein intake, g/kg TBW/day	1.31 [1.06–1.52]	1.19 [0.91–1.40]	< 0.001
Daily protein intake, g/kg FFM/day	1.63 [1.29–1.97]	1.47 [1.09–1.75]	< 0.001
Daily energy intake, kcal/day	1,649 [1,368–2,055]	1,584 [1,266–1,893]	0.010
Daily energy intake, kcal/kg TBW/day	28 [22–34]	26 [20–32]	0.037
Daily energy intake, kcal/kg FFM/day	22 [19–26]	22 [16–25]	0.121

p values < 0.05 were considered statistically significant and are presented in bold

Data are presented as n (%) or median [interquartile range], as appropriate. Comparisons between cohorts were performed using χ² tests for categorical variables and Mann–Whitney U tests for continuous variables. Energy and protein targets represent the mean prescribed targets across the ICU stay. Nutritional intake variables reflect the mean daily intake across all ICU days, and separately across days on which the prescribed target was 100% (i.e., excluding the build-up phase)

Table 3.

Outcome table of non-COVID-19 ICU patients

Variable	TBW cohort (n = 185)	FFM cohort (n = 307)	p value
Mortality Rates, no. (%)
ICU mortality	32 (17%)	47 (15%)	0.649
Hospital mortality	52 (28%)	65 (21%)	0.101
30-day mortality	49 (26%)	71 (23%)	0.464
90-day mortality	66 (36%)	84 (27%)	0.066
Length of Stay, days
ICU stay	10 [6–16]	7 [5–12]	< 0.001
Hospital stay	20 [13–32]	17 [11–26]	0.006
Post-ICU hospital stay	7 [2–13]	7 [2–11]	0.557
Respiratory outcomes
Mechanical ventilation within 24 h	152 (82%)	214 (70%)	0.003
IMV, no. (%)	153 (83%)	197 (64%)	< 0.001
Invasive Ventilation Duration, days	7.4 [4.0–12.0]	5.4 [2.6–9.7]	0.012
Non-invasive mechanical ventilation, no. (%)	64 (35%)	113 (37%)	0.690
Non-invasive mechanical ventilation, days	0.6 [0.1–2.4]	0.8 [0.2–2.8]	0.149
Nutritional targets
Vasoactive medication < 24h	142 (77%)	219 (71%)	0.225
Acute renal failure < 24h	74 (40%)	46 (15%)	< 0.001
CVVH, no. (%)	12 (6.5%)	34 (11%)	0.125
Refeeding syndrome, no. (%)	47 (25%)	67 (22%)	0.423
Nutritional targets
Protein target, g/day	120 [108–135]	114 [95–132]	0.003
Protein target, g/kg TBW/day	1.50 [1.50–1.56]	1.50 [1.37–1.68]	0.008
Protein target, g/kg FFM/day	1.92 [1.77–2.11]	1.85 [1.85–1.85]	< 0.001
Energy target, kcal/day	1,923 [1,749–2,120]	1,922 [1,742–2,204]	0.342
Energy target, kcal/kg TBW/day	25 [24–27]	26 [25–28]	< 0.001
Energy target, kcal/kg FFM/day	31 [28–35]	32 [29–35]	0.037
Nutritional intake during ICU (all days)
Daily protein intake, g/day	70 [48–87]	55 [33–75]	< 0.001
Daily protein intake, g/kg TBW/day	0.94 [0.62–1.13]	0.73 [0.46–1.00]	< 0.001
Daily protein intake, g/kg FFM/day	1.12 [0.76–1.44]	0.93 [0.57–1.27]	< 0.001
Daily energy intake, kcal/day	1,339 [994–1,595]	1,147 [753–1,471]	< 0.001
Daily energy intake, kcal/kg TBW/day	18 [13–21]	16 [10–20]	0.002
Daily energy intake, kcal/kg FFM/day	21 [16–27]	19 [12–25]	0.003
Nutritional intake during ICU (days with 100% target)
Daily protein intake, g/day	96 [73–120]	84 [64–105]	< 0.001
Daily protein intake, g/kg TBW/day	1.30 [1.02–1.51]	1.18 [0.92–1.40]	0.003
Daily protein intake, g/kg FFM/day	1.55 [1.20–1.93]	1.47 [1.10–1.75]	0.019
Daily energy intake, kcal/day	1,613 [1,351–1,946]	1,592 [1,261–1,892]	0.320
Daily energy intake, kcal/kg TBW/day	26 [21–32]	26 [20–32]	0.840
Daily energy intake, kcal/kg FFM/day	21 [17–25]	22 [16–25]	0.841

p values < 0.05 were considered statistically significant and are presented in bold

This study included 620 adult ICU patients (310 per dosing strategy) to compare the outcomes of fat-free mass (FFM)-adjusted protein dosing (1.85 g/kg/day dry FFM) versus total body weight (TBW)-based dosing (1.5 g/kg/day). Only non–COVID-19 ICU patients were included in this analysis. Data are presented as n (%) or median [interquartile range], as appropriate. Comparisons between cohorts were performed using χ² tests for categorical variables, and Mann–Whitney U tests for continuous variables. Energy and protein targets represent the mean prescribed targets across the ICU stay. Nutritional intake variables reflect the mean daily intake across all ICU days, and separately across days on which the prescribed target was 100% (i.e., excluding the build-up phase)

Cox proportional hazards models in the full cohort showed no independent association between cohort and 90-day mortality (FFM vs. TBW: HR 0.78, 95% CI 0.56–1.09; p = 0.145) (Table 4), with similar survival curves (Fig. 2). No significant sex-by-cohort interaction was observed (p = 0.621). Univariate associations with 90-day mortality are presented in Supplementary Table 2. IPTW-adjusted analyses confirmed no difference in 90-day mortality (HR 0.88, 95% CI 0.65–1.21; p = 0.445) (Supplementary Table 3, Supplementary Fig. 3). Cox models similarly showed no association with 30-day mortality (Supplementary Table 4, Supplementary Fig. 4).

Table 4.

Multivariable Cox proportional hazards model for 90-day mortality

Variable	Hazard ratio (HR)	95% CI	p value
Age (per year)	1.03	1.02–1.05	< 0.001
APACHE IV (per point)	1.02	1.01–1.02	< 0.001
UCR (per unit)	1.01	1.00–1.01	< 0.001
COVID-19 (yes vs no)	0.57	0.35–0.94	0.029
Dry FFM (per kg increase)	0.98	0.968–0.997	0.021
Male sex (male vs female)	1.17	0.82–1.68	0.386
Mechanical ventilation at ICU admission (yes vs no)	1.36	1.00–1.86	0.052
FFM cohort vs TBW cohort	0.78	0.56–1.09	0.145

p values < 0.05 were considered statistically significant and are presented in bold

This study included 620 adult ICU patients (310 per dosing strategy) to compare the outcomes of fat-free mass (FFM)-adjusted protein dosing (1.85 g/kg/day dry FFM) versus total body weight (TBW)-based dosing (1.5 g/kg/day). Hazard ratios (HRs) were estimated using a Cox proportional hazards regression model. Continuous variables are expressed per unit increase, unless otherwise specified. COVID-19 and mechanical ventilation at ICU admission were included as binary variables (yes/no). Sex was coded as male vs female. The FFM cohort was compared with the TBW cohort as the reference category. An HR < 1 indicates a lower hazard of 90-day mortality, whereas an HR > 1 indicates a higher hazard

APACHE IV: Acute Physiology and Chronic Health Evaluation IV, CI: confidence interval, FFM: fat-free mass, ICU: intensive care unit, UCR: urea-to-creatinine ratio, TBW: total body weight

Fig. 2.

Adjusted 90-day survival curves for both cohorts. This study included 620 adult ICU patients (310 per dosing strategy) to compare the outcomes of fat-free mass (FFM)-adjusted protein dosing (1.85 g/kg/day dry FFM) versus total body weight (TBW)-based dosing (1.5 g/kg/day). Adjusted survival curves were generated using a Cox proportional hazards model including age, APACHE IV score, urea-to-creatinine ratio, COVID-19 status, dry fat-free mass (dry FFM), and sex. Adjusted survival probabilities were estimated by fixing continuous covariates at their cohort-wide means and categorical covariates at their model values. Solid lines represent cohort-specific adjusted survival curves, and shaded ribbons indicate 95% confidence intervals

Secondary outcomes: clinical and metabolic outcomes

In the full cohort, the FFM cohort had a shorter ICU length of stay (7 [5–12] vs. 10 [6–18] days, p < 0.001) and fewer days of IMV (5.4 [2.6–9.9] vs. 8.0 [4.5–14.9] days, p < 0.001) compared to the TBW cohort (Table 2). Similar patterns were observed in the non-COVID subcohort, with shorter ICU stay (7 [5–12] vs. 10 [6–16] days, p < 0.001) and fewer IMV days (5.4 [2.6–9.7] vs. 7.4 [4.0–12.0] days, p = 0.012) in the FFM cohort (Table 3).

In multivariable regression across the full cohort, the FFM cohort was associated with shorter ICU stay (p < 0.001) and fewer IMV days (p = 0.019), but not with post-ICU hospital length of stay (p = 0.237) (Supplementary Tables 5–7). No effect modification by sex was observed. Patients in the TBW cohort received substantially higher daily protein intakes (Fig. 3). Early target attainment was marginally higher in the TBW cohort, with no subsequent between-cohort differences (Supplementary Fig. 5–6).

Fig. 3.

Daily protein intake per kilogram body weight and fat-free mass during ICU stay. (A) Daily protein intake expressed per kilogram total body weight (TBW), (B) Daily protein intake expressed per kilogram fat-free mass (FFM). The TBW cohort, in which nutritional prescriptions were based on TBW, and the FFM cohort, in which prescriptions were based on FFM, each consisted of 310 patients. Daily protein intake during the first 15 days of ICU admission is shown for both cohorts. Boxplots show the median, interquartile range, and the full range. Numbers below the x-axis indicate the number of patients with available intake data for each study day

Exploratory post-hoc sex-stratified spline analyses suggested U-shaped associations between protein exposure per kg FFM and 90-day mortality in women, but not in men. The lowest estimated risk was observed at approximately 1.8–2.0 g/kg FFM/day. These patterns were seen both in the full cohort and within the TBW cohort (Supplementary Fig. 7–11).

Predicted dry FFM trajectories showed a small initial increase followed by a progressive decline in both cohorts (Fig. 4). At day 30, dry FFM had decreased more in the TBW cohort than in the FFM cohort (-22.3% vs -18.4% in women and -12.6% vs -8.7% in men; difference -3.94 percentage points, 95% CI -6.8 to -1.1; p = 0.007), adjusted for age (Supplementary Fig. 12). Analyses of BIA-derived parameters in the full cohort and among ICU survivors revealed similar temporal patterns across FFM, skeletal muscle mass, fat mass, and water compartments (Supplementary Fig. 13–14). When stratified by sex, the median ratio of dry FFM to TBW was 0.67 (IQR 0.59–0.76) in women and 0.76 (IQR 0.69–0.84) in men (Supplementary Fig. 15). The temporal course of phase angle did not differ between cohorts (time × cohort χ²(4) = 4.70, p = 0.32) (Supplementary Fig. 16).

Fig. 4.

Modelled trajectories of dry fat-free mass during the first 30 days of ICU stay. Predictions were obtained from a mixed-effects model including a cohort-by-time interaction, with time since first bioelectrical impedance analysis (BIA) modelled using natural cubic splines (4 degrees of freedom). The model was adjusted for baseline age and sex and included a random intercept for the patient. The TBW cohort (protein prescription based on total body weight) and the FFM cohort (protein prescription based on fat-free mass) each comprised 310 patients. Solid lines represent cohort-specific predicted trajectories, and shaded areas indicate 95% confidence intervals

UCR increased non-linearly during the first 30 ICU days in both cohorts, with the steepest rise around day 5 followed by stabilisation (Fig. 5, Supplementary Fig. 17). Temporal patterns of UCR differed between cohorts (cohort-by-time interaction χ²(4) = 50.46, p < 0.001). These differences were more pronounced in women, with UCR levels in the TBW cohort continuing to rise for a longer period, whereas levels in the FFM cohort stabilised earlier. In men, cohort-related differences were less evident. Resting energy expenditure changed non-linearly over time, with similar trajectories between cohorts and only modest time-specific deviations (Supplementary Fig. 18). The number of patients contributing data per ICU day, including available BIA, UCR, and resting energy expenditure measurements, is shown in Supplementary Fig. 19.

Fig. 5.

Modelled urea-to-creatinine ratio trajectories stratified by cohort and sex. Predictions are based on a non-linear mixed-effects model that includes natural cubic splines for time and a three-way interaction between cohort, sex, and time, with a random intercept for each patient. The TBW cohort (protein prescription based on total body weight) and the FFM cohort (protein prescription based on fat-free mass) each comprised 310 patients. The UCR was expressed as a percentage change from baseline. Panels show sex-specific predicted trajectories for each cohort over the first 30 days of ICU admission. Solid lines represent cohort-specific predicted changes in urea-to-creatinine ratio (UCR), and shaded ribbons indicate 95% confidence intervals. The horizontal dashed line denotes no change from baseline

Discussion

In the PROGRESS-ICU before–after observational study, FFM-adjusted protein dosing was not associated with differences in 30- or 90-day mortality compared with TBW-based dosing. Secondary analyses showed associations between FFM-adjusted dosing and shorter ICU length of stay, fewer IMV days, attenuation of dry FFM loss, and more favourable UCR trajectories. Notably, similar patterns were observed in the non-COVID subcohort. However, given the observational before–after design and substantial baseline imbalances between cohorts, particularly regarding COVID-19 prevalence, these findings should be interpreted with caution and cannot be considered causal.

The absence of a mortality association aligns with previous studies showing inconsistent effects of higher protein delivery on survival outcomes in critically ill patients [5–7, 16, 17]. Mortality in critical illness is multifactorial and may not be substantially altered by differences in protein dosing within currently accepted ranges [5, 6, 16–18]. The FFM-based dosing cohort was associated with shorter ICU stay and IMV duration. However, this association may reflect residual confounding from temporal changes, case-mix differences, evolving clinical practice, or unmeasured illness severity, rather than a direct effect of protein dosing.

Importantly, the additional non-COVID analyses partly addressed concerns regarding the imbalance in COVID-19 prevalence between cohorts. Similar directional associations for several outcomes were observed in non-COVID patients, supporting the consistency of the findings. However, important baseline imbalances remained present even after exclusion of COVID-19 patients, including substantially different rates of mechanical ventilation at ICU admission, indicating persistent case-mix differences between cohorts. Furthermore, the non-COVID subgroup showed a numerically higher 90-day mortality in the TBW cohort that was not observed in the full cohort, suggesting that the COVID-19 imbalance may indeed have influenced mortality estimates. Consequently, residual confounding cannot be excluded despite extensive multivariable adjustment, IPTW analyses, and sensitivity analyses.

In the FFM cohort, lower UCR trajectories and reduced dry FFM loss may indicate differences in metabolic tolerance or protein handling when protein delivery is more closely aligned with metabolically active tissue. These findings are hypothesis-generating and should not be taken as evidence of improved protein utilisation. Nakamura et al. demonstrated that a higher protein-to-muscle volume ratio correlated with increased blood urea nitrogen concentrations, suggesting limited metabolic tolerance when protein intake exceeds anabolic capacity [19]. Similarly, Umbrello et al. reported that higher protein intake was associated with lower ICU mortality only in patients with greater muscle mass [20]. Importantly, these findings are further supported by data from the EFFORT Protein trial, in which higher protein doses were associated with increased urea levels, suggesting enhanced ureagenesis and potentially limited metabolic tolerance to higher protein delivery in critically ill patients [21]. Together, these findings support further investigation into the relationship between body composition, protein exposure, and metabolic response during critical illness.

The exploratory post-hoc analyses suggesting a U-shaped association between protein targets and mortality in women should be interpreted with substantial caution. These analyses were underpowered, involved relatively few events at the extremes of protein intake, and were not adjusted for multiple comparisons. Therefore, these findings should be considered hypothesis-generating only and not evidence of established sex-specific metabolic effects. Nevertheless, the findings are biologically plausible given known sex differences in body composition and protein metabolism [22, 23], and may warrant further evaluation in adequately powered prospective studies specifically designed to assess sex-related differences in nutritional responses.

We observed that UCR trajectories differed between women receiving TBW-based dosing and those receiving FFM-based dosing. One possible explanation is that women generally have lower FFM relative to TBW, potentially resulting in relatively greater protein exposure when dosing is based on TBW. In the context of inflammation-driven anabolic resistance during critical illness, this mismatch could theoretically contribute to increased ureagenesis rather than muscle protein accretion [14]. However, UCR is influenced by multiple factors beyond protein intake, including renal dysfunction, liver dysfunction, hypovolemia, and heart failure, limiting causal interpretation of these observations [14].

Although energy intake differed modestly between cohorts, both groups were managed according to the same predefined energy targets throughout ICU admission. In clinical practice, actual energy intake is strongly influenced by disease severity and feeding tolerance, including gastrointestinal intolerance, haemodynamic instability, procedures, and interruptions of nutritional support. Consequently, energy intake may partly reflect illness severity rather than solely the nutritional strategy. Moreover, because protein and energy delivery were closely correlated in this observational setting, disentangling their independent effects was not considered methodologically reliable. Residual confounding by energy intake cannot therefore be excluded, and the observed associations may partly reflect differences in overall caloric exposure rather than protein dosing alone. Additional adjustment for time-varying energy intake was not performed, as post-baseline intake may act both as a mediator and a marker of evolving disease severity and feeding tolerance. Adjusting for such a variable could introduce overadjustment or collider bias, and would not reliably separate the independent effects of energy and protein delivery. Moreover, previous research suggests that moderate differences in caloric intake may have limited effects on major clinical outcomes in critically ill patients [24].

Several limitations warrant careful consideration. First, the non-randomised before–after design precludes causal inference and remains highly susceptible to residual confounding by temporal changes in patient characteristics, ICU organisation, evolving treatment strategies, and pandemic-related factors. Importantly, the observed associations with ICU length of stay and IMV duration may have been substantially influenced by these factors rather than the intervention itself. Although extensive multivariable adjustment, IPTW analyses, and non-COVID sensitivity analyses were performed, no statistical adjustment can fully account for large baseline differences between cohorts in a non-randomised study design.

Second, the study may have been underpowered to detect clinically meaningful differences in mortality. The absence of an a priori sample size calculation, which was not performed due to the observational design, limits the ability to exclude smaller but potentially relevant effects, such as the modest difference in 90-day mortality (28% vs. 27%). The post-hoc power analysis for the primary outcome indicated low achieved statistical power (8.6%) and a minimal detectable effect size of Cohen’s h = 0.20. In addition, the confidence intervals remain compatible with both clinically relevant benefit and harm. Therefore, the absence of a statistically significant mortality difference should not be interpreted as evidence of equivalence between strategies. To improve the reliability of future studies, predefined sample size calculations should be implemented to ensure adequate power and strengthen the validity of the findings.

Third, inclusion required at least two BIA measurements, which led to the exclusion of patients who died early in the ICU. This criterion was necessary for the longitudinal analysis of FFM and UCR, but it introduced potential survivorship bias. Post-hoc, we observed that 20 patients in the TBW cohort (0.9% of all ICU admissions) and 17 patients in the FFM cohort (1.1%) died before follow-up BIA. Although these numbers are small, their exclusion could differentially affect the remaining cohort, potentially influencing outcomes across multiple endpoints. Furthermore, although dry FFM estimates were corrected for fluid overload, the applied correction method assumes a fixed ECW/ICW ratio and has not been fully validated in critically ill patients with rapidly changing fluid status [25]. Therefore, interpretation of longitudinal FFM changes should be approached cautiously.

Fourth, to include all patients after BIA became standard of care, this study used a combined retrospective and prospective design for patient inclusion. While data collection was embedded within routine clinical practice to ensure consistency across patients and cohorts, the research team only accessed these data after the study protocol and prespecified analyses had been finalised and the study had been registered. Although all consecutive eligible patients were included according to prespecified criteria, the retrospective component and timing of data access may represent a limitation with respect to potential selection or analysis bias.

Finally, the sex-specific analyses were exploratory and underpowered, functional outcomes were not assessed, and the extent to which the observed differences in ICU stay, IMV duration, or body composition translate into meaningful long-term recovery remains uncertain.

Despite these limitations, this study represents one of the larger evaluations of body composition-guided protein dosing in critically ill patients performed in a real-world ICU setting. The before–after design allowed a pragmatic, real-world evaluation of this practice change, with consecutive patient inclusion and minimal influence on bedside staff behaviour. The findings support further investigation of body composition-informed nutritional strategies in adequately powered randomised trials with predefined subgroup analyses, careful control of confounding factors, and emphasis on functional recovery outcomes rather than short-term surrogate endpoints alone.

Conclusions

In this observational before–after study, FFM-based protein dosing was associated with shorter ICU stay, fewer days of invasive mechanical ventilation, attenuated loss of dry FFM, and more favourable UCR trajectories in secondary adjusted analyses, although no effect on mortality was observed. As the study was not designed or powered to detect specific differences, it may have been underpowered for some outcomes. Given the study design, potential residual confounding, differences in baseline case-mix, and variations in COVID-19 prevalence, these results should be interpreted cautiously. They are hypothesis-generating rather than indicative of a definitive causal benefit. Further well-powered randomised controlled trials are needed to establish whether protein dosing guided by body composition can improve clinically meaningful outcomes in critically ill patients.

Abbreviations

ASPEN

American society for parenteral and enteral nutrition

APACHE

Acute physiology and chronic health evaluation

BMI

Body mass index

BIA

Bioelectrical impedance analysis

ECW

Extracellular water

ESPEN

European society for clinical nutrition and metabolism

FFM

Fat-free mass

HRs

Hazard ratios

IPTW

Inverse probability treatment weighting

ICU

Intensive care unit

ICW

Intracellular water

IMV

Invasive mechanical ventilation

SMD

Standardised mean differences

TBW

Total body weight

UCR

Urea-to-creatinine ratio

VIFs

Variance inflation factors

Author contributions

The PROGRESS-ICU study was designed by MCP and ARHvZ. Patient screening was conducted by MCP. Data collection was carried out by MCP and FH. Data analysis was performed by MCP with verification by FH. The manuscript was written by MCP, and all authors participated in the manuscript review. All authors had full access to study data and had final responsibility for the decision to submit the final manuscript version for publication.

Funding

Not applicable.

Data availability

This article has an online data supplement. The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the local ethics committee (protocol number 2403–017; 11 April 2024) and registered prior to data analysis on ClinicalTrials.gov (NCT07312708; 5 December 2025), the Open Science Framework (https://doi.org/10.17605/OSF.IO/3SU49; 4 December 2025), and the Dutch Central Committee on Research Involving Human Subjects (NL-010953; 26 October 2025). Patients admitted before ethics approval were retrospectively included under a waiver of informed consent, and from April 2024 an opt-out consent procedure was implemented for all new admissions. Data were collected as part of routine clinical care. The study protocol, including prespecified analyses, was finalised and registered before the research team accessed the data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.