Introduction

ICU-acquired weakness is a commonly encountered phenomenon with incidences ranging from 25 to 31%, which may rise up to 70% in frail high-risk elderly patients1. Critical illness and immobilization trigger the onset of acute severe proteolysis and consequential loss of muscle mass2. This can result in an extended period of mechanical ventilation, prolonged hospital stay and a reduction in long-term survival rates, functional capacity and overall perceived health related quality of life3,4,5,6.

Prophylactic and therapeutic interventions like nutritional support, anti-inflammatory agents and exercise are applied to prevent muscle atrophy and weakness and to improve outcomes7. From this perspective the additional use of performance enhancing drugs (PEDs), like anabolic steroids and other peptide hormones such as growth hormone (GH) might be of interest. PEDs have various points of engagement and may act as potential prophylactic and/or therapeutic agents, for instance by their capacity to partly counteract muscle loss by promoting protein and muscle synthesis8. PEDs have shown success in reversing weight loss in patients with human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome with mild/moderate muscle wasting and those suffering from chronic obstructive pulmonary disease. In addition, use of anabolic steroids is associated with improved respiratory function and reduced muscle wasting potentially leading to better clinical outcomes and less adverse events9,10,11,12,13. Literature highlights the positive impact of oxandrolone, an anabolic steroid, in countering muscle catabolism in ICU burn-injury patients and the combination of exercise with oxandrolone suggests a clinical benefit in patients during ICU admission14. Beside anabolic steroids and GH, erythropoietin stimulating agents (ESA) are also considered PEDs. A systematic review of 21 studies holding 5452 patients in total, conducted in 2019, looked into the effects of ESA on mortality and morbidity of critically ill patients admitted to the ICU. Within this patient category the use of ESA was associated with reduced mortality without additional adverse events15.

A comprehensive understanding of the potential effectiveness of these compounds in ICU-patients is crucial, not only for the health of the intended patients, but also for clinicians, researchers and policymakers to make evidence-based decisions in daily practice and target policy actions. This may subsequently lead to improved care and reduced healthcare associated costs through reduced ICU and hospital stay and improvement of self-reliance, activities of daily living (ADL) and health related quality of life (QoL). A systematic review and meta-analysis is therefore warranted to create an opportunity for further research and give new insights into the applicability, effectiveness and safety of these drugs in current clinical practices. The aim of our research is therefore to systematically review and statistically meta-analyse the effects of PEDs in critically ill patients on clinical outcomes such as mortality and comorbidities, length of ICU stay, length of hospitalization, duration of mechanical ventilation and nitrogen balance. Since the effect of ESA in critically ill patients was thoroughly reviewed recently15, these drugs were excluded from this systematic review.

Methods

The study was conducted according to a prespecified study protocol (PROSPERO registration number: PROSPERO 2024 CRD42024500492)16. IGF-1 was removed from the analysis; this seemed to be a relevant drug in the preliminary screening but was not found to be used. Only minor other changes were made compared to the protocol; such as using duration of mechanical ventilation as a secondary outcome. The reason for this was that it was used as an outcome measure in several studies. Adverse events were added after receiving feedback from reviewers. The protocol is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines17.

Eligibility criteria

Studies that meet the following criteria were considered as potentially relevant for this systematic review.

Inclusion criteria

Randomized clinical trials (RCTs) comparing the use of any known PED to a control group were considered for inclusion (see Table 1 for PICO structure).

The patient population examined in the trial should have consisted of critically ill patients admitted to an ICU. As an intervention during their ICU stay, they should have received PEDs in the form of anabolic steroids or peptide hormones based on the full catalogue maintained by the Dutch Doping Authority (see Supplementary material 1 for full list). Primary outcomes of interest were mortality, length of hospitalization and ICU stay. Secondary outcomes were duration of mechanical ventilation, nutritional status defined as nitrogen balance and muscle mass preservation and muscle strength. Adverse events were registered separately, but has been subjected to only minimal analysis in terms of events per PED for the total number of patients when possible.

Table 1 PICO structure.
Full size table

Exclusion criteria

Studies that did not meet the inclusion criteria; not reporting above mentioned primary or secondary outcomes; were not translated to or in English; or did not have a full-text availability were excluded; as well as studies in which PED administration was initiated after ICU discharge. Trials analysing ESAs as intervention were also excluded as this topic was covered in a recently published systematic review15.

Search strategy

We conducted a systematic search of relevant literature using multiple electronic databases, including MEDLINE via OVID, Embase, Web of Science, Cochrane Central Register of Controlled Trials and Google Scholar on the 7th of November 2023. Keywords and Medical database specific indexes were used to identify eligible studies. Search terms consisted of concepts related to “Performance Enhancing Drugs”, “Anabolic Agents”, “Testosterone”, “Intensive Care Unit”, “Critical Illness” and relevant variations. No limits were placed on publication date. All available studies from database inception to November 7, 2023 were considered.

Database specific search strategies were developed with the assistance of the Erasmus Medical Central Medical Library (see Supplementary material 2 for the database characteristics and Supplementary Material 3 for the full executed search strategy).

Study selection

Two independent reviewers (B.B.R. and C.G.H.V.) performed the screening of titles and abstracts to identify potentially eligible studies. Full text of the selected articles was then reviewed for definite inclusion by both reviewers. Additional references were acquired through snowballing, i.e., from the reference lists of analysed studies. The web-based software platform “Covidence” was accessed and used via the Erasmus Medical Centre Medical Library for simultaneous screening of studies by both reviewers. Any differences were resolved in consensus meetings or in case of prevailing dissensus a discussion with a third reviewer (S.E.H).

Data extraction

A standardized data extraction form was developed and used to collect relevant information from the included studies. The primary and secondary outcomes, as well as study characteristics (reference, sample sizes per group and endpoints), population characteristics (age, sex, type of population), intervention details (type of intervention, way of administration of the intervention, type of comparison), outcomes as well as methodological quality (blinding) were extracted per study. Both reviewers (B.B.R. and C.G.H.V.) extracted the data and checked for correctness. Disagreements were once again solved by consensus or discussion with a third reviewer (S.E.H.).

Study risk of bias assessment

The risk of bias within individual studies was evaluated by two independent reviewers (B.B.R. and C.G.H.V.) by applying the Cochrane Collaboration ROB 2 tool18. This tool is used to assess the risk of bias in essential domains within each study and categorize as “low”, “some concerns”, or “high” risk of bias accordingly. Subsequently a conclusion for the overall risk of bias was drawn after assessment of each individual study by both reviewers. Disagreements were solved by consensus or by discussion with a third reviewer (S.E.H.). Finally funnel plots on all outcomes were made to indicate the possible presence of publication bias.

Statistical analysis

Eligible studies were pooled for meta-analysis using a random-effects model with restricted maximum likelihood estimation, due to expected high heterogeneity. Data regarding mortality were reported as Relative Risk (RR) and continuous data as Mean Difference (MD) or Standardized Mean Differences (SMD) depending on the presence of heterogeneity in the measurement of the data (see Supplementary material 4). Data were reported with a 95% confidence interval (CI). Estimated sample mean and standard deviation (SD) were calculated from given median, first and third quartiles and sample size when not reported19. Continuous data reported in hours was transformed to days by dividing the mean/median and SD/Interquartile Range (IQR) by 24. Some studies reported their data with means and standard error of the mean (SEM). SEM was transformed to SD by multiplying the SEM with the square root of the sample size \(\:(SD=SEM\times\:\sqrt{N}\)). Overall pooled results were calculated and expressed in the figures, but not further discussed because of the underlying different mechanisms of the different PEDs. Subgroup analyses per type of PED were performed if possible.

Outcomes of non-eligible studies for meta-analysis were reported separately according to the manner in which they were reported in the corresponding primary studies. Results of two separate studies published as one paper were reported independently20. A P < 0.05 was considered significant.

Heterogeneity was calculated using Q- and I2 statistic. All statistical analyses were performed using R-Studio (Version 2023.09.1 Build 494).

Results

Study selection

Of the 2646 studies identified from literature searches, 230 full-text studies were assessed for eligibility. A total of 23 randomized controlled trials were included in the analysis20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42 of which 20 were acquired from the search and three from other sources, i.e., through snowballing. Results from the search and study selection are illustrated in a PRISMA flowchart43 (see Fig. 1).

Fig. 1
figure 1

PRISMA flowchart.

Full size image

Study and participant characteristics

Studies were published between 1992 and 2023. A total of 1721 participants took part in the included studies. Of those participants 767 were randomly allocated to an intervention group and 954 to a comparison group.

The reasons for ICU admission were reported as follows: sepsis; (multiple) trauma (requiring mechanical ventilation); severe burns; critical/severe illness, not otherwise specified; critical illness and multiple organ failure; critical illness (post-surgery) requiring (prolonged) mechanical ventilation; and critical illness post-(gastrointestinal) surgery.

Sex distribution in the study population consisted of approximately 68% men and 32% women. Age varied across the studies with a range of 0 to 84 years. Four trials were exclusively conducted on pediatric patients aged 0 to 18 years21,36,37,39.

Although all substances of the extensive PED list were allowed for inclusion, only three specific drugs were found to be used in the ICU-setting in RCTs, i.e. GH, nandrolone and oxandrolone. Fifteen studies administered either GH or recombinant human growth hormone (rhGH) as intervention20,21,22,23,24,25,26,27,28,29,31,32,35,37,38,40. The anabolic steroids oxandrolone and nandrolone were administered as intervention in five and two studies respectively. No other studies were found that used other anabolic steroids30,33,34,36,39,41,42. PED administration period varied between three days and one year (see Table 2 for study and participant characteristics).

Table 2 Study characteristics.
Full size table

Risk of bias assessment

A risk of bias assessment was performed through the ROB 2 tool. Six studies were classified as low risk; 13 studies as having some concerns; and four as high risk of bias. High risk was primarily seen in the domain of deviations from the intended intervention; one study had a high risk in the randomization process and one in the domain of missing outcome data (see Fig. 2).

Fig. 2
figure 2

Risk of bias assessment using ROB 2.

Full size image

Meta-analysis

Mortality adult population

Eight studies (with Takala et al.20 being analysed as two separate studies; a national and a multinational study) with a total of 763 participants reported the effects of PEDs on mortality (see Fig. 3)20,22,23,26,30,33,40,41. RhGH was used as an intervention in five studies, oxandrolone in two studies and nandrolone in one study. Subgroup analysis of rhGH and oxandrolone revealed no significant effect on mortality (RR = 1.17, 95% CI 0.64–2.14) and (RR = 0.61, 95% CI 0.17–2.20), respectively. Heterogeneity for all studies was detected (Q statistics = 17.05, P = 0.03, I2 = 60.7%).

Mortality follow-up varied among the included studies. Three studies assessed mortality through 8 days after admission22,23,26. Takala et al. reported mortality through day 21, with most excess deaths occurring within the first 10 days after treatment initiation20. Bulger et al. monitored patients until ICU discharge or death; ventilator days averaged 21.7 ± 10.1 days33. One study followed mortality through day 2830, Shaker et al. through week 840, and Anstey et al. through day 9041.

Fig. 3
figure 3

Forest plot on mortality in the adult population.

Full size image

Length of ICU stay adult population

Four studies (with Takala et al.20 being analysed as two separate studies; a national and a multinational study) with a total of 645 participants analysed the relationship between PED administration and length of ICU stay which varied between 5 and 45 days in the intervention group and 1–65 days in the control group (see Fig. 4)20,30,33,41. RhGH and nandrolone were used as intervention both in one study each20,41 and oxandrolone in two studies30,33. Subgroup analysis revealed a significant increase of ICU stay compared to placebo for rhGH (MD = 8.81 days, 95% CI 3.87–13.75), but not for oxandrolone (MD = 2.85 days, 95% CI -0.77–6.48). Only the single study which used nandrolone showed a significant favourable result compared to intervention (MD= -10 days, 95% CI -14.61 – -5.37). A single study by Duška et al.38 showed no significant difference between GH and standard care (median of 23 and 24 days respectively, p = 0.79). This study was not included in the meta-analysis due to the lack of IQR for the given medians.

Fig. 4
figure 4

Forest plot on length of ICU stay in the adult population. *Data transformed from median to mean.

Full size image

Length of hospital stay adult population

Five studies with a total of 726 participants reported results about the length of hospital stay which varied between 13 and 58 days in the intervention group and 2–54 days in the control group (see Fig. 5)20,30,33,34,41. RhGH was administered as intervention in one study20, oxandrolone in three studies30,33,34 and nandrolone in one study41. Subgroup analyses revealed increased length of hospital stay for rhGH administration (MD = 4.72 days, 95% CI 1.15–8.29). Oxandrolone showed no significant difference in length of hospital stay (MD= -3.24 days, 95% CI -13.27–6.78). The single study performed with nandrolone showed a significant favourable result compared to intervention (MD= -9.74 days, 95% CI -14.48 – -5.00). Another randomized study provided no effect estimates of the results regarding nandrolone administration and hospitalization time, and found no significant difference between the intervention group and placebo, p = 0.92042.

Fig. 5
figure 5

Forest plot on length of hospitalization in the adult population. *Data transformed from median to mean.

Full size image

Nitrogen balance adult population

Seven studies with a total of 222 participants reported results regarding the effect of PEDs on nitrogen balance (see Fig. 6)23,24,25,28,30,35,38. There were no studies which calculated nitrogen balance in the pediatric population. Nitrogen balance varied between − 14.40 ± 8.25 g in the intervention group and − 130 ± 12.75 g in the control group. RhGH was the administered intervention in six23,24,25,28,35,38 and oxandrolone in one study30. Data is expressed as SMD due to different ways of nitrogen balance measurement among the studies (see Supplementary material 4). Subgroup analyses of rhGH showed an increase in nitrogen balance (SMD = 1.4 g, 95% CI 0.47–2.32). Five additional studies analysed the effect of PEDs on nitrogen balance, but they were not included in the meta-analysis because data necessary to conduct a meta-analysis were not reported. In a study consisting of 20 participants, rhGH showed a significant positive effect on nitrogen balance (1.2 ± 6.4 (SD or SEM, unspecified in study) versus − 3.7 ± 3.8 g/day, p < 0.05)22. An additional study measured the nitrogen balance and directly noted the MD when rhGH was administered as intervention and proved no significant difference (MD = 0.7 g/7 days, 95% CI -5–6.4)27. Another study examined the nitrogen balance change for a period of seven days and demonstrated no significant difference between the two groups (-0.7 (SD 12) range (-16 to 20) for rhGH and + 6.6 (11 SD) range (-8.2 to 26) for control, p = 0.17)33. A study consisting of 18 patients examined the effect of rhGH and proved a significantly increased nitrogen balance for the intervention compared to placebo, however no quantitative data for the results was provided other than p < 0.00126. Finally, the Finnish study in the analysis by Takala20 et al. also showed a better nitrogen balance in the GH group than in the placebo group (P = 0.002, P = 0.003, and P = 0.21 on days 7, 14, and 21, respectively). Data was not shown however.

Fig. 6
figure 6

Forest plot on nitrogen balance in the adult population.

Full size image

Mechanical ventilation

Five studies with a total of 506 participants reported results of the impact of rhGH, oxandrolone and nandrolone on length of mechanical ventilation (see Fig. 7)20,28,33,40,41. The duration of mechanical ventilation varied between 4 and 50 days in the intervention group and 4–60 days in the control group. RhGH was administered as intervention in 3 studies20,28,40, while oxandrolone and nandrolone both were administered in one study only33,41. Subgroup analysis on rhGH failed to demonstrate significant improvement (MD= -1.73 days, 95% CI -10.60–7.13). A significant heterogeneity was detected between the rhGH studies (Q statistics = 24.44, P < 0.01, I2 = 93.2%). Another study, not eligible for inclusion in the meta-analysis due to no reported IQRs for the given medians, showed no differences on length of mechanical ventilation with rhGH administration compared to placebo (median 19 versus 17 days, p = 0.49)38.

Fig. 7
figure 7

Forest plot length of mechanical ventilation in the adult population. *Data transformed from median to mean.

Full size image

Pediatric population

The pediatric population21,36,37,39 was analysed separately due to their distinct physiology compared to adults and their potentially different response to GH and anabolic steroids.

Only four studies were found, of which two analysed rhGH21,36 and two oxandrolone37,39.

Mortality pediatric population

Of the four studies, only three analysed mortality in the pediatric population, with a total of 559 participants (see Fig. 8)21,36,39. The single study on the effects of rhGH showed no significant difference in mortality compared to the control group21. Subgroup analysis of oxandrolone also revealed no clear risk reduction on mortality (RR = 0.91, 95% CI 0.30–2.71)36,39.

Mortality follow-up also varied across the pediatric studies. Jeschke et al. reported mortality through approximately day 3036. Ramirez et al. recorded mortality through approximately day 12321. Finally Porro et al. administered oxandrolone for up to 12 months, and tracked mortality over the same period39.

Fig. 8
figure 8

Forest plot on mortality in the pediatric population.

Full size image

ICU stay pediatric population

Two studies37,39 with a total of 457 participants analysed the relationship between oxandrolone administration and length of ICU stay in the pediatric population (see Fig. 9). Analysis revealed a non-significant decrease in ICU stay (MD= -1.29 days, 95% CI -10.10–7.52).

Fig. 9
figure 9

Forest plot length of ICU stay in the pediatric population.

Full size image

Other outcomes

Muscle related outcomes

Hand grip strength was analysed in two studies. Neither study showed a significant difference in grip strength when comparing PED use to placebo20,41. Muscle strength change in six arm muscle groups was measured in a single study analysing the effect of nandrolone versus placebo where participants receiving placebo had a significantly larger increase in muscle strength at end of follow up compared to nandrolone (sum score increase of 17.0 ± 4.7 SD versus sum score increase of 9.3 ± 4.3 SD, p = 0.017)41. Another study analysed the upper arm circumference and sonographic quadriceps changes, which favoured nandrolone, p = 0.008 and p = 0.012, respectively42. In another study, oxandrolone administration led to a two-fold increase peak muscle torque, which was significantly higher than the control group, p < 0.0536.

Protein related outcomes

One study examined the effect of GH on the change in muscle protein fractional synthesis rate (FSR), protein content and muscle free glutamine29. GH proved to have significantly more effect on these outcomes; FSR increased 33% ± 48% in the intervention group, while no significant increase was noted for the control group (p < 0.01), protein content remained the same in the intervention, while it decreased 8% ± 11% in the control group (p < 0.01) and muscle free glutamine decreased with 207% ± 327% in the intervention group, while remaining unchanged in the control group (p < 0.05)29. Two other studies examined the effects of rhGH and rhIGF-I on muscle protein balance. The intervention led to a significant increase of protein balance compared to placebo, p < 0.05 for both studies31,32 .

Adverse effects

All studies were individually screened for potential morbidity or adverse events associated with rhGH, oxandrolone or nandrolone therapy. Marginal statistical analyses were made because of insufficient data. Of the 16 studies20,21,22,23,24,25,26,27,28,29,31,32,35,37,38,40 investigating rhGH, six20,21,23,24,26,27 showed patients developing hyperglycemia; although this was manageable through higher insulin administration. Four studies23,24,26,27 only describe statistically significant hyperglycemia.

Takala20 describes hyperglycemia in 71 (Finnish study) and 58% (multinational study) of GH treated patients, compared to 60 and 36% respectively in the control group. Ramirez21 describes hyperglycemia in 63% of children after burn injury, while the incidence was 41% in randomized placebo treated patients. A further two28,38 studies mention higher insulin requirement to control blood glucose, without strictly mentioning hyperglycemia. The study37 by Jeschke et al., where rhGH was combined with propranolol in a pediatric population, found no hyperglycemia.

Five30,33,34,36,39 studies investigated oxandrolone administration. Gervasio, Bulger and Porro30,33,39 found no significant adverse events. Similarly, the study by Wolf34 et al., found no statistically higher adverse events or complications between the treatment and control group. They do nevertheless mention increases in the liver enzymes alanine and aspartate aminotransferases. However, the principal investigators only reported these findings at their own discretion and no incidences of hepatic insufficiency were reported as adverse events. Jeschke36 et al., found significantly increased serum transaminases with oxandrolone administration, beginning at 17 days post-burn. Liver function was nonetheless not considered to be impaired because increased constitutive hepatic protein levels were also found. There were no other adverse effects on the hypermetabolic, inflammatory, or endocrinologic responses postburn.

Concerning nandrolone, only one study42 found a slightly increased low-density lipoprotein, a type of cholesterol and slightly higher triglycerides; while the other study40 detected no significant adverse events.

Discussion

We performed a systematic review and meta-analysis on the effects of PEDs in critically ill patients, to provide new insights into the applicability, effectiveness and safety of these drugs enabling evidence-based decisions in daily practice. ESA were excluded from this review, as their effects in critically ill patients were comprehensively analysed recently15. Despite the exhaustive list of PEDs (see Supplementary material 1), only studies using (rh)GH and the anabolic steroids oxandrolone and nandrolone were found. Although set up as a meta-analysis, we more specifically clarified the effect estimates for (rh)GH and oxandrolone and nandrolone in critically ill patients on mortality, length of ICU & hospital stay, nitrogen balance and mechanical ventilation; while also taking muscle related outcomes and adverse effects into consideration. Studies in adults and children were analysed separately.

Our review demonstrates that (rh)GH administration improves nitrogen balance in critically ill patients. Unlike the landmark study performed by Takala et al.20, we found no increase in mortality, potentially reopening the opportunity of research into GH therapy. There was however an increase in length of ICU and hospital stay. (Rh)GH, nandrolone and oxandrolone all fail to translate to (other) improved clinical outcomes.

Overall, the data suggest that the risk of morbidity or adverse events is low. However, strict glucose control is necessary when administering rhGH, which may require higher insulin doses. Jeschke et al.37, suggest that concurrent propranolol administration may prevent hyperglycemia in children. Liver enzymes may need to be intermittently monitored in patients receiving oxandrolone, with the knowledge that it can induce increased release of hepatic transaminases; although it is explicitly stated by both Jeschke et al.36 and Porro et al.39 that liver function was not affected because there was a significant increase in constitutive proteins, while liver enzymes were not elevated.

There could be several explanations why a positive nitrogen balance in this review does not correlate with improved clinical outcomes. At the outset, of the seven studies23,24,25,28,30,35,38 that both investigated nitrogen balance and could be analysed in the group, only one rhGH intervention also analysed mortality23 and one oxandrolone intervention30 investigated mortality, length of ICU stay and hospitalization. One could therefore conclude that it is not possible to draw definitive conclusions, because not only were the studies few in number, but because the studies that investigated nitrogen balance are not the ones that also studied other clinical outcomes. Furthermore, both studies23,30 were powered to primarily investigate the effects on nitrogen balance, not to investigate mortality, length of ICU stay or length of hospitalization. Our finding that rhGH improves nitrogen balance but fails to improve clinical outcomes, is in contrast to some studies that suggested reduced ICU mortality in case of improved nitrogen balance44,45. These studies focussed solely on positive protein intake however, and neither employed RCTs nor incorporated any form of PED. Conversely, nitrogen balance in itself could possibly be considered an outdated concept as a suitable measure of nutritional status or a tool for increased survival. Current guidelines46 on clinical nutrition in the ICU make specific statements about caloric and protein intake, that are largely dependent on timing but also necessitate dynamic re-evaluation. The guidelines make a clear distinction between the acute inflammatory phase in the early stage of critical illness, often spanning the first seven days, and the post-acute phase typically beyond the seven days. This distinction is pivotal in avoiding harm from overfeeding early on while ensuring sufficient support during recovery. In the first 48 h of ICU admission and the transitioning phase after, focus lies on minimizing risks of malnutrition and catabolism, while avoiding metabolic overload and preventing risk of overfeeding or refeeding syndrome. This translates to gradual feeding with hypocaloric nutrition (e.g. 70% of energy expenditure) and a cautious approach to protein provision (approximately 1.3 g/kg/day). After the acute inflammatory response subsides, and there is a progressive shift toward anabolism and recovery, protein targets can be increased (e.g. 1.5–2.0 g/kg/day), and caloric intake adjusted to match full energy requirements of patients46.

Our subgroup analysis showed no effect of oxandrolone on mortality, length of hospitalization and length of ICU stay. This was contrary to earlier RCTs and reviews in, importantly, non-critically ill patients where treatment with oxandrolone was associated with clinical benefit and potential reduction of adverse events9,13,14. This discrepancy could be explained by differences in patient populations (critically ill vs. non-critically ill) and in therapy initiation timing (acute vs. post-acute), with the acute phase being characterized by profound catabolism, especially in critically ill patients.

Importantly, our review shows no increase in mortality when administering rhGH. This stands in stark contrast to the landmark study by Takala et al. in 199920, which demonstrated increased mortality and subsequently let to near complete cessation of research into the potential beneficial effects of rhGH in critical care. A possible clarification for this discrepancy can be found in an observational study performed by Knox et al.47 that showed that GH administration can reduce mortality rates and shorten duration of mechanical ventilation in critically ill patients. They, however, suggested that age above 60 years and severity of critical illness could be associated with increased mortality with rhGH use47. Other studies that investigated the effect of GH are mainly conducted in critically ill, non-septic patients. Unlike Takala et al., these studies showed no significant impact on mortality21,22,23,26,40. This discrepancy might therefore be due to a lower severity of the illness in patients treated in the other studies. Although patients were reported to have no septic shock at the time of enrolment; multiple-organ failure, septic shock and uncontrolled infection were all identified as major causes of mortality in Takala’s landmark study20. An explanation for this can possible be found in GH’s modulatory effects on immune function. Depending on the underlying condition, both GH excess or deficiency can lead to a disruption of the intricate balance between proinflammatory and anti-inflammatory processes, resulting in a pathological inflammatory status48.

Given that regular physiological processes are disturbed in the critically ill, various mechanisms could be proposed to further clarify our findings. Consequently, it could also be reasoned which categories of patients might and/or might not benefit from PED administration. Plasma GH levels are known to be elevated at time of ICU-admission and non-survivors have been shown to have a seven-fold increase in GH value compared to survivors49,50. GH administration, especially in higher dosages, could therefore hypothetically be detrimental instead of beneficial, in accordance with the finding of Takala20. There was considerable variability in the administered GH dosages, between 0.05 and 0.30 mg/kg/day24,38. Although the study by Takala et al.20 showed higher mortality with dosages up to 0.13 mg/kg/day, Ramirez et al.21 used dosages of 0.20 mg/kg/day in pediatric patients without a detrimental effect on mortality. Pediatric patients are however known to tolerate higher GH dosages49. Similarly, we showed that oxandrolone in children does not seem to increase mortality or ICU stay.

The pulsatility by which GH is normally released, up to 10 times per day, is also disrupted during critical illness49,51,52,53. Although Duška et al.37 have tried to mimic normal pulsatility, many study designs fail to mimic normal physiology. Finally, there was considerable variation in timing and duration of therapy. Interventions were initiated between day 2 and 21 of ICU admission and therapy lasted between 3 days and 1 year in total. How all this translates to optimizing dosage, timing and duration of GH administration can be challenging. However, this could imply that to effectuate positive effects, dosages must be adjusted according to age, and administering PEDs may need to be delayed in the acute phase.

Furthermore, there was variability in the administration route. ICU patients are known to experience gastrointestinal problems and have altered metabolism resulting in erratic enteral uptake. Similarly, both fat and muscle mass may substantially decrease in the cachectic ICU patient. Fat solubility and overall fat content on one hand and compound characteristics and overall muscle mass on the other hand can interfere with the absorption of subcutaneously and intramuscularly administered substances52. Subcutaneous injection was the primary choice of administration of rhGH, although continuous intravenous administration and intramuscular injections were also applied in the included studies. The variable results could indicate that the different administration methods did not have an influence on the chance of adverse events. Nonetheless, for comparison in future investigations it is wise to use similar administration methods. Our recommendation would be to use intravenous injections; this method is independent of variability in gastrointestinal uptake, fat content or muscle mass and guarantees the substance is delivered correctly.

Despite our best efforts we are confronted by several limitations stemming from either the investigated studies or by our own choices.

A key limitation of this review is that 14 of the 17 analysed trials were published more than 15 years ago. Although we deliberately included all available studies to provide the most comprehensive overview of the existing evidence, the age of many of these trials raises concerns about the direct applicability of their findings to current critical care practice, which has evolved substantially over the past two decades. To strengthen the evidence base and ensure that future systematic reviews and meta-analyses reflect contemporary practice, the generation of new, high-quality trials is strongly recommended. This would allow subsequent reviews to incorporate more recent data and provide an even more complete and clinically relevant synthesis.

Furthermore, the number of studies that could eventually be analysed as a group was small. We have explicitly stated that pooled results would not be discussed, we have however calculated and expressed them in the figures as examples. This combination in a single analysis could be called questionable, because, as mentioned previously in the method section, different PEDs are characterised by different mechanisms of action and patterns of adverse effects. A concerning risk of bias was present among the studies; there were some methodological concerns primarily regarding randomization and blinding (see Fig. 2). It is nonetheless likely that most studies implemented proper randomization but failed to report the process in accordance to the Cochrane ROB 2 tool, suggesting that the actual quality of the studies might surpass our initial assessment. In addition, funnel plots illustrated the possible presence of publication bias primarily for the studies that analysed mortality, length of ICU stay and length of hospitalization (see Supplementary material 5). As mentioned previously there was methodological heterogeneity, primarily concerning interventions (dosage, duration and mode of delivery). There was however also statistical heterogeneity; eligible studies were pooled for meta-analysis using a random-effects model, due to the high heterogeneity caused mainly by small sample sizes and possibly inconsistent effect sizes. This heterogeneity makes it more difficult to draw general conclusions.

Furthermore, several studies required a transformation of the medians and IQR to estimated means and SD for the statistical analysis19. Although this transformation was necessary to conduct a meta-analysis, it may have caused small deviations from the original data, presenting a potential source of bias19.

Finally, we decided against more rigorous study selection even though some studies continued administering PEDs after discharge from the ICU and analysed the results during the recovery phase21,34,39. Similarly, mortality follow-up varied greatly, ranging from 8 days in the adult population22,23,26, to as long as 12 months in pediatric populations39. We also decided against subgroup analysis of dosages and converting to standardized units to prevent further deviation from the original data. Nonetheless, these choices raise some points of concern.

Based on the currently available evidence we assume that (higher dosages of) rhGH should not be routinely administered to adult patients during the acute setting of critical illness, similar to current clinical guidelines on nutrition. However, for certain subgroups PED therapy may be beneficial. We surmise that patients who are no longer in the acute setting of infection or critical illness, but who have developed ICU-acquired weakness, may benefit from PED therapy through its capacity to promote protein and muscle synthesis, as supported by studies in other patients with sarcopenia9,10,11. Consequently, future studies could focus on subgroup analysis of PED administration in the acute versus post-acute phases. A set of well-designed trials are needed to investigate drug timing, dose response relations and, eventually, cost-effectiveness of PED administration in critically ill patients in the context of modern critical care.

Conclusion

Our systematic review and meta-analysis shows that GH has a positive effect on nitrogen balance, but GH, oxandrolone and nandrolone don’t positively impact clinical outcomes such as mortality, length of ICU stay, length of hospitalization and duration of mechanical ventilation. This conclusion, however, is based on significant heterogeneous outcome measurements, patient populations and treatment schemes and therefore questionable methodological quality. A possible subgroup that may benefit from PEDs could be patients who have surpassed the acute phase of critical illness. That is why further research on the diverse fields of PEDs, including types, dosages as well as the timing and duration of therapy is required to acquire new insights on potential therapeutic applications of PEDs in the rehabilitation of the critically ill.