The authors used a structured descriptive approach rather than a formal guideline process. For each recommendation, they identified the predominant study design, judged the consistency of findings across available studies, summarised residual uncertainties, and assigned a descriptive evidence-strength label of high, moderate, low, or insufficient. This transparency is important, because the headline conclusion is sobering: the evidence underpinning many widely practised rehabilitation strategies remains limited and heterogeneous, preventing firm recommendations and leaving clinicians to navigate a field where well-established physiological reasoning is not always matched by rigorous trial data.

Underpinning the syndrome is a cascade of biological injury that begins in the acute phase. The strongest evidence points to early neuromuscular damage and rapid skeletal-muscle wasting during critical illness. ICU-acquired weakness, the hallmark of physical PICS, reflects a combination of critical illness myopathy, polyneuropathy, and disuse atrophy, driven by enhanced proteolysis through the ubiquitin-proteasome and autophagy pathways, suppressed protein synthesis, oxidative stress, mitochondrial dysfunction, and impaired excitation-contraction coupling. Beyond the loss of muscle mass, persistent alterations in muscle metabolism, mitochondrial bioenergetics, and gene expression have been documented in survivors, and emerging work implicates epigenetic changes such as altered DNA methylation detectable years after ICU admission.

The brain is similarly vulnerable. Systemic inflammation can compromise the blood-brain barrier, allowing peripheral mediators to trigger central neuroinflammation. Activated microglia and disrupted neuronal networks contribute to cognitive dysfunction and mood disturbance, while hypoxaemia, impaired cerebral autoregulation, and microvascular injury exacerbate neuronal vulnerability. Structural imaging frequently reveals white-matter abnormalities associated with long-term cognitive deficits, and critical illness episodes have been linked to accelerated cognitive decline resembling that seen after moderate traumatic brain injury. Neuroendocrine disruption affecting the hypothalamic-pituitary-adrenal, thyroid, and gonadal axes may contribute to catabolism, metabolic dysregulation, and impaired recovery, persisting in some patients for five years or more. Frailty occupies a bidirectional position, serving both as a risk factor for critical illness and as a consequence of it, further complicating attribution of long-term impairments.

Defining the syndrome itself remains contested. The 2012 consensus definition requires impairment in at least one of the three core domains, and the authors argue against tightening this threshold, noting that multidomain impairment is relatively uncommon, affecting roughly a quarter of survivors at three months and about 21 percent at twelve months, while single-domain impairment still carries substantial functional cost. PICS overlaps with post-sepsis syndrome, chronic critical illness, post-COVID conditions, and myalgic encephalomyelitis/chronic fatigue syndrome, and attribution is further complicated by pre-existing cognitive impairment, psychiatric illness, and physical disability. Because no dedicated International Classification of Diseases code exists for PICS, coding, reimbursement, and sustainable staffing remain barriers, although a national code has recently been approved in Germany.

On assessment, the review recommends a pragmatic two-step pathway based on expert consensus rather than validated diagnostic criteria. Brief screening tools, including the PHQ-4 for mental health, the Mini-Cog for cognition, and the Timed Up-and-Go for physical function, are applied to all survivors, with comprehensive multidisciplinary evaluation reserved for those who screen positive, ideally around three months after hospital discharge and with reassessment at six and twelve months. Additional clinically relevant domains, including fatigue affecting more than half of survivors in the first year, post-extubation dysphagia, persistent pain reported by roughly a third to a half of survivors, sleep disturbance, and health-related quality of life, are inconsistently captured. Return to work integrates all three core domains and is starkly affected: roughly two-thirds of previously employed survivors remain jobless at three months, and about 40 percent at twelve months.

Prevention starts in the ICU with the ABCDEF bundle, which combines pain assessment, spontaneous awakening and breathing trials, light sedation, delirium prevention, early mobilisation, and family engagement. Higher bundle performance is consistently associated with less delirium, shorter mechanical ventilation, reduced length of stay, and a higher likelihood of discharge home, although effects on mortality are inconsistent and long-term functional outcomes remain insufficiently studied. Early mobilisation, defined as mobilisation within 72 hours of ICU admission including passive and active components, shows short-term benefits in trials and meta-analyses, but long-term evidence is conflicting and appears to depend on dose and comparator. No trial has directly compared mobilisation doses, a gap identified as the highest research priority and now being addressed by the ESICM’s first international study, ERUPT. Adverse-event rates are below three percent and typically transient, with safety demonstrated even during extracorporeal membrane oxygenation, renal replacement therapy, and vasopressor support.

The transition from ICU to the general ward represents a high-risk window. In a UK survey of 25 centres, 98 percent of patients required ongoing physiotherapy and 70 percent were at risk of malnutrition, yet ward-based rehabilitation is frequently fragmented by competing priorities, lower staffing ratios, and poor handovers, creating a so-called post-ICU gap that may erode gains achieved in intensive care. Structured multidisciplinary continuity of care, explicit rehabilitation goals, and standardised handovers covering mobilisation, nutrition, and outstanding risks are therefore emphasised. After hospital discharge, individualised multimodal rehabilitation across inpatient, outpatient, community, and home-based models is supported, and meta-analyses show fairly consistent improvement in aerobic capacity, rated high-certainty in one synthesis of fourteen randomised trials, but effects on quality of life are inconsistent. Three components remain chronically under-delivered: structured pharmacist-led medication review of psychoactive drugs carried forward from the ICU, explicit treatment rather than mere measurement of pain, fatigue, sleep disturbance, and dysphagia, and vocational rehabilitation to support return to work.

The organisation of post-ICU recovery services varies dramatically across health systems, from well-integrated national follow-up programmes in parts of Europe to very limited provision elsewhere. A cautionary finding emerges from a 2024 randomised trial in which a hospital-based, intensivist-led multidisciplinary consultation model produced worse one-year outcomes than standard follow-up, possibly because the intervention was largely diagnostic and referral-based, with treatment changes in fewer than eight percent of patients and modest attendance. The authors argue this does not undermine survivorship care itself, but rather exposes the danger of assuming that a single clinic-centred model suits every survivor. Emerging approaches, including telehealth, virtual reality, and peer-support programmes, aim to widen access, though evidence remains preliminary and digital barriers persist for frail or remote patients.

On the family dimension, the review reframes PICS-family through a systems perspective in which critical illness affects the family as an interdependent biopsychosocial unit, shaping and being shaped by the patient’s recovery trajectory. Nurse-led, multicomponent family interventions delivered during and after the ICU stay appear promising for improving communication, shared decision-making, and early psychological outcomes, whereas ICU diaries have not shown consistent benefit. The review closes with a research agenda: harmonised diagnostic criteria, longitudinal phenotyping of recovery trajectories, determination of the optimal dose and timing of rehabilitation, rigorous head-to-head comparison of follow-up models, mechanistic studies incorporating biomarkers and epigenetics, fair and transparent use of artificial intelligence for risk prediction, culturally and spiritually responsive care, and a universal ICD code. Until that agenda is fulfilled, the authors conclude, post-intensive care syndrome should be understood not as a static post-discharge diagnosis but as a dynamic recovery process spanning the entire continuum from intensive care admission to long-term survivorship.

Subject of Research: Post-intensive care syndrome and rehabilitation across the ICU recovery continuum

Article Title: Rehabilitation and the post-intensive care syndrome across the recovery continuum: a narrative review

Article References: Rehabilitation and the post-intensive care syndrome across the recovery continuum: a narrative review. (n.d.). https://doi.org/10.1007/s00134-026-08600-5

Image Credits: AI Generated

DOI: 10.1007/s00134-026-08600-5

Keywords: post-intensive care syndrome, critical illness recovery, ICU-acquired weakness, rehabilitation, ABCDEF bundle, early mobilisation, cognitive impairment, PICS-family, ICU follow-up, neuroinflammation, skeletal muscle wasting, survivorship care

The trial, registered under numbers NL6179 and NTR6326, was designed as a prospective non-randomised cluster study and enrolled one hundred adults between July 2018 and June 2023. Rather than randomising individual patients, the investigators assigned whole institutions to different care models: the University Medical Centre Groningen, a tertiary referral centre, delivered a structured goal-directed nutritional support protocol known as GDNS, while the Hospital Group Twente, a secondary hospital, provided usual care. This cluster design was chosen deliberately to minimise the risk that the intervention protocol would contaminate routine practice at a single site. Eligible patients were over eighteen years old, had histologically confirmed, previously untreated oesophageal cancer, and were scheduled for curative-intent chemoradiotherapy and surgery. Patients undergoing salvage oesophagectomy, those with cervical lymph node involvement or distant metastases, post-cricoid tumours, poor performance status, or an inability to complete questionnaires were excluded. Fifty patients were included in each arm.

The intervention itself was built on three technical pillars. First, each patient in the GDNS group was assigned a dedicated dietitian acting as a case manager, who monitored dietary intake continuously and performed all nutritional assessments from baseline onward. Second, energy requirements were measured rather than merely estimated: indirect calorimetry was performed in 88 percent of patients at baseline, complementing standard predictive equations, alongside the Patient Generated-Subjective Global Assessment. Third, nutritional support—oral nutritional supplements, enteral tube feeding, and parenteral nutrition where necessary—was provided proactively rather than reactively. Usual care, by contrast, involved dietitians at multiple locations who monitored weight and intake during chemoradiotherapy and initiated support when deemed necessary, with research nurses conducting assessments using the short form of the PG-SGA, but without routine recording of intake or measured energy expenditure.

The study’s primary endpoint was the change in appendicular skeletal muscle index, or ASMI, the mass of limb skeletal muscle normalised to squared height, expressed in kilograms per square metre. Because computed tomography, the reference standard for muscle quantification, was only systematically available in the intervention group, the researchers relied on bioelectrical impedance analysis using a Seca mBCA 525 device, applying Sergi’s validated prediction formula and cross-checking the estimates against two independent biomarkers: skeletal muscle area measured on abdominal CT scans at the third lumbar vertebra level, and urinary creatinine excretion from 24-hour collections. The correlations were convincing—bioelectrical ASMI tracked CT-derived skeletal muscle index with correlation coefficients rising from 0.48 at baseline to 0.96 at twelve months, and urinary creatinine correlated at 0.58 and 0.72 at baseline and one year, respectively. Measurements were taken at seven or more timepoints: baseline, one week and three to six weeks after the start of chemoradiotherapy, between radiation completion and surgery, at oesophagectomy, before discharge, and at three, six, and twelve months postoperatively.

The central finding is a distinctive decline-recovery-decline pattern in muscle mass that played out identically in shape across both groups. During chemoradiotherapy, ASMI fell by 2.53 percent in the GDNS group and 3.20 percent under usual care—a modest loss compared with the pooled figure of roughly minus 6.69 percent reported in meta-analyses of neoadjuvant therapy. Between the end of radiotherapy and surgery, muscle mass actually rebounded in both arms, gaining 0.95 percent with GDNS and 2.51 percent with usual care, consistent with a recovery window that prior prehabilitation research has also documented. Then, after oesophagectomy, the pattern reversed: by twelve months, ASMI had fallen by 5.75 percent in the intervention group and 7.09 percent under usual care. Notably, in the peri-operative window alone—the stretch from the preoperative visit to early recovery—the usual care group lost 4.88 percent of limb muscle while the GDNS group lost only 0.93 percent, the sole between-group comparison that reached nominal statistical significance (P = 0.038), although this signal dissolved after adjustment for confounders such as surgical duration, blood loss, and hospital stay.

Formal statistical modelling reinforced the picture of a real but statistically non-significant difference favouring the intervention. An analysis of covariance incorporating sex, age, smoking, body mass index, and complications found no significant between-group difference in relative ASMI change at any phase, and sensitivity analyses adding sepsis to the covariates confirmed the result. Generalised linear mixed-effects modelling, which best captured the non-linear trajectory with a random-intercept model containing a quadratic time term over a mean seventeen-week interval from baseline to surgery, likewise detected no intervention effect. The investigators had powered the study to detect a five percent improvement in muscle mass during chemoradiotherapy; the observed differences, though directionally consistent, fell short of that threshold. Reduced sample size—driven partly by attrition, with only 16 and 21 patients respectively completing the full twelve-month follow-up, and by the logistical strains of conducting research during the COVID-19 pandemic—likely constrained the study’s ability to confirm what the trends suggest.

Beyond the primary endpoint, the goal-directed approach left clearer fingerprints on other measures of nutritional status. Energy intake in the GDNS group rose from 23.4 to 28.0 kcal/kg and protein intake from 1.03 to 1.32 g/kg during chemoradiotherapy, values that closely match international ESPEN guideline targets of 25 to 30 kcal/kg and at least 1.0 to 1.5 g protein/kg. Body mass index declined less steeply during radiotherapy in the intervention group, and while waist circumference and fat mass index increased slightly under GDNS, they decreased under usual care. Urinary creatinine, a biochemical surrogate of total muscle mass, remained stable in the intervention group across the first six months but declined significantly under usual care—a between-group difference that did reach significance. Malnutrition defined by the GLIM criteria nonetheless rose in both arms during chemoradiotherapy, from 34.1 to 58.5 percent in GDNS and from 35.4 to a striking 85.4 percent under usual care, and quality of life, assessed with the EORTC QLQ-C30 questionnaire, was significantly better at the end of radiotherapy in the intervention group before recovering in both arms after surgery.

The study also surfaced uncomfortable truths about translating measured physiology into clinical targets. Although indirect calorimetry revealed that measured resting energy expenditure at baseline—1854 kcal on average—significantly exceeded the 1674 kcal predicted by standard equations, the measured targets were actually applied in clinical practice in only 22 percent of cases at baseline and 37.5 percent later in treatment, apparently reflecting dietitians’ habitual reliance on predictive equations and a blanket 30 percent physical-activity correction that overlooks interindividual variation. Surgical outcomes added further complexity: the GDNS group experienced longer operations, greater blood loss, longer hospital stays, and a higher rate of postoperative sepsis (19.4 versus 2.6 percent), which the authors attribute plausibly to a learning curve, since robot-assisted oesophagectomy was being introduced at the tertiary centre during the study while the secondary hospital had already mastered it. One-year overall and disease-free survival did not differ between the arms.

Set against the wider literature, the modest muscle loss observed in this trial is itself noteworthy. Previous cohorts of oesophageal cancer patients undergoing neoadjuvant chemoradiotherapy have reported substantially larger declines, and the present figures are more comparable to those seen with neoadjuvant chemotherapy alone, hinting that both intensive support and improving usual care may contribute to attenuation. The findings also echo the growing consensus that nutrition alone cannot fully protect muscle: the PERFECT exercise trial found that resistance training preserved fat-free mass after oesophagectomy but was thwarted by inadequate postoperative protein intake, exactly the deficit the Dutch team observed between three and six months after surgery in their intervention group. The authors argue that this points squarely toward personalised, integrated interventions combining dietetics and physical therapy, with particular attention to the postoperative phase, where the steepest losses occurred in both arms.

Ultimately, the study delivers a measured verdict. A dedicated dietitian, calorimetry-based targets, and proactive tube feeding and supplementation produced better nutritional intake, more favourable body composition trends, and improved quality of life, and smaller—but not statistically significant—muscle loss during chemoradiotherapy and at one year after surgery. Survival was unchanged, and unexpected surgical complications in the intervention cohort cloud the risk-benefit calculus. What the trial establishes most firmly is the biological narrative: muscle mass in oesophageal cancer follows a predictable rhythm of loss during radiation, partial recovery before surgery, and renewed decline afterward, and the postoperative period is where the greatest opportunity for intervention still lies. For clinicians managing this notoriously catabolic disease, the message is that even intensive nutrition cannot be a substitute for a genuinely integrated, individualised recovery strategy.