Intensive care unit-acquired weakness and mechanical ventilation: A reciprocal relationship

Abstract

Intensive care unit-acquired weakness (ICU-AW; ICD-10 Code: G72.81) is a syndrome of generalized weakness described as clinically detectable weakness in critically ill patients with no other credible cause. The risk factors for ICU-AW include hyperglycemia, parenteral nutrition, vasoactive drugs, neuromuscular blocking agents, corticosteroids, sedatives, some antibiotics, immobilization, the disease severity, septicemia and systemic inflammatory response syndrome, multiorgan failure, prolonged mechanical ventilation (MV), high lactate levels, older age, female sex, and pre-existing systemic morbidities. There is a definite association between the duration of ICU stay and MV with ICU-AW. However, the interpretation that these are modifiable risk factors influencing ICU-AW, appears to be flawed, because the relationship between longer ICU stays and MV with ICU-AW is reciprocal and cannot yield clinically meaningful strategies for the prevention of ICU-AW. Prevention strategies must be based on other risk factors. Large multicentric randomized controlled trials as well as meta-analysis of such studies can be a more useful approach towards determining the influence of these risk factors on the occurrence of ICU-AW in different populations.

Key Words: Critical illness; Meta-analysis; Morbidity; Myopathy; Polyneuropathy; Risk factors

Core Tip: Intensive care unit-acquired weakness (ICU-AW; ICD-10 Code: G72.81), an unspecified neuromuscular weakness in critically ill patients, continues to be a key concern in ICU patients and survivors. There is a definite association between the duration of ICU stay and mechanical ventilation (MV) with ICU-AW. However, the interpretation that these are modifiable risk factors influencing ICU-AW, appears to be flawed, because the relationship between longer ICU stays and MV with ICU-AW is reciprocal and cannot yield clinically meaningful strategies for the prevention of ICU-AW. Prevention strategies must be based on other risk factors.

TO THE EDITOR

Intensive care unit-acquired weakness (ICU-AW; ICD-10 Code: G72.81), an unspecified neuromuscular weakness in critically ill patients, continues to be a key concern in ICU patients and survivors[1]. It is a syndrome of generalized weakness described as clinically detectable weakness in these patients with no other credible cause and is one of the prime etiologies behind chronic impairment and insufficient functional recovery[2]. Advances in modern medicine have been unsatisfactory in reducing post-ICU morbidity and impaired quality of life[3].

In a recent major review, the risk factors for ICU-AW were enumerated to be modifiable and non-modifiable. Modifiable risk factors include hyperglycemia, parenteral nutrition, vasoactive drugs, neuromuscular blocking agents, corticosteroids, sedatives, some antibiotics, and immobilization. Non-modifiable risk factors include the severity of disease, septicemia and systemic inflammatory response syndrome, multiorgan failure, prolonged mechanical ventilation (MV), high lactate levels, older age, female sex and pre-existing systemic morbidities[3].

In their case-control study, Wang and Long[4] tried to stress preventive measures based on the risk factors identified by the machine learning algorithm. The length of ICU stay has been shown as the most important factor contributing to ICU-AW (100.0%), followed by the duration of MV (54.9%). The prediction of ICU-AW by the neural network model showed 92.2% sensitivity and 82.7% specificity. We believe that the machine-learning model has calculated the association between the duration of ICU stay and MV with ICU-AW correctly. However, in our view, the interpretation that minimizing ICU stay and MV duration could be a clinically useful primary preventive strategy against ICU-AW appears to require a revisit. This is because the relationship between longer ICU stays and MV with ICU-AW is reciprocal.

Continued use of assisted ventilation raises the possibility of diaphragmatic dysfunction and ICU-AW, which in turn raises the possibility of unsuccessful weaning from assisted ventilation[5]. Thus, it cannot be ascertained whether extended periods of MV lead to ICU-AW, or if the converse is true. As the total length of ICU stay can be related to the duration of MV, and as ICU-AW would also lead to more prolonged ICU stay, the correlation between these factors cannot imply the causation of ICU-AW due to these factors.

The methodology adopted by the authors also raises some questions. One of our concerns is the genesis of the cut-off of 14 d determined to study the clinical parameters of patients in the ICU-AW group. The groups of ICU-AW and non-ICU-AW patients selected by the authors are not comparable at baseline. The data from the patients in the group that did not develop ICU-AW includes patients whose length of ICU stay had an interquartile range of only 1 to 4 d. How, then, was it possible to compare the clinical parameters in these patients given that their period of admission to the ICU was shorter than 14 d? Is it also possible that patients not classified as having ICU-AW at day 14, could subsequently develop ICU-AW? These concerns need to be addressed in future studies.

We are also apprehensive about the confounding effect of the difference in the dose of midazolam used in the two groups. It is questionable whether patients in both groups were at the same level of consciousness while testing muscle strength. It is known that the Medical Research Council Scale is most accurate in conscious patients, as voluntary limb muscle strength is used for assessment. This would be lower, given the administration of midazolam in the patients diagnosed to have ICU-AW in the authors’ sample, while the non-ICU-AW group received no sedatives[3]. A true assessment could be made if the extent of sedation of patients in the two groups was similar[6].

The conclusion drawn from the study does not seem justified given the study methods. Even if there is an association between the length of ICU stay, MV and ICU-AW, it cannot be asserted that minimizing ICU stay and MV duration could be a clinically useful primary preventive strategy against ICU-AW. If even a brief duration of MV may trigger ICU-AW, and if it is implied that decreasing the duration of MV is preventive[4], to what extent is it clinically relevant that MV can be reduced for prevention of ICU-AW?

Utilizing iterative machine learning techniques to assess the predictive performance of a multilayer perceptron neural network model for ICU-AW and using the receiver-operator-characteristic curve as done by the authors does not seem to be providing any advantage over already identified risk factors and established preventive measures[5]. Several randomized controlled trials have demonstrated the protective role of avoidance of hyperglycemia, early parenteral nutrition and reducing excessive sedation[7]. The present approach encourages liberal glycemic control (7.8-10.0 mmol/L) in very sick patients[8]. Neuromuscular electrical stimulation applied to limb muscles also has a positive effect in mechanically ventilated patients, and this effect is enhanced by combining it with physical therapy[9]. For patients on MV, an earlier institution of occupational and physical therapy, along with reduction of sedation results in more ventilator-free days in the ICU and better functional status at hospital discharge[10]. We believe that large multicentric randomized controlled trials as well as meta-analysis of such studies can be a more useful approach towards determining the influence of these risk factors on the occurrence of ICU-AW in different populations.

Multicentric randomized controlled trials are advantageous in having high internal validity, a standardized protocol with appropriate blinding and statistical power, as well as the ability to determine the effect of covert risk factors in selected cohorts. They are, however, cost and time-intensive, and may have ethical concerns, questionable external validity and limited generalizability outside their selected patient cohorts[11]. Meta-analyses, on the other hand, can identify patterns or trends that may not be apparent in discrete studies, and thus provide increased statistical power for the assessment of individual risk factors amounting to one of the highest levels of evidence[12]. One such meta-analysis concluded that the Acute Physiology and Chronic Health Evaluation II score and drugs like neuromuscular blocking agents and aminoglycosides were observed to have a significant association with ICU-AW[13]. Meta-analyses are also cost and time-efficient. They, however, may be restricted by the characteristics of the studies they include, their heterogeneity and publication bias, as well as the risk of errors in data abstraction[12]. Machine-learning-based studies are again limited by dependency on the quality and representativeness of the input data, sometimes overfitting their interpretations to the training data leading to poor generalizability. Deep learning makes it challenging to interpret how these algorithms arrive at their conclusions. There are also concerns about patient data privacy and security when using these algorithms for health research[14].

In conclusion, though the study by Wang and Long[4] is relevant, the authors seem to have relied heavily on the machine-learning algorithm, disregarding the effect of the reciprocal relationship between MV and ICU-AW which is also related to the length of ICU stay. Randomized controlled trials and meta-analyses encompassing preventive strategies addressing systemic risk factors would give a more conclusive verdict. It may be possible to develop a multifactorial algorithm to predict patients’ likelihood of developing ICU-AW thereby positively affecting patient outcomes.