Changes in muscle ultrasound for the diagnosis of intensive care unit acquired weakness in critically ill patients

To test diagnostic accuracy of changes in thickness (TH) and cross-sectional area (CSA) of muscle ultrasound for diagnosis of intensive care unit acquired weakness (ICU-AW). Fully conscious patients were subjected to muscle ultrasonography including measuring the changes in TH and CSA of biceps brachii (BB) muscle, vastus intermedius (VI) muscle, and rectus femoris (RF) muscles over time. 37 patients underwent muscle ultrasonography on admission day, day 4, day 7, and day 10 after ICU admission, Among them, 24 were found to have ICW-AW. Changes in muscle TH and CSA of RF muscle on the right side showed remarkably higher ROC-AUC and the range was from 0.734 to 0.888. Changes in the TH of VI muscle had fair ROC-AUC values which were 0.785 on the left side and 0.779 on the right side on the 10th day after ICU admission. Additionally, Sequential Organ Failure Assessment (SOFA), Acute Physiology, and Chronic Health Evaluation II (APACHE II) scores also showed good discriminative power on the day of admission (ROC-AUC 0.886 and 0.767, respectively). Ultrasonography of changes in muscles, especially in the TH of VI muscle on both sides and CSA of RF muscle on the right side, presented good diagnostic accuracy. However, SOFA and APACHE II scores are better options for early ICU-AW prediction due to their simplicity and time efficiency.

Muscle ultrasound is a convenient approach to investigate the muscle changes over time after admission in ICU 14 . Some muscle ultrasound studies have been able to detect reduction tendency of the cross-sectional area (CSA) 15,16 or decreasing pennation angle 17 , decrescent muscle thickness (TH) 15,18 and increase in echo intensity 17,19,20 in patients who were critically ill. Nevertheless, the relation between those muscle parameters and ICU-AW remains unclear. Witteveen 's research tested the accuracy of neuromuscular ultrasound 21 and found that receiver operating characteristics with a calculated area under the curve (ROC-AUC) of muscle parameters showed the promising possibility in differentiating patients with and without ICU-AW. The hypothesis of the present study is the changes in muscle ultrasound over time may show better diagnostic efficiency in the occurrence of ICU-AW. Consequently, we carried out the present study at a single center to test the diagnostic accuracy of the changes of muscle ultrasound over time in differentiating patients with and without ICU-AW.

Methods
Population and design. This longitudinal observational study was designed to be carried out at a single center, a general ICU in Shanghai, China from June 2019 to May 2020. The study was duly approved by the Rui Jin Hospital Ethics Committee and was performed in accordance with relevant guidelines and regulations. Written informed consent (either directly or through an appropriate surrogate) was obtained from all patients. Patients aged ≥ 18 years with an anticipated ICU stay of at least 2 days were eligible for screening after being evaluated daily for awakening and reaction to simple verbal commands. Exclusion criteria comprised individuals with prior diagnosed diseases characterized by generalized or regional weakness or with any diagnosis at the time of admission making patients abnormal muscle strength and unable to follow commands (e.g., cardiac arrest, stroke, spinal injury, traumatic brain injury, or intracerebral infection), or delirium or dementia during the ICU stay. Additionally, the patients experiencing edema of upper and lower limbs and patients who did not have arms or legs for muscle strength testing or ultrasound or had wounds, fractures, lesions, burns, or bleeding at the measurement points were excluded as well. Finally, patients who received early mobilization or physical therapy during the observation period were removed from the statistics.
Clinical data collection. Baseline data were collected after ICU admission and included age, sex, Body Mass Index (BMI), hand dominance, admission diagnosis, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology, and Chronic Health Evaluation II (APACHE II) score, risk factors for polyneuropathy or myopathy (restraints, surgery, nutritional supports, mechanical ventilation, glucose peak concentration, glucocorticoid, use of sedative and analgesic) and comorbidities (cardiac dysfunction, respiratory failure, liver dysfunction, acute kidney injury, hypertension, diabetes mellitus and Multiple Organ Dysfunction Syndrome (MODS).
Ultrasound protocol. Two researchers who were trained and qualified to measure the muscle parameters immediately on the admission day by using a Philips ultrasound machine (IU22, USA) and a linear probe (frequency: 10-13 MHz) which enabled acquiring high-resolution images of clear structures of muscles 22 . Before performing, the patient must be in a supine position with extended elbows, wrists, knees and relaxed muscles, meanwhile the palms and toes of patients were facing or pointing to the ceiling 23 . The ultrasonography of muscles included TH and CSA of biceps brachii (BB) muscle, vastus intermedius (VI), and rectus femoris (RF) muscles (Fig. 1). All the muscles were measured bilaterally and scanned in the transversal (cross-sectional) image. The transducer was oriented transversally in relation to the longitudinal axis of the arm or thigh for obtaining a cross-sectional image, thus creating a right angle to the skin surface. Landmarks for ultrasound image acquisition were at standardized anatomical points, including the midpoint between supraglenoid tubercle and radial tuberosity for BB 24 , the second third of the distance between the anterior inferior iliac spine (AIIS), and the midpoint of the proximal border of the patella for RF, and the midline of the same distance as RF for VI 23 . The correlation coefficients of measurement accuracy of the two researchers were 0.88, 0.90, and 0.91 for BB muscle, RF muscle, and VI muscle respectively. When performing ultrasonography, the pressure on the skin was kept minimal, and adequate coupling agents were used for obtaining the images 25 . To enhance the accuracy of the measurement of target muscles, all the CSA and TH were measured three times continuously and an average was www.nature.com/scientificreports/ calculated as the final value. The whole muscle ultrasound procedure was repeated on day 4, day 7, and day 10 after ICU admission to know the changes of muscle TH and CSA.
Muscle strength assessment. Another two researchers who were blind for the results of quantitative measurement of muscle parameters were in charge of assessing the conscious patients for muscle strength by using the MRC score on the 10th day after ICU admission 26 . The MRC score is extensively utilized for diagnosing the ICU-AW and its good interobserver reliability in critical settings has been confirmed in a previously published study 12 .For the patients mechanically ventilated with sedatives, if the RASS (Richmond Agitation Sedation Scale) fell anywhere between − 1 and 1 27 and they showed a positive reaction to 5 verbal commands with facial muscles, we considered them feasible for muscle strength assessment 12 . Twelve muscle groups were tested for the calculation of MRC score including elbow flexion, wrist extension, shoulder abduction in upper limbs, and dorsiflexion of the foot, hip flexion knee extension in lower extremities. Examined subjects whose total MRC score < 48 were categorized as the ICU-AW group according to the international consensus statement 13 .
Sample size. According to the equation of diagnostic experiment 28 , a significance level of 0.05 for a twosided level and test power of 80% were assumed, and the expected sensitivity and specificity of muscle ultrasonography to be 0.8. Based on previous studies, ICU-AW was about 50% prevalent in critically ill patients 4 . 36 examined subjects was determined after considering the loss of 10% of the sample.

Statistical analysis.
Kolmogorov-Smirnov's normality test was employed for evaluating continuous variables' distribution. Data acquired from continuous variables with a normal distribution were expressed either as standard deviation or mean or as the interquartile range (IQR) or median in case if they had a non-normal distribution. Mann-Whitney test, Student t-test, exact Fisher test, and chi-squared test were employed to assess the differences among patients with and without ICU-AW diagnosis according to the distribution and type of the variable. Additionally, repeated measurement analysis of variance was used for testing the differences of changes in sonographic TH and CSA of observational muscles between groups. The discriminative power of changes of muscle ultrasound over time was examined with a 95% confidence interval (CI) using ROC-AUC (receiver operating characteristic curves with calculated area under the curve). The discriminative power of AUC values among 90 and 100 percent have been described as < 60 percent as failed, 60-70 percent as poor, 70-80 percent as fair, and 80-90 percent as good 29 . The change of CSA and TH are represented by ΔCSA and ΔTH respectively, and was calculated using the formula: Based on ROC curve analysis, the specificity, sensitivity, positive and negative predictive values (PPV, NPV) for muscle ΔCSA and ΔTH were calculated. The optimal cutoff value was confirmed by calculating the Youden Index. Youden Index = (specificity + sensitivity − 1). When the Youden Index is maximum, the corresponding value is the optimal cutoff value 30 . A significant two-level p-value taken for all analyses was < 0.05. SPSS version 19 was used for all statistical analyses.

Results
In total, 106 patients were enrolled and their informed consent was obtained. Among them, 37 patients finally went through all 4 times muscle ultrasonography measurement successfully, of whom 24 had ICU-AW. The flowchart of screening and inclusion is shown in Fig. 2. Table 1 enlists the patient characteristics.
Whether or not patients suffered from ICU-AW, all the groups presented a descending trend of both TH and CSA bilaterally. In the upper limbs, the changes of CSA in BB on the right side showed more statistical differences at different observation time points between groups. Moreover, ICU-AW patients had a greater degree of declination in CSA of RF bilaterally, and a remarkable reduction of TH in VI as well. Many significant differences between groups were found at different points in time (Fig. 3).
The ΔTH of BB on both sides had higher ROC-AUC than ΔCSA of BB in the upper limbs, with ROC-AUC ranging from 0.702 to 0.792. The ROC-AUC of ΔCSA of BB on both sides were not significant except ΔCSA day4 of BB on the right side. In the lower limbs, most ROC-AUC of ΔTH and ΔCSA of RF were not significant on the left side. While the ROC-AUC of ΔTH and ΔCSA of RF on the right side were significantly higher, ranging from 0.734 to 0.888, especially ΔCSA day10 of RF (ROC-AUC: 0.888, p < 0.001). Besides, ΔTH day10 of VI had fair ROC-AUC values that were 0.785 on the left side and 0.779 on the right side (Table 2; Fig. 4).
Following that, we compared the diagnostic power of SOFA, APACHE II, and certain muscle parameters that showed good diagnostic performance as previously mentioned. The SOFA (ROC-AUC: 0.886, p < 0.001) and APACHE II scores (ROC-AUC: 0.767, p < 0.05) at the time of admission to the ICU showed close diagnostic efficacy compared to the changes in muscle parameters (Fig. 4).
Further, using specific thresholds (15% for ΔTH day10 of BB, RF and VI muscle, 12% for ΔCSA day10 of BB and RF) in term of Youden Index of ΔTH day10 and ΔCSA day10 that were calculated based on the ROC curve, the sensitivity, specificity, PPV, NPV, and accuracy were confirmed and are presented in Table 3. Diagnostic accuracy of ΔTH day10 and ΔCSA day10 of RF on the right side and ΔTH day10 of VI on both sides was high and ranged from 75.7 to 78.4%.

Discussion
The present study confirmed that patients with ICU-AW had a significant reduction of muscle TH and CSA than those of patients without ICU-AW, especially in the lower extremities. Moreover, for a 15% threshold for ΔTH day10 and a 12% threshold for ΔCSA day10 , muscles in lower extremities showed a good diagnostic accuracy of the diagnosis of ICU-AW, particularly on the right side. More importantly, changes in muscle ΔTH day10 and ΔCSA day10 of the lower extremities were found to have close diagnostic validity to SOFA and APACHE II scores at the time of ICU admission. In this study, of all evaluated patients, 64.9% were found to have ICU-AW. Whether or not patients had ICU-AW, all patients presented the descending trend of both TH and CSA bilaterally. Moreover, patients with ICU-AW had an obviously greater degree of declination in CSA of RF bilaterally, and a remarkable reduction of TH in VI of both sides as well. Turton et al. carried out a study on 22 ICU patients who were mechanically ventilated and performed an ultrasonographic assessment of the flexor compartment of the elbow joint, the vastus lateral muscle, and the medial head of the gastrocnemius muscle on admission and 10 days later. The results showed that the loss of muscle mass mainly occurred in the lower extremities and there was no change in the size of the flexor compartment of the elbow joint. Such data helps to justify the argument regarding investigating the lower extremities further as peripheral muscles, in patients with critical illnesses have more chances to undergo early disuse atrophy 31 . In particular, a 3-week follow-up study employed ultrasonography to evaluate RF muscle in terms of the morphological changes and found severe muscle mass loss in CSA and muscle diameter experienced by all the ICU trauma patients. By day 20, approximately 45% of rectus femoris muscle mass was lost 32 . Consequently, in comparison to upper limbs, lower limbs muscles experienced earlier and greater atrophy. The potential reason was given in an earlier study that assessed rectus femoris CSA and protein/DNA ratio over time. The results showed that during the first week, virtually all cases decreased in muscle mass. Lower limb muscle atrophy is considered to be the result of net catabolism due to the decrease of muscle protein synthesis and the simultaneous increase of protein decomposition relative to protein synthesis 16 .  www.nature.com/scientificreports/ Few previous studies tested muscle ultrasonography for the diagnosis of ICU-AW or prediction of prognosis or similar symptoms. One study diagnosed skeletal muscle loss by measuring the CSA of RF using ultrasound and compared it with frailty to predict the prognosis of critically ill patients. The outcomes of the study clearly show that the prediction value of adverse discharge tendency by bedside ultrasound in the diagnosis of skeletal myopenia was consistent with frailty 33 . Moreover, a prospective observational study found that the larger the CSA in RF on the day of admission, the lower the occurrence of the muscle fiber necrosis and muscle waste of RF 34 . In addition, Greening et al. demonstrated that an independent risk factor for unscheduled readmission or death is a smaller quadriceps muscle size measured by ultrasound 35 . These studies showed the potential diagnostic possibility for ICU-AW diagnosis. Further, Witteveen's study performed ultrasonographic TH of the tibialis anterior muscle, biceps brachii muscle, flexor carpi radialis muscle, and the rectus femoris muscle thus finding that for these muscles, the diagnostic accuracy of muscle TH was rather low with ROC-AUC ranging from 21 51.3 to 68.0%. Nevertheless, CSA which is considered as a crucial property for contraction and strength of muscle was not fully explored for its relation with ICU-AW 14 . According to results, the changes in quantitative muscle ultrasound had good performance when analyzed on MRC criteria for the diagnosis of ICU-AW and the best cutoff ratio of reduction in muscle parameters for diagnosing ICU-AW using ultrasound is more than 15% for ΔTH day10 and more than 12% for ΔCSA day10 in the lower extremity of the right side, which endorses the use of muscle ultrasound as a supplementary tool for ICU-AW diagnosis.
Although changes in some muscle parameters over 10 days presented good diagnostic efficacy, a comparison showed that SOFA and APACHE II scores at the time of ICU admission had a more adequate advantage. Given the time efficiency and implementation efficiency, the ROC-AUC of SOFA and APACHE II scores were shown to be near to the change in muscle parameters, making it appear unnecessary to predict the occurrence of ICU-AW by a 10-day muscle observation. Many predictors for the occurrence of ICU-AW have been confirmed, and the SOFA and APACHE II scores can be considered as indicators of multiple high-risk factors integrated together 36,37 . However, from previous reports, SOFA and APACHE II scores alone did not present sufficient diagnostic efficacy 38,39 , so further validation of these results is needed considering the limited sample size of the present study.
Some limitations of this study deserve the necessary commentary. First, due to the limited availability of biopsy or electroneuromyography in the ICU, we could not classify patients with ICU-AW into the three subcategories (critical illness neuromyopathy (CINM), critical illness myopathy (CIM), and critical illness polyneuropathy (CIP)). Second, we did not observe other ultrasonographic characteristics of muscle in recognition of ICU-AW, for instance, pennation angle and echo intensity which may have better diagnostic value. Third, it was impossible for ultrasound examiners to be completely blind to the MRC score because the absence or presence of spontaneous movements gave the impression of muscle strength. Therefore, to improve the accuracy of muscle measurement, all CSA and TH were measured three times in a row, and the average value is calculated as the final value to minimize the performer's deviation.

Conclusion
Ultrasound measurement of muscles can be used as a tool to assist in the recognition of ICU-AW, especially for unconscious critically ill patients. Changes in TH and CSA of RF on the right side and the changes in TH of VI on both sides had good diagnostic accuracy for diagnosis of ICU-AW. However, considering the convenience and time efficiency, SOFA and APACHE II score are better options for early prediction of ICU-AW.

Variable
No ICU-AW n = 13 ICU-AW n = 24 t/χ 2 P value   Figure 4. Comparison of ROC curves among SOFA score, APACH II score, ΔTH day10, and ΔCSA day10 of muscles. (a) Comparison of changes in TH of biceps brachii muscle, vastus intermedius muscle and rectus femoris muscles; (b) comparison of changes in CSA of biceps brachii muscle, vastus intermedius muscle, and rectus femoris muscles; (c) comparison among changes in TH and CSA of muscles, SOFA and APACHE II score. BB, biceps brachii; RF, rectus femoris; VI, vastus intermedius; TH, thickness; CSA, cross-sectional area; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment.