Statistical analysis

In the previous Catastim 1 trial [24], MRC muscle strength was evaluated for hip flexion: mean changes in MRC score from the first postoperative day until ICU discharge were −0.80 points [standard deviation (SD), 0.70 points] in the NMES group (n = 20) versus −1.75 points (SD, 1.52 points) in the control group out of a maximum MRC score of 5 points (n = 20) (P = 0.015). Based on these results, sample size calculation with Stata yielded 25 patients per group at a two-sided significance level of α = 0.05 and 80 % power. Assuming a dropout rate of 20 %, 60 patients needed to be randomized.

All data was analyzed according to the intention-to-treat principle with no imputation for any missing data. Normality was assessed with Kolmogorov-Smirnov and Shapiro-Wilk statistics. If quantitative variables were normally distributed, they were expressed as mean ± SD or 95 % confidence interval (CI). If they were not normally distributed, median (range) was indicated.

In previous muscle ultrasound studies, linear regression analyses were performed [7, 27, 28]. Similarly, we presented results of actual raw data (Additional file 2 and Table 7) and results based on linear regression modeling (Tables 2, ​,3,3, ​,4,4, ​,5,5, ​,66 and Additional file 3). Linear mixed models were used to assess the treatment effects while accounting for the repeated measurements per patient. Thus, the random effect ‘patient’ was defined in each model. Linear mixed models were calculated for mean MLT, mean MRC score of all muscle groups, quadriceps mean MRC score and grip strength. Fixed effects of interest were postoperative day or study day, NMES/control group, interaction between NMES/control group and postoperative day, and daily fluid balance in the ICU. Fixed effects were either considered as quantitative or categorical variables: for example, the postoperative day or the daily fluid balance at the ICU were quantitative variables. The study day (preoperative day, first postoperative day, ICU discharge, hospital discharge) or the group (NMES group, control group) were categorical variables. By definition the results presented as linear mixed models are derived statistics by linear regression modeling (‘proc mixed’ in SAS software version 9.4; SAS Institute Inc., Cary, NC, USA).

Linear mixed model for MLT from the first postoperative day for a maximum of 14 postoperative days (53 patients, 183 observations)

MLT muscle layer thickness, CI confidence interval, NMES neuromuscular electrical stimulation, ICU intensive care unit

Linear mixed model for MLT on four important study days (53 patients, 141 observations)

MLT muscle layer thickness, CI confidence interval, ICU intensive care unit, NMES neuromuscular electrical stimulation

Linear mixed model for mean MRC of all muscle groups^a from the first postoperative day for a maximum of 14 postoperative days (51 patients, 220 observations)

Days of ICU and hospital discharge, where no NMES was applied anymore, were excluded in this model

The linear mixed model for mean MRC in Table 4 reads as follows:

MRC = 4.10 + (0.02 × postoperative day) - (0.45 × NMES group) + (0.09 × postoperative day × NMES group)

0.02 is the slope of MRC time variation in the control group, which is the reference group: for each postoperative day, MRC increases by 0.02 points (95 % CI, −0.02 to 0.05 points) in the control group (P = 0.40)

0.45 represents the lower starting point in the NMES group on the first postoperative day before the NMES intervention began: on the first postoperative day, MRC was about −0.45 points (95 % CI, −0.88 to −0.03 points) lower in the NMES group than in the control group (P = 0.04)

0.09 is the slope of MRC time variation in the NMES group: the slope of MRC time variation is 4.5 times higher than the slope in the control group (P = 0.002)

MRC Medical Research Council, CI confidence interval, NMES neuromuscular electrical stimulation

^aAccording to the MRC scale [33], mean MRC score ranges from a minimum of 0 to a maximum of 5 points

Linear mixed model for mean MRC of all muscle groups on four important study days (51 patients, 130 observations)

The linear mixed model for mean MRC in Table 5 reads as follows:

On the first postoperative day mean MRC was about −0.57 points (95 % CI, −0.78 to −0.36 points) lower than on preoperative day, which is the reference day (P < 0.001)

At ICU discharge mean MRC was about −0.27 points (95 % CI, (−0.48 to −0.06 points) lower than on preoperative day (P = 0.01)

At hospital discharge mean MRC was not different from mean MRC on preoperative day (P = 0.43)

On preoperative day, at the first postoperative day, at ICU discharge and at hospital discharge there were no differences in mean MRC between the NMES and control group (P = 0.92)

MRC Medical Research Council, CI confidence interval, ICU intensive care unit, NMES neuromuscular electrical stimulation

^aAccording to the MRC scale [33], mean MRC score ranges from a minimum of 0 to a maximum of 5 points

Linear mixed model for grip strength measured by hand dynamometer on four important study days (49 patients, 127 observations)

MRC Medical Research Council, CI confidence interval, ICU intensive care unit, NMES neuromuscular electrical stimulation

Functional outcomes on four important study days (12 patients)

ICU intensive care unit, FIM Functional Independence Measure, MCS-12 mental component score of the SF-12, PCS-12 physical component score of the SF-12

^a P for change from preoperative day to ICU discharge: 0.002

^b P for change from ICU discharge to hospital discharge: 0.002

Because of bandages to protect arterial catheters, patients were not always able to perform every movement in all 12 muscle groups. If there were missing values for muscle groups, mean MRC score was calculated for the muscle groups that could be assessed. Similarly, due to armboards because of radial arterial catheter, patients were not always able to perform the hand dynamometer test on both sides. Thus, either the right or left grip strength was chosen for statistical analysis per patient according to the lowest number of missing values over all study days on each side. If the number of missing values was equal for right and left grip strength for a patient, either the right or left grip strength was chosen according to the highest mean grip strength over all study days on each side.

In addition, intraoperative fluid balance, changes over time in MLT or mean MRC score were compared between both groups with independent-groups t tests or Mann-Whitney U tests as appropriate. Change in MLT was correlated with the cumulative fluid balance in the first three postoperative days with Pearson product-moment correlation.

The secondary functional outcomes were evaluated in a sensitivity analysis for patients seen both on preoperative day and at hospital discharge (n = 12) in order to evaluate whether preoperative functional levels could be regained at hospital discharge. Changes over time in secondary functional outcomes were analyzed with Wilcoxon signed-rank tests. Changes in functional outcomes were compared between both groups with Mann-Whitney U tests. The patient satisfaction related to the intervention was evaluated in 42 patients with Pearson’s chi-squared test or Fisher’s exact test as appropriate.

Significance level was set at 0.05. All P values were two-tailed. For statistical analysis, SPSS version 22 (IBM Corp., Armonk, NY, USA) and SAS version 9.4 were used. For figure construction, GraphPad Prism version 6.0 (GraphPad Software, Inc., La Jolla, CA, USA) was used.