Cardiorespiratory response to early rehabilitation in critically ill adults: A secondary analysis of a randomised controlled trial

Introduction Early rehabilitation is indicated in critically ill adults to counter functional complications. However, the physiological response to rehabilitation is poorly understood. This study aimed to determine the cardiorespiratory response to rehabilitation and to investigate the effect of explanatory variables on physiological changes during rehabilitation and recovery. Methods In a prospectively planned, secondary analysis of a randomised controlled trial conducted in a tertiary, mixed intensive care unit (ICU), we analysed the 716 physiotherapy-led, pragmatic rehabilitation sessions (including exercise, cycling and mobilisation). Participants were previously functionally independent, mechanically ventilated, critically ill adults (n = 108). Physiological data (2-minute medians) were collected with standard ICU monitoring and indirect calorimetry, and their medians calculated for baseline (30min before), training (during physiotherapy) and recovery (15min after). We visualised physiological trajectories and investigated explanatory variables on their estimated effect with mixed-effects models. Results This study found a large range of variation within and across participants’ sessions with clinically relevant variations (>10%) occurring in more than 1 out of 4 sessions in mean arterial pressure, minute ventilation (MV) and oxygen consumption (VO2), although early rehabilitation did not generally affect physiological values from baseline to training or recovery. Active patient participation increased MV (mean difference 0.7l/min [0.4–1.0, p<0.001]) and VO2 (23ml/min [95%CI: 13–34, p<0.001]) during training when compared to passive participation. Similarly, session type ‘mobilisation’ increased heart rate (6.6bpm [2.1–11.2, p = 0.006]) during recovery when compared to ‘exercise’. Other modifiable explanatory variables included session duration, mobilisation level and daily medication, while non-modifiable variables were age, gender, body mass index and the daily Sequential Organ Failure Assessment. Conclusions A large range of variation during rehabilitation and recovery mirrors the heterogenous interventions and patient reactions. This warrants close monitoring and individual tailoring, whereby the best option to stimulate a cardiorespiratory response seems to be active patient participation, shorter session durations and mobilisation. Trial registration German Clinical Trials Register (DRKS) identification number: DRKS00004347, registered on 10 September 2012.


2/18
"First, we aimed to analyse physiological variables from before, during and after rehabilitation to describe both the response to rehabilitation and to estimate recovery. Second, we aimed to characterise effects of a-priori selected explanatory variables on cardiorespiratory reactions from before to during rehabilitation and from before to after rehabilitation." P2, L37: would you please add the information of the seven different session types, with increasing activities. Please use a stringent reporting.
We are unfortunately limited by the allowed word count within the abstract (maximal 300 words, initially 292 words). We therefore cannot add all explanatory variables. Moreover, to add only 'session type' would unduly elevate the variable above the others. Our hypotheses about the main drivers of the cardiorespiratory response to early rehabilitation also included mobilisation level, exercise modality and duration. These explanatory variables should therefore be considered as important as session type. It is also important to note, that the seven session types are not reflecting increasing activities. They are based on commonly used, combined interventions in clinical practice. For example, early mobilisation is often combined with respiratory management, because sitting on the edge of bed might help to clear secretions or because deep-breathing exercises might calm patients after an effort to sit up. More detailed information on the seven session types is provided within the supplement (S1 File).
However, we concede that the term rehabilitation is used differently across the world and that some clarity of our concept might be needed. While our manuscript includes a definition of rehabilitation (page 17, lines 348-349), we agree that the concept must be clear within the abstract. We therefore slightly changed the suggested sentence to include the pragmatic study design and some type of interventions within the abstract (page 2, lines 37-38): "…, we analysed the 716 physiotherapy-led, pragmatic rehabilitation sessions (including exercise, cycling and mobilisation)." P2, Abstract: I am fan of clear structures, e.g. using always the same order of reported outcome parameters: MV, VO2, MAP. Within the sentences, this order changes. Would you please re-arrange the order, and keep it? See also lines 153-156.
We have amended the order throughout the manuscript to 'HR, MAP, MV, SpO2 and VO2' when it made sense to do so. For example, we did not change the order, when the magnitude of the change was discussed (e.g., page 10, lines 246-247), but changed tables 2 and 3 to the same order. When one variable was not discussed we left this variable out, for example, we wrote 'HR, MV and VO2'. P2, Abstract: I guess all results are mean differences? If so, please add this information.
Indeed these are mean differences, we added this description to the results in the abstract (page 2, lines 49-51): "Active patient participation increased MV ( Yes, the reviewer's assumption is correct. Our reasoning is as follows: This is a secondary analysis of an RCT that did not show any significant or clinically relevant differences in the prespecified primary and secondary outcomes. The two groups' intervention did therefore not affect outcomes. In consequence the two treatments can be considered equal, which allowed us to combine both groups into one population to answer our research questions. The current dataset therefore includes the whole original population using data from all trial participants who had at least one rehabilitation session. We slightly revised the manuscript to clarify this part (page 5, lines 121-125) "No significant differences were found in the primary or secondary outcomes with the exception of improved mental health six months after hospital discharge for the experimental group [19]. Consequently, this secondary analysis considered the two randomised groups as one population using data from all trial participants with at least one rehabilitation session.
Nevertheless, we did not ignore the allocated intervention within our analysis. Again, the reviewer is correct that the intervention group had significantly longer rehabilitation sessions. Also, only the intervention group cycled. However, these factors are accounted for within our analysis by the explanatory variables 'session duration' and 'session type'. We decided apriori which explanatory variables to include within our model. These variables were based on clinical reasoning and earlier research to limit confounding (also see revised lines 196-197 based on comments of reviewer #2). During this process we considered a variable 'randomisation group'. However, we discarded it to avoid multicollinearity within the model 4/18 and because we were rather interested on how the variables themselves affected the cardiorespiratory response (e.g. 'session duration').
Nevertheless, based on the reviewer's comments, we performed a sensititvity analysis that includes the variable 'randomisation group' in our model. Compared to our main analysis, the estimates barely changed for MAP, MV and VO2. While we found a significant effect for 'randomisation group' for HR ('during': mean difference 1.47bpm (0.17 to 2.78), p=0.029; 'after': 2.1bpm (0.79 to 3.41), p=0.002), the estimates for our chosen explanatory variables changed only slightly when compared to our main analyses. Most importantly, our results and conclusions do not change but rather reinforce our earlier reasoning to exclude this variable. We therefore limit the results of this sensitivity analysis to this letter, but did not add them to the supplementary material.

L125 "physiological data (n=35)" You don't mean that you analyzed 35 data, but data of 35 patients, right?
This is correct. We revised this to (page 5, lines 125-127): "A preliminary safety analysis of the physiological data of the first 35 subjects indicated a moderately increased workload with increased heart rate and oxygen consumption but stable oxygen saturation from before to during rehabilitation [20]."

L126: please insert a line break between "…physiotherapy [20]." and "The local…"
This has been implemented as suggested.

L130: please add "Population", and a line break. I guess you mean "equal or older than 18 years"?
We added a new heading 'Population' (line 132). Accordingly, we deleted 'population' from the heading 'study design'. We further revised the sentence as follows (page 5, lines 133-134): "Participants were ≥ 18 years old, functionally independent before ICU admission and expected to remain ventilated for ≥ 72 hours."

10/18
Section results is fine. The tables are hard to read, but I have no idea how to improve readability. The "examples for interpretation" are really helpful.
As described previously, we changed the order of the column headings to HR, MAP, MV and VO2 to give a clear structure throughout the manuscript. We did not perform any other changes as readability should increase with article formatting.

L316-319: would you please summarize the main results in their plain meaning, eg. "complex exercise and mobilisation led to increased VO2", or else
We concede that tables 2 and 3 are complex. We believe that the examples for interpretation and the highlighted relevant 95% confidence intervals will help readers to interpret results. To further increase understanding, we give a full summary here (based on 95% confidence intervals) along with a brief summary within the manuscript.
Full summary for 'during rehabilitation' • HR: increased by 'session duration' (per each additional minute of therapy), active participation (when compared to passive therapy), and by sedatives (when compared to none). • MAP: increased by female gender (when compared to male), mixed or active participation (when compared to passive therapy), decreased by vasoactive drugs (when compared to none) and 'session duration' (per each additional minute of therapy) • MV: increased by active participation (when compared to passive therapy), decreased by female gender (when compared to male) and by 'session duration' (per each additional minute of therapy) • VO2: increased by active participation (when compared to passive therapy), decreased by age (per each additional year of age), female gender (when compared to male) and 'session duration' (per each additional minute of therapy) Full summary for 'after rehabilitation' • HR: increased by age (per each additional year of age), session types 'mobilisation' and 'complex exercise and mobilisation' (when compared to 'exercise) and by sedatives (when compared to none), decreased by mobilisation level 'edge-of-bed' and 'out-of-bed' (when compared to 'in-bed) and by daily SOFA score (per each additional point) • MAP: decreased by BMI (per each additional unit), daily SOFA score (per each additional point) and by vasoactive drugs (when compared to none) • MV: decreased by female gender (when compared to male) • VO2: increased by mobilisation level 'edge-of-bed' (when compared to 'out-of-bed'), decreased by female gender (when compared to male) and by vasoactive drugs (when compared to none) We newly summarised this in the manuscript as follows (page 16, lines 333-338): Yes, there are several research questions arising from our analysis. Most importantly, we found that 'treatment modality' and 'session duration' were the major drivers of cardiorespiratory changes during early rehabilitation. However, a sufficient cardiorespiratory response does not automatically translate into an adequate neuromuscular response or functional benefits. Consequently, future trials should investigate if any of these factors can improve patient-centred, functional outcomes. For example, shorter sessions were associated with an increased cardiorespiratory response. We propose to investigate whether shorter, more frequent sessions improve functional outcomes, when compared to only one session per day.
We revised the manuscript to include this more clearly (page 17, lines 363-365):

"This strategy was associated with improved 3-month outcomes after stroke [27] and should be investigated in future randomised controlled trials."
Patient participation (treatment modality) is dependent upon sedation but also on patients' motivation or their perceived rate of exertion. There are a few studies who specifically investigate the impact of sedation on patient participation and functional outcomes [1], but patient fatigue is often a limiting factor [2]. A qualitative study that explores enabling factors and barriers to being active while critically ill might give insights into how future clinical trials should be shaped to achieve intensity-targets.
Additionally, we suggest to perform prospective studies that explore the optimal target for cardiorespiratory parameters and how this relates with patients' perceived rate of exertion as well as neuromuscular activation (proof-of-concept study). For example, we chose a conservative cut-off of 10% for our analysis. However, this might not be sufficient to elicit a neuromuscular benefit. We would therefore propose to investigate the feasibility of different physiological targets (e.g., 10% versus 20% versus 30%) preferably using oxygen consumption. (Rationale: oxygen consumption seems the most important variable, when talking about exercise that is defined as a planned, structured, repetitive bodily movement produced by skeletal muscles that results in energy expenditure and aims to improve or maintain one or more components of physical fitness [3]). Research questions would include: • Can patients achieve and sustain these intensities?
• Are higher intensities associated with improved muscle activation (for example when measured with electromyography)?

12/18
• If these physiological targets prove feasible (and remain safe as within our study), what is the effect of these different targets on functional outcomes?
We revised the manuscript to express some of these thoughts (page 19, lines 423-425): "Future trials should therefore investigate the feasibility and efficacy of different physiological training intensities as well as their association with neuromuscular activation and patients' perceived rate of exertion." Finally, our findings should be validated in a prospective study with prospectively planned hypotheses (also see revision based on comments of reviewer #2: lines 418-419).
Not part of the review and just a suggestion: You started very early, at day 2, and the muscle loss might not be severe in early days. I wonder if there are any differences in cardiorespiratory responses to PT's rehab when delivered within early <7 days or late ≥7 days?
Our analysis accounted the time from ICU admission to each session (median of 9 days with an interquartile range from 4 to 20 days, which includes the 7 day difference). The estimated effect on cardiorespiratory parameters was very minimal, clinically irrelevant and not statistically However, cardiorespiratory parameters are a poor surrogate measure for muscle mass, thus to infer about potential neuromuscular effects, a study using electromyography might be necessary.
Very fine work, thank you.

Reviewer #2
Thank you for the opportunity to review this paper. The authors present the cardiorespiratory response in critically ill adult patients in response to various early rehabilitation interventions. This paper is well written and contributes novel data to the field, highlighting that early rehabilitation is safe. The authors adequately identify limitations to the study. The heterogeneity in both the population and interventions and the impact on the conclusions is mentioned.
We thank the reviewer for taking the time to review our manuscript and are grateful for the constructive and valuable feedback. All comments have been answered below.
I have detailed a few minor points below: 1. It may be valuable to clarify/better define the "explanatory" variables. How were these decided on or chosen?
We agree, that this is a highly relevant point. As mentioned within the comments of Reviewer #1, we extensively discussed parameters within the research team and weighted them against the previous evidence. We modified the manuscript accordingly (page 7, lines 196-197): "Explanatory variables were prospectively determined using extensive clinical reasoning and previous evidence to account for confounders. They included…" 2. If understood correctly, the clinically relevant cutoff of >10% is related to safety; however further justification as to why this was chosen and not 20% is required. This will perhaps be important when weighing up safety versus adequate intensity to bring about physiological change. In the discussion it seems as if this clinically relevant response is linked to response to exercise and not safetyplease clarify.
Indeed, we chose the 10% cut-off based on safety recommendations that generally consider an increase of >20% as an adverse event. To the best of our knowledge, there is currently no guidance on the dose and safety of any training intensity available in the literature. We therefore chose the more conservative cut-off that was considered safe in previous studies. This is clearly a limitation, that we acknowledge within the manuscript.
As the reviewer mentions, 20% might be a more adequate intensity to elicit a neuromuscular response potentially improving functional outcomes. To this end, we performed a sensitivity analysis using a 20% cut-off (S2 File: S2 Table), whereby 1 out of 10 sessions was affected. Adverse events within our trial were not based on fixed cut-offs, but rather on individually set limits by the treating physician. Overall, we reported only few, transient adverse events. Accordingly, we believe that a 20%-threshold might be safe and should be considered in future trials as a potential training intensity. We further included this in the manuscript which has been revised to: Additionally, it is important to note that the main aim of this study was to describe and explore the physiological response (changes) by early rehabilitation. The response to exercise is therefore an important topic within our discussion. To avoid any confusion between terms, we further substituted 'clinically relevant change' to 'clinically relevant variation' throughout the manuscript.
Finally, we incorporated subheadings that align with the results section to clarify that the 10% cut-off belongs to the safety analysis.

Stronger motivation is needed for not adjusting for multiple testing is needed.
We concede that the chance of false-positive findings is increased without adjustment for multiple testing. However, this is a secondary, hypothesis-generating analysis and not a primary hypothesis-testing study (null versus alternative hypothesis). Adjustments for multiple testing might therefore discard potentially useful observations [4,5]. We therefore do not think that multiple testing is appropriate for our type of analysis. However, we concede that we need to clearly state this limitation to readers.
We therefore revised the manuscript as follows: Page 19, lines 418-419: "These results therefore provide important information for future trials, but need to be validated in prospective studies." Table 1, please could you clarify what is meant by "Time from ICU admission to start of each session (days)"?

In
We investigated 716 rehabilitation sessions. The 108 participants had a median of 3 [2][3][4][5][6][7][8] sessions (line 226). Accordingly, the time to each session varied in respect to ICU admission. For example, the first session could have been on the first day after admission, the second on the second day and the third session on the fifth day. This variable accounts for these different time periods.
We renamed this variable to "Time from ICU admission to start of individual session (days)" and added a footnote to better explain the variable (Page 10, lines 234-235): " a number of sessions varied between patients, this variable takes into account the time from ICU admission to the start of each, individual session in the individual patient."

While recommendations are made for clinical practice in terms of monitoring and
recovery, what may be more helpful to clinicians and advancing the cause of early rehabilitation in the ICU is know how these physiological responses link to various outcomes. As highlighted by the authors, there is a gap in terms of adequate prescription (FITT) in the critically ill patients. Such data sets, could help shed light on factors such as the frequency and intensity of rehabilitation interventions to effect change on outcomes. Authors highlight these results enhance clinical decision making around frequency and intensity, however, I think without looking at outcomesthis enhancement is limitedit may be safe, but there is no way to know it is effective with the data provided.
We agree with the reviewer that studies linking physiological responses to functional outcomes are needed. However, before this research can be conducted safety should be established. Additionally, we would need proof-of-concept studies, for example, is a higher intensity associated with a better neuromuscular response? However, our data-set was primarily obtained to answer the question about the efficacy of two different interventions. This analysis would not be in line with our initial research questions and conclusions. We therefore advocate for future prospective studies. We integrated this within the manuscript (page 19, lines 423-425): "Future trials should therefore investigate the feasibility and efficacy of different physiological training intensities as well as their association with neuromuscular activation and patients' perceived rate of exertion." We further added this limitation to the manuscript (page 19, lines 419-420): "Fourth, while our results inform clinical decision-making on the intensity and duration of early rehabilitation, they cannot establish the effect on functional outcomes."