Twenty-six patients commenced exercising at the OTT between January 2012 and November 2019 without active acquisition. No adverse events occurred. Mortality data were not accessible. However, considering the low 5-year relative survival rate of MM patients  (www.krebsdaten.de; RKI 2021: Zentrum für Krebsregisterdaten > Krebsarten > Multiples Myelom), the drop-out rates in the study population which range between 25.4 percent (drop-out within 6 weeks of exercise), 30.2 percent (drop-out within 12 weeks of exercise/less than 12 sessions) and 50 percent (less than 36 sessions) might have been affected by mortality. Additionally, previous studies systematically excluded patients at increased risk of bone fractures . In the present study, at baseline a high percentage of patients exhibited disseminated bone lesions (69.2 percent) and bone fractures (42.3 percent) not at risk for instability (Table 1). The impact of these factors on admission rate and exercise adherence remains speculative due to the real-life nature of the data and the missing data of determinants for early dropout. In general, safety and feasibility of exercise therapy with MM patients at the OTT are basically given.
4.2 Adherence and Efficacy
Absolute indicators of exercise adherence such as participation period and total attendance by week and session are indicators for sustained participation or durability. Relative values indicate exercise frequency. Most strikingly the level of BMI, is negatively associated (p < 0.01) with exercise frequency by week. A high BMI might indicate psycho-social as well as physical barriers to regular physical activity, such as a generally rather unfavorable health-related lifestyle, lower motivation, and affinity for exercising as well as increasing risks for serious systemic secondary diseases . In contrast, absolute indicators of exercise adherence are not-significantly correlated with BMI. Additionally, Figure 3 illustrates extreme outliers with three BMI values above 30 and two under 20 accounting for one third of the sample size while two thirds of the values scatter around a BMI of 25, suggesting a statistical distortion effect. A high diarrhea (single-item) and a high role functioning scale score (multi-item; e.g. "Have you been restricted at your work or at other daily activities?") reflect the disability of MM patients and might pose additionally obstacles to a long-term maintenance of an exercise routine as indicated by the negative correlations with absolute adherence values. Significant negative correlations found for moderate work-related activity and absolute indicators of exercise adherence are most likely caused by an overrepresentation of zero values as only six out of 15 patients exhibit available data on that item. Moreover, the limited informative value oft the GPAQ, particularly in patiens with cancer has already been reported . The positive correlation of fatigue and participation period, with respect to existing research, might depict an interesting and promising finding. Cancer-related fatigue is often perceived as one of the highest burden of MM patients [21, 22] and can be mitigated by exercise more effectively than by drug therapy . A recently published meta-analysis confirms this effect in hemato-oncological entities . Accordingly, high fatigue burden and efficacy of reduction through exercise might positively impact long-term exercise motivation with a positive trend seen in the EC (N = 8).
The associations reported in the literature between bone fractures, major surgery, and low training adherence [21, 25] could not be supported due to non-significant results via grouping tests for surgery and bone fractures. Furthermore, the movement-reducing effect of stem cell transplantation [26, 27] could not be investigated due to the small sample size and the unequal distributed grouping variable (see Table 1).
The physical tests yielded hardly evaluable data also because of the heterogeneity of the performed strength tests among patients and MPs. Consequently, comparability was hardly producible. The application of a feasible and quick test like the handgrip strength test could solve this problem in future studies.
The most interesting and promising result of this investigation is the statistically significant change in physical functioning over a training period of 16 to 30 exercise sessions between MP1 and MP2 in the EC (N = 8). Existing interventional studies point in a similar direction regarding beneficial effects on physical functioning by exercising in MM patients [9, 14, 25, 27]. Nevertheless, there are several remarkable factors with respect to this result. First, the present exercise period between patients of the EC is very heterogeneous and significantly shorter than in existing interventional studies on exercise therapy in MM patients . Furthermore, the smaller dispersion between values of realized exercise sessions (range: 16 - 30 sessions) and observation period (range: 8 - 46 weeks) between MPs (see Figure 4), in the context of FITT criteria, might indicate higher importance of total session count or time criteria respectively over exercise frequency in real-life circumstances. Accounting for the small sample size of eight patients, either a very large effect size or factors of chance and positive selection bias lead to statistical significance. However, the fact that data from a reference sample show similar values for baseline physical functioning scores  contradicts the presence of a positive sample. Recently, we demonstrated a highly significant improvement of physical activity and functioning in patients with the precursor condition monoclonal gammopathy of undetermined significance under whole-body vibration exercise training .
Notwithstanding the promising nature of various findings, there are some limitations particularly resulting from the applied patient flow allocation to different cohorts and generally from the evaluation of real-life data. The allocation of patient flow to specific cohorts was carried out to enable cross-sectional and longitudinal data evaluation. Although the number of realized exercise sessions between measurements was not a relevant allocation criterion, the applied procedure resulted in an unintended incremental increase of mean exercise adherence by absolute and relative indicators among cohorts (Table 2), and therefore, most likely led to a positive selection bias. Moreover, the consecutive reduction in cohort sizes limits the statistical representation of the underlying total population of MM patients restraining the range of findings.
Additional limitations result from the real-life nature of the present study. Firstly, it cannot be precluded that a distortion of the patient selection already existed at the time of admission to the OTT. It is unclear whether all eligible patients were informed and basically had access to the OTT or whether the participating patients already possessed certain attributes that did not adequately reflect the characteristics of the underlying population (e.g. motivation, training affinity). Furthermore, there is no precise knowledge of possible confounders during exercise therapy such as disease or psycho-social burdens. Ultimately, defining statistical measures post-hoc generally leads to an incalculably size of alpha error, hence, increasing likelihood for coincidental findings. The findings should be further examined in confirmatory study designs.