Subjective and objective assessment of physical activity in multiple sclerosis and their relation to health-related quality of life

The amount of daily physical activity was reduced in PwMS compared to HC. This was expected from previous reports [28‐30] as well as from the theoretical conceptualization of PA as the behavioral correlate of disability in MS [31, 32]. Given that the vast majority of evidence on PA in PwMS is from US cohorts, this is also the first confirmatory study in German MS subjects.

When measured objectively with SWAmini, group differences were seen for all parameters except light PA and were most pronounced for higher intensities, indicating the generally lower participation of PwMS in strenuous physical activities. The three most distinctive parameters were step count, mean METs and active METs. This supports the description of changes in PA as (1) reduction in global walking mobility and (2) reduced activity-related EE [7, 33, 34].

Effect sizes ranged from 0.63 to 0.86, which is comparable to previous reports of large MS cohorts [28, 35‐37].

While daily step counts were similar to reports using other devices [36, 38, 39], the mean duration of MVPA activities observed in this study was higher than reported from US cohorts (MVPA < 23.5 and < 33.8 min/day in PwMS and HC) [35, 36] but lower than in other non-diseased populations [40]. Besides different recruitment bias and assessment devices, differences in activity lifestyle may also contribute.

An exploratory analysis of parameter inter-correlations (Additional file 2: Table S2) revealed that mean METs were most tightly associated with MVPA, active METs most strongly related to the amount of vigorous PA while step count showed only moderate correlations to both, mean METs and MVPA. Its lack of association with active METs and vigorous PA indicates that locomotor activity is predominantly performed within the low to moderate intensity segments.

This underlines that step count and EE capture complementary aspects of PA: subjects with gait or balance impairment may walk similar distances as HC but at lower speed (lower intensity) or – conversely - walk less efficiently, thus increasing the energy expended per step, which has indeed been shown in PwMS [41, 42].

Although our cohort included rather physically active PwMS, we observed a generally low amount of vigorous activities in PwMS with only one subject performing very vigorous PA, which impeded statistical comparison for this parameter. Concerning light activities, our results imply that this activity segment represents inevitable ADL activities, performed to a similar extent by HC and PwMS. It should further be noted that the group differences in PA relate to a segment of only about 10 min per hour, while both groups spend the largest part of their wake-time physically inactive. According to the MET definition of light PA, this may comprise low-intensity locomotor activity as well as sedentary time except sleeping which explains the seeming discrepancy to reports of increased sedentary time in PwMS [43, 44]. The effect of occupational status on PA was somewhat unexpected in its direction (tendency to even larger weekend – weekday difference in unemployed subjects) and small effect size, as others reported employment as a relevant factor for PA among PwMS [5, 35, 44, 45], however, in much larger cohorts. From our observation of a lower proportion of PwMS employed (Table 1), that even seems to decrease with higher EDSS scores (Additional file 3: Figure S1), we conclude that effects of occupational status and MS progression on daily PA may be difficult to distinguish or even counteract each other.

Group differences in PA could also be detected with the subjective IPAQ long form and were observed for the amount of MPA and MVPA, but not for the time spent walking. Convergent to SWAmini data, PwMS perceive themselves as (and according to SWAmini “are”) less engaged in higher intensity physical activities at group level. This is remarkable given the small group size and previous reports on a tendency for unfit (otherwise healthy) individuals to overestimate their participation in MVPA activities by up to 37% [46]. The questionnaire was easily applicable in our cohort and data exclusion or truncation rules had to be applied at similar rates in HC and PwMS.

With respect to IPAQ walking, it has to be considered, however, that IPAQ data refer only to reported bouts of over 10 min, although most locomotor activity, i.e. steps accumulated during the day, is likely to occur in shorter bouts. When relating results from subjective to objective assessment, we therefore subsumed reported walking activity within moderate PA.

The observed convergence for this matched parameter pair (r .507) was higher than expected from previous results obtained with other means of objective and subjective PA assessment in PwMS [36, 39, 47‐49]. Similarly, the relation of IPAQ PA levels to selected SWAmini PA parameters roughly converge with commonly used cut-offs to detect low and highly active lifestyles by step count or mean METs (step count low level < 5000 steps/day, moderate level 5000–7500 steps/day and high level > 7500 steps/day; mean METs: very low level < 1.2 METs, low level 1.2–1.4 METs, moderate level 1.4–1.6 METs and high level > 1.7 METs) [23, 38]. This supports the validity of this easily applicable classification, e.g. for screening purposes. The trend seen for group difference (p = 0.15) is likely due to small sample size. Only one person with MS was classified as highly active, probably due to high IPAQ Walking durations despite a not specifically high step count around 7000/day and mean METs of 1.5.

This part of analysis further suggested that step count rather differentiates high from moderate PA levels, which may apply early in disease, while mean METs (related tightly to MVPA) is more suited to distinguish between moderate and low PA level. This further implies that increase in PA is mediated by different types of activity according to PA levels and that different PA parameters may be suited to monitor MS populations of different symptom severity.

Another aspect of IPAQ with relevance for intervention design is the context of PA. From our data, it seems unlikely that increases of PA in PwMS will be added as leisure/sports activity, whereas supporting active transportation or integrate PA into subjects’ lifestyle [50, 51] seem more appropriate. In this respect, it is worth re-considering and exploring the role of occupational status in daily PA and, specifically, active transportation. With regard to behavioral interventions, these probably need to be designed and evaluated specific to cultural settings according to activity lifestyle differences, for example between U.S. and German populations.

SWAmini parameter correlations to EDSS were moderate in magnitude, which is similar to previously reported coefficients between −.34 and −.70 [33, 39, 52‐54]. Again, inspection of the parameter plots per EDSS score (Fig. 2) suggests reduced step count as an early feature in MS while decline in the daily amount of MVPA occurs beyond EDSS > 3, i.e. at stages with manifest decrease in walking mobility.

Corresponding IPAQ – EDSS plots revealed, that despite lower duration of MVPA in PwMS at group level, among those less severely affected, self-report of moderate activities was even higher than in HC. This paradoxical relation was even more pronounced for subjectively assessed walking duration and total EE. As discussed above, this may be interpreted as either a larger perceived exertion when physically active (despite unchanged or lower activity counts on “objective” assessment) or a systematic reporting bias in subjects aware of physical limitations, which has been reported in different populations [13]. Despite this and with only a small number of subjects remaining for IPAQ MPA and MVPA to EDSS analysis, correlations with symptom severity were similar to those observed with SWAmini. Further, IPAQ PA level assignment was related to EDSS scores. Both support the validity of IPAQ results in our PwMS cohort, although interpretations from individual absolute IPAQ results cannot be recommended.

With respect to HAQUAMS, we observed distributions of subscores similar to those reported in another German MS cohort [26] as well as expected correlations with EDSS.

However, the amount of daily moderate PA or daily step count were not related to HAQUAMS sum score, but seemed only related to the perceived impact of MS on lower limb function. This result was irrespective of the means of PA assessment (Tables 3 and 4) and adds to previous findings that PA in PwMS is only indirectly related to hrQoL, most likely mediated through physical function [55], although data on this topic are not abundant. One interventional study reported improved HAQUAMS total scores with intervention along with improvements in lower limb coordination [56]. Vice versa, decreased hrQoL as such seems unrelated to decrease in PA but related to a multitude of not directly disease-related factors [55, 57‐61]. This implies, that interventions aiming to increase PA should target walking ability and physical limitations of the lower limbs, while the quantitative assessment of lower limb function may be a potentially useful surrogate of PA for intervention monitoring.

Our findings, in line with previous reports, support the notion that both objective and subjective PA assessment may be appropriate depending on the purpose and resources of the study [39], a major limitation of both being the inherent variability in PA that requires long-term recording over several days. Further, specific limitations of different devices - questionnaires or activity monitors - have been studied in different populations [18, 42, 62, 63] and rather point to an approximation of “true” PA from different perspectives instead of one golden assessment standard. Thus, individual PA parameters may rather be interpreted as reflecting a subject’s activity level than as meaningful quantitative parameters per se. Accordingly, clinically meaningful differences as determined for step count, for example, are expectedly large given the generally high standard deviations within groups of healthy subjects (here: 2397/day). It seems therefore justified to evaluate the validity of more amenable and standardized quantitative performance measures as predictors of PA in PwMS. Our data add evidence to previous findings that suggest measures related to walking and lower limb function as most promising candidates.