Physical activity measurements in individuals with interstitial lung disease: a systematic review and meta-analysis

Description of studies

Details of the included studies are available in table 1.

Results of the search

The search yielded 15 407 citations and ended with 40 studies [3–5, 7–10, 21, 22, 29–59] from 49 citations (figure 1).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Flowchart of literature review. ILD: interstitial lung disease.

Hur et al. [60] and Hur et al. [8] used the same cohort data. All PA metrics reported by Hur et al. [60] were shown in the other study [8]. Thus, we excluded Hur et al. [60]. Three studies by Vainshelboim et al. [6, 7, 61] used the same participant data (Registration No. NCT01499745). A PA metric (overall PA calculated in metabolic equivalent of tasks (METs) min·week⁻¹) shown in two studies [6, 61] was not reported in the other one [7]. Therefore, we combined the two studies [6, 61] with Vainshelboim et al. [7] and excluded them [6, 61]. Two studies by Dale et al. [44, 62] used the same cohort data (Registration No. ACTRN12608000147381). We excluded the latter one [62]. Two reports by Bahmer et al. [5, 63] used the same participant data (Registration No. DRKS00006170) and Bahmer et al. [5] reported complete data at baseline. Thus, we excluded Bahmer et al. [63]. We excluded four studies that did not show the participant characteristics with ILD separately from other participants [64–67].

Included studies

Table 1 shows the characteristics of the included studies. Of the 40 studies, 11 studies [9, 21, 34, 35, 37, 46, 49–52, 58] were intervention studies, 13 studies [3, 7, 8, 22, 29, 30, 33, 36, 41, 43, 47, 54, 56] were cohort studies and 16 studies [4, 5, 10, 31, 32, 38–40, 42, 44, 45, 48, 53, 55, 57, 59] were cross-sectional studies. Of the 16 cross-sectional studies, one was a validation study of a PA questionnaire [8].

15 studies included patients with ILD due to any causes [8, 21, 30–34, 37, 46–48, 51, 53, 55, 59], 15 studies included only idiopathic pulmonary fibrosis (IPF) [5, 7, 9, 22, 29, 40–43, 45, 49, 50, 54, 56, 58], seven studies included only sarcoidosis [4, 35, 36, 38, 39, 52, 57] and three studies included another ILD subtype [3, 10, 44]. The total sample size of each study ranged from 13 [46] to 629 [57]. The mean or median age was more than 65 years in 26 studies (65%).

PA measurement

Of 40 studies, 31 (78%) studies used only activity monitors [3, 5, 10, 21, 22, 29–32, 34–45, 48–52, 55–59], seven (18%) used only questionnaires [7, 9, 33, 46, 47, 53, 54] and two (5%) used both [4, 8]. Thus, 33 (83%) of 40 studies used activity monitors and nine (23%) used questionnaires.

Table 2 summarises the activity monitor-based PA measurements. Of 33 studies, 14 studies used SenseWear Armband [3, 5, 21, 22, 35–37, 39, 40, 44, 49, 50, 55, 56], nine studies used ActiGraph [8, 10, 34, 38, 48, 51, 52, 58, 59] and four studies used Lifecorder [29, 32, 42, 45].

12 studies did not report the sensor location [21, 31, 32, 35, 37, 38, 40, 43, 50, 52, 57, 58], nine studies located the monitors at the arm (upper arm: seven studies [3, 22, 36, 39, 44, 55, 56]; not specified: two studies [5, 49]), three studies at the wrist [30, 34, 51], four studies at the waist [29, 42, 45, 59], two studies at the hip [41, 48], one study at the upper thigh [4], one study at the wrist and waist [8], and one study wrist or waist [10].

Of nine studies that used questionnaires, five studies used the International Physical Activity Questionnaire (IPAQ) (long form: one study [8]; short form: three studies [4, 7, 9]; not reported: one study [54]), four studies used the Human Activity Profile [46], the Rapid Assessment of Physical Activity Questionnaire [47], a part of frailty assessment by the Fried Frailty Criteria [33], and the Minnesota Leisure-Time Physical Activity Questionnaire short form [53] (table S1 in supplementary material 2).

Reporting quality of PA measurements

Table 3 summarises the reporting quality of PA measurements using activity monitors (33 studies). Accelerometer brand and model, the number of days of data collected, PA metrics, and the number of people not meeting wear-time criteria were well reported in most studies. However, epoch length, placement of the accelerometer (especially on the side of the body), the number of participants receiving the accelerometer, distribution method of the accelerometer, criteria for defining nonwear time, minutes requirement for a valid day, and the number of valid days needed were poorly reported (table S2 in supplementary material 2).

Risk of bias

Table S3 in supplementary material 2 shows the results of the risk of bias. The domains with the highest risks of bias were item 7 regarding the method of PA measurements and items 5 and 6 regarding confounding factors.

PA metrics

Of 40 studies, 35 studies (activity monitor only: 29 studies [3, 5, 10, 21, 22, 29–32, 35–45, 48–52, 55, 56, 58, 59]; questionnaire only: four studies [7, 46, 53, 54]; both: two studies [4, 8]) displayed at least one PA metric of patients with ILD. The PA metrics used were heterogeneous (tables S4 and S5 in supplementary material 2).

Of 29 studies that used activity monitors, 28 studies (97%) reported steps [3–5, 8, 10, 22, 29–32, 35–42, 44, 45, 48–50, 52, 55, 56, 58, 59], 16 studies (55%) MVPA [3, 5, 8, 10, 21, 22, 31, 35, 37, 41, 44, 48, 50, 51, 55, 56], eight studies (20%) TEE [3, 21, 22, 31, 35, 36, 44, 45], four studies (10%) AEE [8, 22, 41, 45] and nine studies (23%) ST [8, 10, 21, 31, 41, 43, 50, 55, 56]. Definitions of MVPA and ST were heterogeneous (table S4 in supplementary material 2). For example, seven of 16 studies defined moderate intensity of PA as >3 METs [5, 22, 31, 37, 44, 55, 56], two studies defined it as >2.5 METs [3, 35] and others used acceleration magnitude [8, 51] or EE [41] for defining MVPA.

Of nine studies that used questionnaires, four studies presented IPAQ overall EE (MET min·week⁻¹) [4, 7, 8, 54], two studies IPAQ walking EE (MET min·week⁻¹) [4, 7] and ST (min·day⁻¹) [7, 8], and one study IPAQ MVPA EE (MET min·week⁻¹) [8] (table S4 in supplementary material 2). Other metrics measured by questionnaires are described in table S5 in supplementary material 2.

Estimation of PA levels in steps, MVPA, TEE and ST

The pooled mean (95% CI) of steps was 5215 (4640–5791) steps·day⁻¹ (I²=97 (97–98) %) (figure 2a). Subgroup analyses found that people with IPF or ILD took fewer steps than those with sarcoidosis. Activity monitors worn on the wrist or upper arm showed higher steps than those worn on the waist or lower extremity. Additionally, people with lower FVC, D_LCO, 6MWD and LTOT exhibited fewer steps (figures S1–S8 in supplementary material 3). However, I² only slightly improved in subgroup analyses by the ILD subtype.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Forest plots of estimation of overall means of a) steps (steps·day⁻¹), b) moderate to vigorous physical activity (MVPA) (min·day⁻¹) and c) sedentary time (ST) (min·day⁻¹). Weight is calculated by the random-effects model. Wrist and waist refer to the activity monitor location. C: control; I: intervention.

The pooled mean of MVPA was 82 (58–106) min·day⁻¹ (I²=99 (99–99) %) (figure 2b). Subgroup analyses revealed that people with sarcoidosis spent more time in MVPA than the other ILD subtypes (figures S9–S17 in supplementary material 3). MVPA measured by SenseWear (91 (61–121) min·day⁻¹) appeared to be higher than by ActiGraph (49 (3–96) min·day⁻¹), but it was not significant. Patients with lower FVC or LTOT spent a shorter time in MVPA than those with preserved FVC or without LTOT (all p<0.05). A mean of MVPA, defined as 2.5 METs or more, was approximately 2.5 times longer than that defined as 3.0 METs (152 (123–181) min·day⁻¹ versus 67 (51–82) min·day⁻¹) with a slight improvement in heterogeneity.

The pooled mean of TEE was 3574 (1684–5464) kcal·day⁻¹ (I²=100 (100–100) %) (figure S18 in supplementary material 3). We did not perform a subgroup analysis by FVC because no study was classified into the low FVC group. Activity monitors worn on the upper extremity showed higher TEE than those worn on the waist or lower extremity (figures S19–S25 in supplementary material 3). Heterogeneity was improved only in subgroup analysis by the ILD subtype.

The pooled mean of ST was 605 (323–887) min·day⁻¹ (I²=100 (100–100) %) (figure 2c). Types of activity monitors showed significant differences in ST. People with worse FVC or D_LCO exhibited longer ST than those with better FVC or D_LCO. ST significantly differed between studies that used different definitions of ST (figures S26–S34 in supplementary material 3). There was no improvement in I² in all subgroup analyses.

Sensitivity analysis

Subgroup analyses by reporting quality of PA measurements found that studies with low reporting quality reported longer MVPA than those with adequate reporting quality (p=0.02) (figures S35–S38 in supplementary material 4). Substantial heterogeneities were observed in all PA metrics.

After omitting two studies [21, 22], the pooled mean value of TEE was 2130 (1847–2412) kcal·day⁻¹ (figure S39 in supplementary material 4). I² was unchanged (96 (94–97) %). Subgroup analyses showed similar results (figures S40–S47 in supplementary material 4).

Reference data of healthy controls

Of 40 studies, seven studies recruited healthy controls (tables S6 and S7 and figures S48 and S49 in supplementary material 5). The pooled mean of steps from six studies was 10 167 (8433–11 901) steps·day⁻¹ (I²=88 (76–94) %). The pooled mean of TEE from three studies was 2618 (2505–2730) kcal·day⁻¹ (I²=15 (0–91) %). The means (sd) of MVPA in three studies were 261 (118), 132 (72) and 86 (8) min·day⁻¹. No study reported ST.

Quality of the body of evidence

The quality of the body of evidence on steps, MVPA, TEE and ST was very low, mainly due to severe inconsistency and imprecision (table S8 in supplementary material 6).

Discussions

This systematic review revealed that 1) measurement procedures varied tremendously between studies, 2) reporting quality of PA measurements was poor in most studies, 3) types and definitions of PA metrics were heterogeneous and influenced the PA metrics values, and 4) use of PA questionnaires is limited in patients with ILD. Additionally, there was very low-quality evidence in the pooled means of steps, MVPA, TEE and ST. Therefore, clinicians and researchers should improve the quality of PA measurements.

Reporting quality of PA measurements

Four systematic reviews have assessed the reporting quality of PA measurements using accelerometers in the general population [15], cancer survivors [68] and patients with chronic heart failure [69] or COPD [70]. Although direct comparisons of our results to these reviews are difficult due to the differences in methodology of reporting quality assessment, we selected the same tool used in two reviews [15, 69], enabling us to compare the reporting quality in patients with ILD to other populations.

More than 50% of studies included in this review failed to report six of 12 items related to data collecting and processing. Specifically, 88% of the included studies failed to report the criteria for defining nonwear, compared with 69% in general populations [15], 49% in cancer survivors [68] and 80% in heart failure [69]. Moreover, included studies also failed to report how accelerometers were distributed to participants (79%), placement of accelerometer (76%) and epoch length used (73%). The percentages are worse than those in the general population (69, 51 and 36%) [15], cancer survivors (46, 7 and 52%) [68] and heart failure (0, 64 and 60%) [69]. In contrast, 48% of the included studies did not report the number of valid days needed and 42% failed to report how many minutes were needed to be considered a valid day. The percentages are similar to those in the general population (52 and 50%) [15], cancer survivors (43 and 38%) [68] and heart failure (76 and 78%) [69]. Burtin et al. [70] revealed that only 37 of 110 (34%) studies with COPD patients fulfilled the following minimal preferred methodologic quality of PA assessment: measurement period ≥7 days; minutes needed to be considered a valid day ≥8 h·day⁻¹; ≥4 consecutive or nonconsecutive valid days; and invalid days excluded from analysis. In our review, only 36% met the criteria. We are reluctant to use these results to claim that the reporting quality in studies with ILD is similar or inferior to that in other populations. However, we believe that reporting on PA measurements should be improved because these factors are crucial for replicating and comparing studies. Therefore, we encourage clinicians and researchers to report data collecting and processing methods following the checklist by Montoye et al. [15], as shown in table 3 and table S2 in supplementary material 2. We have developed a template for adequate reporting quality (table S9 in supplementary material 7), which can be used to report required information as supplementary materials.

Risk of bias

The studies included in this review recruited well-defined patients with ILD using objective measures following international guidelines. In contrast, the included studies did not adequately define and deal with confounding factors. Difficulties in recruiting enough patients with ILD may be a possible reason for this, because a large sample size is required to control confounding factors. Furthermore, most studies failed to measure PA validly and reliably. Therefore, defining and dealing with confounding factors and PA measurements are at high risk of bias, leading to the heterogeneity of the PA metric values.

PA metrics

This review showed the pooled mean values of four commonly used PA metrics (steps, MVPA, TEE and ST) in patients with ILD. However, we need to be cautious when interpreting these values due to the substantial heterogeneity caused by differences in the activity monitors used, PA measurement methods and participant characteristics. For example, although most studies used validated activity monitors and PA metrics in COPD [71–74], there was no validation study of them in patients with ILD. However, we believe that these values are valid because PA metrics are broadly distributed in the general population or people with chronic diseases.

First, the pooled mean value of steps in ILD (5231 (95% CI 4577–5885) steps·day⁻¹) was similar to that in patients with COPD (5723 (sd, 3768) steps·day⁻¹) [75] and about two times lower than that in their healthy counterparts (10 195 (95% CI 8023–12 367) steps·day⁻¹). The mean value is slightly higher than a physical inactivity threshold (5000 steps·day⁻¹) [76], but the 95% CI includes 5000 steps·day⁻¹. Moreover, subgroup analyses revealed that type of activity monitor, sensor location, ILD subtype, FVC, D_LCO, 6MWD and use of LTOT are associated with steps, which aligns with previous findings. Although caution is required when interpreting the values due to variations in how activity monitors count steps [77], about half or more patients with ILD are inactive in terms of walk-related PA.

Second, the pooled mean value of MVPA in ILD was 97 (95% CI 64–130) min·day⁻¹. This value is lower than healthy controls in included studies (132 and 261 min·day⁻¹) and a previous study recruited similar-aged people (156 min·day⁻¹) [78]. Subgroup analysis showed that the mean was 152 (95% CI 123–181) min·day⁻¹ in studies which defined MVPA as time spent in PA of ≥2.5 METs, while the mean of MVPA defined as >3.0 METs was 67 (95% CI 51–82) min·day⁻¹. Patients with COPD spent similar times in MVPA, defined as >3.0 METs (65 (sd 11) min·day⁻¹) [55]. These results suggest that using a consistent definition of MVPA across studies and populations helps to ensure valid comparisons.

The mean values of MVPA in patients with ILD and healthy controls were dramatically higher than the international guideline recommendation of 150 min·week⁻¹ (approximately 21 min·day⁻¹) [2]. Similar values were reported in other populations (e.g. COPD, healthy controls) [55]. This highlights the need for a new PA recommendation based on MVPA measured by activity monitors. Moreover, calibration studies are needed to establish ILD-specific cut points to distinguish between various PA intensity levels using indirect calorimetry for the following reasons. First, cut points are drastically different between populations (e.g. children versus adults versus elderly) [79, 80]. Second, using cut points derived from healthy people in patients with exercise intolerance (e.g. COPD, ILD) may be flawed.

Third, the pooled mean of TEE was 2917 (95% CI 1370–4463) kcal·day⁻¹. This value was higher than that in healthy controls in included studies (2618 kcal·day⁻¹) and older adults (2501 kcal·day⁻¹) [81]. After omitting two studies with outliers, the mean was changed to 2130 (95% CI 1847–2412) kcal·day⁻¹. TEE of 2130 kcal·day⁻¹ could be referred to as a more valid value of TEE because the means of TEE in the omitted studies might be outliers.

Fourth, the pooled mean value of ST was 587 (95% CI 253–920) min·day⁻¹. Subgroup analyses suggested that possible sources of heterogeneity were the differences in ST definition, sensor types and D_LCO. Additionally, differences in calculation method and wear-time requirements may affect ST. For example, the means of ST in Hur et al. [8] and Alexandre et al. [31] are 349 and 536 min·day⁻¹. Hur et al. [8] measured PA 24 h·day⁻¹ and calculated daytime ST by excluding sleeping time. Alexandre et al. [31] measured daytime (12 h·day⁻¹) PA and calculated daytime ST. Interestingly, similar ST was observed in healthy older people in studies that measured daytime ST [82, 83]. Thus, standardisation of definition, wear-time requirement and calculation method are required to compare studies.

Finally, the use of questionnaires is limited in ILD. The most common questionnaire was IPAQ. Concurrent validity, internal consistency and responsiveness of IPAQ long-form were acceptable, and they estimated a minimally important difference [8]. Thus, IPAQ long-form can be used in combination with activity monitors or as an alternative in situations where activity monitors are unavailable.

Quality of the body of evidence

The quality of a body of evidence on steps, MVPA, TEE and ST was very low due to serious inconsistency and imprecision. Their sources were not only participant characteristics but also PA data collection and processing. Thus, future studies which measure PA using a standardised method with high reporting quality will change the pooled mean values of PA metrics.

Limitations

There are limitations in this review. First, the authors did not have access to Embase, which may reduce its comprehensiveness. Although the Cochrane Handbook for Systematic Reviews of Interventions encourages researchers to search Embase if accessible, Cochrane regularly searches Embase for trial reports and includes them in CENTRAL [84]. Thus, our review included at least relevant interventional studies, contributing to the comprehensive search. Second, ILD subtypes were inconsistent among included studies and affected the pooled value of PA metrics. Third, non-English studies were excluded during the selection process, potentially causing language bias. However, language restrictions during the selection process appeared to have little impact on language bias [85]. Fourth, studies that did not report separate participant characteristics and PA metrics for ILD patients were excluded. Fifth, multiple subgroup analyses were performed to investigate the sources of heterogeneity. Multiple subgroup analyses could reduce statistical power, leading to the potential of overlooking significant differences between subgroups [86]. Finally, we used several estimated mean values of individual studies not reporting the mean and sd of PA metrics to estimate the pooled means. These limitations could lead to biased results.

Practical recommendations

• 1) Researchers and clinicians are encouraged to report details of PA measurements following the checklist by Montoye et al. [15] or the template presented in table S9 in supplementary material 7.

• 2) Unfortunately, there is no standardised method for measuring PA in patients with ILD. Following expert consensus on objectively measured PA in COPD [87, 88] may be crucial to facilitate interpretation, pooling of PA data and comparisons with COPD.

• 3) The same MVPA and ST definitions should be used. We propose time spent in ≥3.0 METs PA as a definition of MVPA and time spent in ≤1.5 METs as a definition of ST, following the World Health Organization guidelines [89].

• 4) For identifying nonwear time, using activity monitors with automated algorithms for calculating nonwear time [90, 91] is encouraged if available. Alternatively, a wear diary could be used.

• 5) Patient characteristics, including ILD subtype, the severity of ILD, pulmonary function, exercise capacity and the number of patients with LTOT should be reported because these variables influence PA.

• 6) Validation studies are necessary to ensure the accuracy of major activity monitors and PA metrics in patients with ILD. Additionally, calibration studies are required to establish an ILD-specific cut-off for counting steps and distinguishing different intensities of PA.

Conclusions

In this systematic review of 40 studies using activity monitors or questionnaires in patients with ILD, we found severe heterogeneity in the methods used for PA data collection, processing and definition of PA metrics. The heterogeneity makes it difficult to interpret and pool PA data and compare results in ILD with other diseases. Therefore, we encourage researchers and clinicians to improve the quality of PA measurements. Our recommendations could be helpful when measuring and reporting PA in patients with ILD.