The CADENCE-Adults study is the first calibration study to employ a sex-and-age balanced sampling approach to establish heuristic cadence thresholds associated with increasing absolutely-defined intensity during walking. Using two distinct analytical methods, we confirm that 100 steps/min is a reasonable heuristic threshold associated with absolutely-defined moderate intensity (i.e., 3 METs) ambulation in 21–40 year olds. We also provide further evidence for additional cadence thresholds associated with incremental MET-defined intensity up to and including 130 steps/min as a heuristic threshold associated with 6 METs. These additional heuristic values are important indices useful for public health purposes to guide 1) generalized cadence-based walking recommendations and 2) analysis and interpretation of minimally processed ambulatory data obtained from contemporary wearable technologies.
Heuristic values are evidence-based, practical, rounded numbers that are grounded in evidence, but may not be necessarily precise. They serve as useful and easy to recall mental short cuts, quickly conveying generalized or broadly representative information to guide decisions. A simple daily-use example of a heuristic value is the estimated time it would take to drive between two cities. Other common public health-related examples of heuristic values include “eat 5 fruits and vegetables per day”, “be active 30 min/day”, and “limit time spent watching TV to 2 h/day”. It bears emphasizing here that heuristic values, while evidence-based and thus appropriate for public health purposes, are by definition not individualized.
We first proposed the heuristic value of 100 steps/min as a proxy indicator of moderate intensity in 2005, based on a linear regression model of treadmill walking [9]. A number of other studies [6,7,8, 10, 11] subsequently confirmed this heuristic value, despite acknowledging evidence of a tolerable range of inter-individual variation. Notably, these studies have been generally small, included predominantly younger samples, did not always employ a direct observation criterion standard of step counting, and employed various analytical approaches. This initial article focused on 21–40 year olds from the CADENCE-Adults study represents the largest sex-and-age structured sample to date employing a direct observation standard and using both regression and ROC analysis to inform evidence-based but generalized heuristic cadence values associated with absolutely-defined moderate and vigorous intensity. The 100 steps/min threshold for absolutely-defined moderate intensity continues to be confirmed for this age group. The stability of this heuristic across the adult lifespan up to 85 years of age will continue to be tested as part of the CADENCE-Adults study as data collection is completed.
To date, there have been three studies that have reported values congruent with a heuristic value of 130 steps/min associated with 6 METs (i.e., absolutely-defined vigorous intensity) in ostensibly healthy adults [6, 9, 22]. Herein, the optimal absolutely-defined vigorous intensity cadence thresholds were 129 and 120 steps/min, identified using regression and ROC analyses, respectively. Both algorithms are commonly accepted means of determining associations between independent and dependent variables and establishing thresholds. However, both analyses have different assumptions, and therefore different limitations. Regression models may be overly influenced by outliers, while ROC curves are organized in a rank-order fashion. By incorporating both methods, we provide more robust support for the heuristic thresholds reported herein. With that said, setting a lower threshold increases sensitivity, but reduces the specificity and PPV; the opposite is true for higher thresholds. Considering these trade-offs, we settled on a final heuristic threshold of 130 steps/min for absolutely-defined vigorous intensity.
The heuristic thresholds of 100 and 130 steps/min demonstrated good-to-excellent classification of absolutely-defined moderate and vigorous intensity ambulation, with an overall accuracy (true positive and true negative rates of > 85%). Moreover, for individuals walking ≥100 steps/min (~ 53.6–67.1 m/min or ~ 2.0–2.5 mph; Table 2), the probability (PPV) of achieving an absolutely-defined moderate intensity was 91.4%. For 130 steps/min (107.3 m/min or ~ 4.0 mph), the probability (PPV) of achieving an absolutely-defined vigorous intensity was 70.7%. This value is less than ideal and may be influenced by the lower number of participants (n = 49) who achieved 6 METs. However, this number still reflects 65% of the participant pool, and the associated NPV of 95.8% conversely suggested a very high probability that individuals walking at < 130 steps/min were at an intensity < 6 METs. Overall, this evidence supports the use of 100 and 130 steps/min, corresponding to absolutely-defined moderate and vigorous intensity ambulatory activity, respectively, as direct translations of public health recommendations for the minimum desired ambulatory intensity required to achieve health and fitness improvements [12, 13].
In the current study, we employed an absolutely-defined measure of intensity (i.e., METs), as opposed to a relatively-defined measure of intensity (e.g., %VO2Reserve, % Heart Rate Maximum [HRmax] or Heart Rate Reserve [HRR]). This approach is consistent with previous studies that have determined accelerometer activity count cut points related to absolutely-defined moderate and vigorous intensities [23,24,25], and also with U.S. Federal physical activity guidelines [13, 26] and the American College of Sports Medicine position stand [27] that express their physical activity recommendations (intended for public health applications) using METs (e.g., 500–1000 MET-min/week). However, the use of absolute intensity may not be ideal for all adults, especially individuals who are older or have low fitness levels, whereby an indicator of absolute intensity represents a higher percentage of maximal capacity (relative to a younger or fitter adult) [27]. Few studies have examined the cadence-intensity relationship using relatively-defined measures of intensity, which may be more suitable for clinical or other types of individualized applications. For example, Serrano et al., [28] and Slaght et al. [29] reported cadence thresholds of 115 ± 10 and 114 ± 11 steps/min, respectively, associated with 40% of VO2reserve. In addition, Pillay et al., [30] found that 122 ± 37 steps/min corresponded to 60% of HRmax, whereas O’Brien et al., [11] reported that ~ 120–125 steps/min corresponded to 40% METmax, dependent on the modelling technique and the covariates included in the model (e.g., height, leg length). The differences observed between these cadence thresholds (employing different relative indicators of intensity) and those reported herein (absolutely-defined) reflect the inconsistencies between the implemented intensity definitions. Unlike absolute intensity measures, for which there is consensus in the literature regarding what constitutes a moderate or vigorous intensity (3 and 6 METs respectively) [26, 27], there appears to be less consensus regarding relatively-defined intensity [31]. Using a single example of %HRmax, moderate intensity has been defined as 64–76% HRmax [27], 55–69% HRmax [32], and 60% HRmax [30]. While there are strengths to using a relative intensity approach, especially for clinical and other types of individualized applications, there are also weaknesses, such as the need for a maximal fitness test to establish relative moderate and vigorous intensity levels based on individualized maximal/peak VO2 or HR values. Although it is possible to use equations to estimate %HRmax or HRR [33,34,35,36,37], such estimates are based on assumptions that may introduce an additional source of error. Indeed, there is no universally accepted HR-based equation with a minimal and acceptable (< 3 bpm) level of error [38]. Furthermore, some equations may be age (e.g., Åstrand [37]) or sex specific (e.g., Gulati et al., [34]), so care must be taken when applying these equations to various populations. Collectively, this makes such indicators of relative intensity less practical for public health applications including translations of physical activity guidelines as they are currently expressed [13, 26]. In summary, we believe our approach to using absolutely-defined intensity is reasonable and defensible given the consistency with previous studies and with public health guidelines. Still, we anticipate future research will be able to delve into the utility and limitations of individualizing cadence-based exercise prescriptions for clinical and more individualized applications (e.g., personal training).
An innovation of this study includes proffering a more comprehensive set of incremental cadence-intensity thresholds, including optimal and heuristic cadence thresholds for the intermediary values of 4 and 5 METs. Notably, with each increasing intensity level, the precision estimates (prediction intervals for regression; confidence intervals for ROC curve) tended to narrow, suggesting greater confidence that individuals walking at higher cadence thresholds will indeed achieve the desired higher intensity level. Based on the values presented herein, it is reasonable to conclude that, starting from 100 steps/min, each 10 steps/min increase is roughly associated with an increase in intensity of 1 MET, confirming the findings of a small preliminary study conducted in 2005 [9]. Notably, based on the regression and ROC optimal thresholds (both 112.5 steps/min) corresponding to 4 METs, we may have selected either 110 or 115 steps/min. However, considering our definition of a heuristic threshold (not only evidence-based, but also practically useful) and the complete set of cadence-intensity thresholds being put forth herein, we settled on 110 steps/min. In numerical terms, this was associated with a decrease in the PPV (8.3%) and increase in the NPV (4.6%) for this intensity level. Notably, these cadence thresholds, including that associated with 6 METs, are all achievable within the range of walking cadences for healthy adults; the walk to run transition occurs at ~ 140 steps/min [39]. Moreover, in the current study we deliberately excluded the bouts where 15 participants transitioned to running, so the evidence presented herein solely arises from walking cadence. With walking being the most commonly reported and widely accessible form of physical activity [40], this intentional focus greatly improves the utility of this set of cadence-intensity thresholds for application in the general population.
Regarding precision of regression predictions, we chose to report prediction intervals (PIs). While confidence intervals are more commonly reported, PIs are more appropriate for repeated measures dataset regressions, as they account for not only the uncertainty of the actual population mean, but also the overall spread of the data. For this reason, PIs appear wider in distribution compared to confidence intervals. Cadence PIs for 3 METs were seemingly large (45.9–111.2 steps/min). It is important to note that we intentionally included all walking bouts (e.g., starting at 0.5 mph) in order to incorporate a maximal range of ambulatory speeds. However, extremely slow speeds (e.g., 0.5 and 1.0 mph) may be considered non-ecological, as young healthy adults do not typically walk at these slow speeds and we observed our own participants struggling to find a comfortably paced execution of these speeds. In a different study, even when instructed to walk ‘rather slowly,’ healthy young adults (19–39 years old) chose to walk at a pace of 2.1 ± 0.4 mph [41]. When excluding the two slowest walking speeds employed herein, the mean cadence associated with 3 METs slightly decreases (96.4 steps/min), but more importantly the PIs tighten considerably (72–114 steps/min).
While the purpose of this analysis was to establish heuristic cadence-intensity thresholds in 21–40 years olds using group aggregate data, we acknowledge that inter-individual variability exists and that any heuristic threshold will have limited precision in terms of applicability to any single individual. While we accounted for the potential influence of both leg length and sex in the overall model fit across all participants, these additional variables did not change the model prediction (RMSE 0.68 ± 0.10 and 0.69 ± 0.10, respectively, compared to 0.68 ± 0.10 for the base model). Furthermore, the addition of leg length only marginally improved the model fit (R2 = 0.85; compared to the basic model, R2 = 0.84). Notably, the regression model including leg length predicted only a 0.58 MET difference at a given cadence between participants with the longest versus shortest leg length (95.5 cm vs. 65.7 cm, respectively). Similarly, when BMI was added to the regression model, the model fit did not change (R2 = 0.84), and there was only a 0.57 MET difference in predictions for participants with the highest and lowest BMI (36.9 vs 19.4 kg/m2, respectively). Given the limited change in model accuracy when adding these additional factors, we considered it reasonable to only include cadence in the final model. We acknowledge that any remaining variance in intensity at a given cadence may be better explained by other factors. In addition, we did not measure VO2peak or VO2max in this study, and as such are unable to make any conclusions regarding fitness and its impact on our study outcomes, or provide cadence thresholds corresponding to relative intensity measures. It bears repeating here, however, that the goal of establishing cadence-based thresholds corresponding to absolutely-defined intensity levels is to provide clear guidelines with little or no additional individual information required. Finally, we also acknowledge that cadence is specific to bipedal locomotor movements and further that these thresholds are most applicable to walking behaviors that are characteristically rhythmic, purposeful, continuous, and advancing forward through space.
Despite these limitations, cadence thresholds associated with absolutely-defined moderate and vigorous ambulatory intensity can serve as important heuristic values in efforts to measure and modulate adult walking behaviors, thus extending the potential utility of contemporary wearable technologies that offer step counting and cadence tracking features. One clear application of these cadence thresholds is for implementation in walking interventions. In our recent systematic review [42], we identified a limited number (n = 9) of intervention studies that had used a cadence-based goals to modulate walking behavior, or used cadence thresholds to quantify physical activity intensity from accelerometers and wearable device data. Based on the small number of studies and the observed associated high risk of bias, we concluded that it was premature to synthesize their findings. Rigorously designed walking intervention studies that utilize these cadence thresholds to convey and evaluate ambulatory behavior are required to elucidate the associated health benefits (e.g., improvements in aerobic fitness, blood pressure and glucose levels, body composition). In addition, future research should also explore ways to individualize cadence-based intensity prescriptions (e.g., using indicators of relative intensity) similar to Slaght et al., [29] and to modulate intensity in predictable ways (e.g., manipulating cadence using rhythmic auditory cueing [music or metronome]).