The electronic pedometer is a valuable tool for researchers interested in quantifying ambulatory activity and is typically used to record steps. Previous studies have demonstrated the validity of these devices in normal weight individuals under controlled laboratory conditions (^2,5) and in the field (¹³). However, there is concern regarding pedometer validity in obese individuals. Two studies indicate that accuracy is compromised in overweight and obese individuals (^10,14). However, Swartz et al. (¹⁵) assessed pedometer accuracy in normal (BMI < 25 kg·m⁻²) overweight (BMI 25–29.9 kg·m⁻²), and obese (BMI ≥ 30 kg·m⁻²) participants and found that BMI category had no effect on pedometer accuracy. Elsenbaumer and Tudor-Locke (⁶) also found that BMI category is relatively unimportant and has little effect on the percent error of pedometers at self-selected normal walking speeds.

Some of the inconsistency in the literature may be due to previous studies stratifying the participants based on BMI, which is not the best indicator of abdominal adiposity (^1,11). It is possible that waist circumference or waist-to-hip ratio may have a greater influence on pedometer accuracy. Shepherd et al. (¹⁴) propose that in individuals with a large amount of abdominal adipose tissue, the soft tissue may dampen the vertical accelerations of the trunk, which could contribute to lower step counts. Furthermore, accuracy in these individuals may be compromised because the pedometer is often tilted away from the vertical plane, thus increasing friction in the spring-suspended lever arm and resulting in a failure to register all steps.

Several studies have used the Yamax Digiwalker SW-200 (SW) in normal, overweight, and obese populations (^{6,12,15,16,18}). The SW uses a spring suspended horizontal lever arm that moves up and down in response to the hip’s vertical accelerations. This movement opens and closes an electrical circuit; the lever arm makes an electrical contact and a step is registered. For this pedometer to work correctly it must be placed in a vertical plane, perpendicular to the ground. In contrast, the New Lifestyles NL-2000 (NL) uses a piezo-electric accelerometer mechanism that has a horizontal cantilevered beam with a weight on the end, which compresses a piezo-electric crystal when subjected to acceleration. This generates voltage proportional to the acceleration and the voltage oscillations are used to record steps. Thus, this mechanism could be less susceptible to errors that occur due to tilt.

Although the NL costs slightly more than the SW ($50 vs $25), there are distinct advantages for using the NL in research. The NL has the ability to store seven days worth of data in 1-d epochs, and automatically resets itself to 0 steps at midnight. This allows a researcher to seal the pedometer before giving it to the participant, which reduces the problem of “reactivity” (increased physical activity due to knowledge of one’s steps). The SW has to be manually reset, and although it can store data over several consecutive days, it does not distinguish between days of data collection.

Because pedometers are often used in research studies and physical activity interventions aimed at overweight and obese individuals, it is important to examine the accuracy of waist mounted pedometers in these populations. Therefore, the purpose of this study is to examine the effects of BMI, waist circumference, and pedometer tilt angle on the accuracy of the Yamax Digiwalker SW-200 and the New-Lifestyles NL-2000 in overweight and obese individuals under controlled and free-living conditions.

METHODS

Subjects.

Forty adult participants (20 men, 20 women) from the Springbrook Wellness Center at Blount Memorial Hospital and The University of Tennessee and surrounding community volunteered to participate in the study. Those with a BMI below 25 kg·m⁻² or who were unable to walk at 4 mph on a treadmill were excluded from the study. The protocol was approved by the University of Tennessee institutional review board and the Blount Memorial Hospital investigational research and review committee. Each participant signed a written informed consent and completed a Physical Activity Readiness Questionnaire (PAR-Q) before participating in the study.

Anthropometric measures.

Participants had their height and weight measured (in light clothing, without shoes) using a stadiometer and calibrated physician’s scale, respectively. Body mass index was calculated according to the formula: body mass (kg) divided by height squared (m²). Circumference measures of the waist, hip, and abdomen were performed according to the guidelines established by Lohman, Roche, and Martorell in the Anthropometric Standardization Reference Manual (⁴). Briefly the waist was taken at the narrowest portion of the torso between the most inferior rib and iliac crest, the hip was taken at the maximal circumference of the buttocks above the gluteal fold, and the abdomen was taken at the level of the umbilicus. All measurements were taken by the same tester using a Gullick measuring tape (Creative Health Products, Inc, Plymouth, MI). Circumference measures (recorded to the nearest 0.1 cm) were taken in duplicate, with an average of the measurements used for analysis. A third measurement was taken if the first two were different by more than 1.0 cm, and the lowest two values were then averaged.

Treadmill walking.

Participants walked on a treadmill (Quinton Instrument Co., Q55XT, Bothell, WA) at speeds of 54, 67, 80, 94, and 107 m·min⁻¹ for 3 min at each speed. The treadmill speed was calibrated by measuring the belt length (3.190 m) and the time required to complete 25 revolutions of the treadmill belt. This was verified by using a handheld digital tachometer (Nidec-Shimp America Corp., DT-107, Itasca, IL) that was calibrated to an accuracy within ± 0.1%. A carpenter’s level was used to calibrate the treadmill grade at 0.0%.

Before walking on the treadmill a New-Lifestyles NL-2000 (NL) and a Digiwalker SW-200 (SW) were placed on the left and right side of the body, respectively. The pedometers were placed at the midline of the thigh on the waistband of the participant’s clothing. Once the pedometers were in place, a protractor (Sears Craftsman magnetic professional) was used to measure the pedometer tilt angle. A negative tilt indicates that the top of the pedometer was tilted away from the body, whereas a positive number indicates that the top of the pedometer was tilted towards the body. During the treadmill protocol, an investigator tallied actual steps with a hand-tally counter. Between trials the participant straddled the treadmill belt so that the step values from the pedometers could be recorded and reset to 0 for the next trial.

24-h free-living study.

A subset of 36 participants volunteered to wear the pedometers for a 24-h period. The NL and SW were worn on the left and right hip, respectively, in the same position as during the treadmill walking. The participants were instructed to wear the pedometers for 24-h except when sleeping, showering, and swimming. Participants were given written and verbal instructions on how to work the pedometers and the proper placement of the pedometers. They were also given a data sheet to record when the pedometers were put on and taken off, and how many steps were recorded by each pedometer at the end of the day.

Statistical treatment.

Statistical analyses were carried out using SPSS version 12.0 (SPSS Ins., Chicago, IL). For all analyses, an alpha 0.05 was used to indicate statistical significance. All values are reported as mean ± standard deviation. Initially, a two-way repeated measures ANOVA (speed × pedometer brand) was used to compare mean difference scores (actual steps minus pedometer counts) for steps taken. To examine the effects of obesity on pedometer accuracy the participants were divided into tertials for each category; BMI (25–29.9 kg·m⁻², 30–35 kg·m⁻², and >35 kg·m⁻²), waist circumference (<95.7 cm, 95.7–106.5 cm, and >106.5 cm), and absolute pedometer tilt angle (≤10°, 10.1–15°, and >15°). Because the direction (towards or away from the trunk) of pedometer tilt was not an important factor affecting the pedometer error score, we chose to use the absolute value of the pedometer tilt angle (i.e., absolute tilt ≤10° meant that it was within ±10° of vertical). Three-way repeated measures ANOVA (BMI × speed × pedometer brand), (waist circumference × speed × pedometer brand), and (pedometer tilt × speed × pedometer brand) were then used to compare mean difference scores for steps taken. Simple contrasts were used to examine significant differences from actual steps. Pairwise comparisons with a Bonferroni adjustment were performed to locate significant differences when necessary.

Independent t-tests were used to examine the difference between genders for anthropometric variables. Pearson correlation coefficients (r) were computed between the pedometer error score and BMI, waist circumference, and absolute pedometer tilt angle at each speed. Partial correlations were computed between pedometer error score and pedometer tilt angle at each speed while controlling for BMI and waist circumference. Paired t-tests were performed to examine differences between pedometers for the 24-h free-living study.

Modified Bland–Altman Plots were used to graphically show the variability in pedometer error scores (³). This allowed for the mean error score and the 95% prediction interval to be shown. Devices that are accurate will display a tight prediction interval around zero. Data points below zero signify an overestimation, whereas points above zero signify an underestimation.

RESULTS

Physical characteristics.

The physical characteristics of the participants are shown in Table 1. The males had significantly greater body mass (P = 0.001), waist circumference (P = 0.001), abdominal circumference (P = 0.002), waist-to-hip ratio (P < 0.001), and height (P < 0.001). There were no differences between males and females for age, BMI, and hip circumference (all P ≥ 0.05).

Pedometer accuracy during treadmill walking.

Figure 1 shows the percentage of actual steps for the NL and SW during treadmill walking. The SW recorded significantly fewer steps than the NL at all walking speeds (P < 0.05) and significantly underestimated actual steps at speeds of 54–94 m·min⁻¹ (P < 0.05). The NL underestimated actual steps by 7%, on average, at 54 m·min⁻¹ (P < 0.05), but was within 3% at speeds of 67–107 m·min⁻¹ (P ≥ 0.05).

Figures 2–4 show the effects of BMI, waist circumference, and the absolute pedometer tilt angle on the percentage of steps taken during treadmill walking. In general, the SW became less accurate with increasing BMI, increasing waist circumference at speeds of 80 m·min⁻¹ and slower, whereas the pedometer tilt angle was a more important factor across all speeds. The accuracy of the NL was not affected by BMI, waist circumference, or the pedometer tilt angle.

Figures 5 and 6 show modified Bland–Altman plots for the NL and SW for each pedometer absolute tilt group at all walking speeds. The NL, on average, undercounted by 1.2 ± 34.3 steps (0.1 ± 11.8%) in those with an absolute tilt less than or equal to 10°, by 1.2 ± 13.9 steps (0.5 ± 4.6%) in those with an absolute tilt between 10.1° and 15°, and by 7.2 ± 31.7 steps (2.6 ± 10.6%) in those with an absolute tilt greater than 15°. The SW, on average, undercounted by 15.8 ± 73.4 steps (5.6 ± 25.4%) in those with an absolute tilt less than or equal to than 10°, by 45.6 ± 92.9 steps (14.4 ± 28.9%) in those with an absolute tilt between 10.1° and 15°, and by 139.6 ± 148.6 steps (41.3 ± 44.1%) in those with an absolute tilt greater than 15°.

Tables 2 and 3 list the Pearson correlation coefficients (r) between the SW and NL error score, respectively, at each speed with BMI, waist circumference, and absolute pedometer tilt angle. The absolute tilt angle of the SW pedometer was significantly correlated with the SW error scores at all speeds (P < 0.05). However, BMI was not significantly correlated with the NL or SW error scores at any speed (P ≥ 0.05). Furthermore, waist circumference was not significantly correlated with pedometer error scores at any speed (except at 67 m·min⁻¹, for both pedometers). When waist circumference and BMI were controlled for, the absolute tilt angle of the SW pedometer was still negatively correlated with the error scores at speeds between 67 and 107 m·min⁻¹. In contrast, the absolute tilt angle of the NL pedometer was not correlated with the error scores at any speed (P ≥ 0.05).

Pedometer steps during 24-h study.

The participants wore the pedometers for an average of 14.3 ± 2.0 h, which is similar to previous studies (^8,9,17). On average the NL recorded 7662 ± 3035 steps, which was significantly greater than the SW, which recorded 6632 ± 3355 steps (P < 0.001). Figure 7 shows a modified Bland–Altman plot for the NL and SW step counts during the 24-h measurement.

DISCUSSION

Electronic pedometers are growing in popularity among researchers due to their ease of use, objectivity, and low cost. Due to the increasing prevalence of overweight and obesity in the U.S., it is desirable that these devices be accurate in those populations. The main finding of this study is that the NL pedometer is more accurate than the SW pedometer during treadmill walking in those who are overweight and obese. The primary factor affecting SW accuracy appears to be pedometer tilt, although BMI and waist circumference also contribute to the SW miscounting steps.

Previous studies looking at pedometer accuracy in overweight and obese individuals have used BMI to classify individuals and have found conflicting results. Shepherd et al. (¹⁴) reported that a spring-levered pedometer (Sportline 345) missed 4.9 ± 6.1% of steps in obese individuals during a brisk 400-m walk, whereas the average error was only 1.1 ± 1.0% in the nonobese. When the participants walked 10 m slowly, the average error increased to 18.7 ± 23.7% in the obese and 11.4 ± 13.6% in the nonobese. In contrast, Swartz et al. (¹⁵) found that BMI had no effect on pedometer error during treadmill walking from 54 to 107 m·min⁻¹. They found that the Yamax SW-200 significantly undercounted steps at 54 and 67 m·min⁻¹, but was accurate at 80 m·min⁻¹ and above. Differences between these studies that could have contributed to the discrepant results were the pedometer brands used, and the fact that some participants in the study of Shepherd et al. (¹⁴) had gait abnormalities due to lower extremity surgery.

Melanson et al. (¹⁰) recently conducted a two-part study that examined the accuracy of various pedometers. The first part of the study examined the accuracy of the Yamax Digiwalker SW-200 during overground walking at self-selected normal and brisk walking speeds. The second part of the study examined a piezo-electric pedometer (Omron HF-100), and two spring-levered pedometers (Walk-4-Life LS-2500, and Step Keeper HSB-SKM), during treadmill walking at speeds of 26.8–69.7 m·min⁻¹ and during overground walking at one self-selected speed. In the first part of the study they found that the accuracy of the Yamax Digiwalker SW-200 decreased with increasing age, body weight, and BMI at both normal and brisk walking speeds, but they were not able to determine which variable was the most important factor responsible for decreased pedometer accuracy. In the second part of the study they reported that a piezo-electric pedometer (Omron HF-100) was more accurate than spring-levered pedometers (Walk-4-Life LS-2500, and Step Keeper HSB-SKM) at slow walking speeds. Thus, they suggested that the Omron HF-100 pedometer would be a more accurate pedometer in obese individuals because they tend to walk at slower speeds (<80 m·min⁻¹). However they did not directly compare the Omron HF-100 and Yamax Digiwalker SW 200 against each other. The Omron HF-100 pedometer is no longer available, but a newer model (Omron HJ-112) uses a similar piezo-electric mechanism, although it has not yet been validated.

Melanson et al. (¹⁰) and Swartz et al. (¹⁵) recognized that their studies were limited by not measuring waist circumference and pedometer tilt, respectively. The current study extends their findings by directly comparing a piezo-electric pedometer (NL) and a spring-levered pedometer (SW) against each other under treadmill and free-living conditions. In addition, anthropometric measurements were taken and pedometer tilt was measured so that factors affecting pedometer accuracy could be examined more closely.

During the 24-h measurement the SW undercounted steps compared with the NL in this overweight and obese population. Schneider et al. (¹²) found that over a 24-h period the NL recorded 2.2% more steps than the SW (P > 0.05) in nonobese individuals with an average BMI of 25.8 ± 4.1 kg·m⁻². A separate study from our laboratory found that the NL recorded 4.3% more steps than the SW (P > 0.05) in a group with an average BMI of 26.0 ± 6.1 kg·m⁻² (⁷). These two studies confirm that the NL and SW record similar values in nonobese individuals. In the present study we found that on average the NL recorded 16.5% more steps than the SW in a group with an average BMI of 32.6 ± 4.8 kg·m⁻².

We considered whether moving the SW pedometer to the midaxillary line in individuals with large waist circumferences (to put the pedometer in a more vertical position) would improve the accuracy. Although we did not test this, Swartz et al. (¹⁵) showed that moving the SW to the midaxillary line worsened the underestimation of steps, compared to wearing it in the midline of the thigh. Thus, the alternative for obese individuals is to use a piezo-electric pedometer rather than altering the position of a spring-levered pedometer.

Considering that approximately 65% of U.S. adults are overweight or obese, this study has implications for researchers. For example, pedometers are often used in weight loss studies to monitor physical activity, and thus it is desirable to select devices that are accurate in this population. However, the current study does have some limitations. Specifically, the small sample size and the participants’ ethnicity (predominantly Caucasian) limit the generalizability of the findings.

This study showed that a piezo-electric pedometer (NL) is more accurate than a spring-levered pedometer (SW) for counting steps in overweight and obese individuals, especially at slower walking speeds. The most important factor affecting the accuracy of the spring-levered pedometer (SW) is pedometer tilt angle; waist circumference and BMI are of lesser importance. In contrast, a piezo-electric pedometer (NL) was not affected by any of these variables. Researchers and practitioners need to take this information into consideration when choosing a pedometer for use in overweight and obese individuals.