Validation Study: Hydra-Pedometer for Predicting Fluid Loss

The purpose of this project was to validate the use of a hydration pedometer that could give accurate estimations of fluid loss in a diverse population of users, across a varied range of walking and running speeds.  The monitor could be worn during a single exercise session only, or worn throughout the entire day to estimate total fluid loss associated with an individual’s normal daily activity.


Thirty subjects volunteered and signed informed consent forms and confidentiality agreements to participate in the study.  Sixteen men and fourteen women were recruited to represent a diversity of body types, ethnicity, age, and fitness level.  The mean weights of the males and females were closely matched to the October 2004 Center for Disease Control statistics for the average height and weight for American males and females (males=175.8 cm, 86.1 kg, females=162 cm, 74.0 kg).  The exercise histories of the thirty subjects included approximately 63.1% of their weekly exercise coming from running and/or walking, with the average length of a workout approximately 72 minutes.  The study subjects exercised an average of 4.7 days per week, at an intensity of 6.4 on a scale from 1-10, (1- very light, 10 – very hard).

Among the study participants, ethnicity included: Afro-American (2), Asian (2),  East Indian (2), Hispanic (1), and Pacific Islander (1).  The descriptive statistics of the study population are displayed below.


Initial Assessment:  Baseline Data

On an initial visit the subjects reported to the exercise laboratory and were acquainted with the testing protocol.  Urine specific gravity (USG) was collected on arrival to determine the subject’s current state of hydration using a refractometer (Misco model PA203X , Cleveland, OH).  Height was measured to the nearest 0.1 cm, using a Seca Accu-Hite wall mounted stadiometer.  Weight was taken in the nude, to the nearest 50 grams using load cell scale, (Befour PS-6600, Saukville, WI).  Skinfold measurements were taken with a Harpenden skinfold caliper at 7 sites, and employed the equations of Jackson & Pollock (1980,1978), for the determination of the subject’s body fat.

Baseline hemodynamic measurements were taken using the BioZ impedance cardiography machine, (Cardiodynamics, San Diego, CA).  A maximum assessment of aerobic capacity was determined using a graded exercise test.  Respiratory gas analysis was measured using a CPX Express system, (Medgraphics, St. Paul, MN) integrated to a Quinton Q55 treadmill.  Subjects were fitted with a Polar Vantage XL downloadable heart rate monitor and were acquainted with the treadmill protocol.  A flat protocol was selected with the subjects starting with a 3 mile per hour walk.  The speed was increased 0.5-2.0 mph every two minutes until the subject could no longer continue.  All graded exercise test protocols were able to achieve volitional exhaustion for the subjects between 8 and 15 minutes.  The subjects were scheduled for a second assessment as soon as their schedule allowed.

Second Assessment:  Measurements of Fluid Loss

A Urine Specific Gravity was recorded prior to testing.  Subjects were given a T-shirt, shorts, and socks, and were allowed to wear their own undergarments and shoes.  Clothing and two towels, (hand towel and bath towel) were weighed before and after the assessment to collect sweat loss in the clothing during the assessment.  The subjects were weighed nude to the nearest 50 grams, and a second BioZ impedance cardiography examination was repeated to verify the subject’s baseline hemodynamic status.

Wearing the provided clothing, the subjects were led to an area where their stride lengths would be determined for both a walking and running gait.  A fiberglass tape was laid out on the floor, and the subjects were instructed to take 10 walking and 10 running strides from a moving start.  The average of three trials were recorded in centimeters and programmed into a hydration pedometer.  The pedometer was fastened to the waist band of the subject’s shorts, approximately 3 inches from the midline of the body, pursuant to the Manufacturer’s recommendations.  The pedometer was started and the subject was escorted to a comfortable room where they were kept sedentary for two hours watching a movie, reclined on a sofa.

During this time the subject’s heart rate, temperature, humidity and steps were recorded at fifteen minute intervals.  Each subject was permitted to urinate if the need arose, and no food or drinks were permitted during the assessment period.  At the end of a 2 hour resting period, 5 minutes of resting metabolic gas analysis and resting heart rates were recorded.  The subjects were then reweighed nude to the nearest 50 grams.

The subject’s were then prepared for a one hour exercise protocol on a treadmill incorporating respiratory gas analysis of oxygen consumption and Respiratory Exchange Ratio (RER) measurements for each stage of exercise.  The first fifteen minutes the subject was instructed to walk and 2.0 miles per hours, and during the second fifteen minutes the speed was increased to 3.5 miles per hour.  For the third stage of fifteen minutes, subjects were required to run at 5.0 miles per hour.  If they were not physically able to perform the third stage they were advanced to the fourth and final stage.  The last stage was selected at 50% of the subject’s maximum oxygen consumption as determined from their maximal treadmill test taken on assessment one.  The speed was selected to achieve the determined MET display value equal to 50% of the subject’s maximum oxygen consumption on the Quinton 645 Treadmill controller.

Each fifteen minute stage the subject’s heart rate and perceived exertion was recorded using a Borg chart.  Temperature, humidity, step count and fluid loss values were recorded at each fifteen minute interval.  All steps were also recorded and verified using a Panasonic MS-1 Super VHS video camera.   At the conclusion of the exercise the subject was placed on a padded plinth and a final BioZ impedance cardiography assessment was performed, while still connected to the respiratory gas analysis equipment.

At the end of 15 minutes of recovery, the subjects were disconnected from the metabolic cart, and another Urine Specific Gravity (USG) was obtained.  The subjects were then instructed to weight nude, and to weigh their clothing.   A brief questionnaire was completed at the end of the assessment to determine their exercise history and habits, and their perception of their current level of thirst.  The subjects were then outfitted with a pedometer and instructed to record their beverage and food intake over the next 24 hours.  At the end of the twenty-four hour period the subjects were re-weighed and a final USG was obtained.


Statistical Analysis was performed utilizing the Statistical Package for Social Sciences (SPSS).  Comparisons between variables will utilize the bivariate correlation function and Paired Sample T-Test for reporting significance.  Comparisons will also employ the use of simplified box plots grouped with a summary of individual variable comparisons.  A level of significance for the validation study will be set a p=<.05.


Each subject was allowed to relax reclined on a couch for two hours of inactivity.  The last fifteen minutes of the two hour interval the subject was led to the laboratory and 5 minutes of respiratory gases were collected.  The subjects were reweighed and they were given a one hour exercise session on a treadmill.  Only 22 of 30 subjects were able to complete the third exercise stage which involved running at 5.0 mph for 15 minutes.  Two more subjects started the stage but were not able to complete the entire fifteen minutes, (9 and 10 minutes).  On the last stage the exercise intensity was chosen to approximate 50% of the maximum oxygen consumption, (56% or Heart Rate Reserve), and ranged from 2.2 to 4.6 mph.     A summary of the heart rate and respiratory data is displayed in Table 2.

Step Count

Validation of the step count mechanism showed a correlation of .716, (p<.001) with actual steps obtained by counting steps from a video recording of the exercise session.  Mean step counts taken from the video for the thirty subjects averaged 6806 steps, and the pedometer indicated a mean step count of 8094 steps, (+18.9%).  An overestimation of the step counts did not appear to be effected by the difference in walking or running speeds.

It should also be noted that ten of the subjects used in this study showed an overestimation of 15.1% using a similar pedometer, (6723 steps versus 5839 steps), manufactured by the same manufacturer on an earlier evaluation.  The correlation between the percentage error rates of the step counts on the ten subjects in the last study versus the step counts of the same ten subjects in this study was .487, but was non-significant (two tailed, p=.067).

It appears the step counts may be the largest introduced error in the prediction of fluid loss, and the overestimation of step counts have been fairly consistent across both studies performed thus far.


The reference stride lengths were calculated by determining the distance traveled on the treadmill during each 15 minute interval, at speed of 2.0 mph, 3.5 mph, 5 mph, and 50% of max VO2, (ranging from 2.4 to 4.6 mph), and dividing this by the video step count, (Treadmill Distance).   For the pedometer, the stride lengths were determined as a percentage of the subject’s body height and a programmed adjustment to stride length was made, depending upon the speed of the walking or running gait.   The resulting stride lengths were multiplied by the step counts recorded by the pedometer and the distance estimates were made, (%Ht-Ped Steps).  A high correlation between the two methods of determining distance was noted at r=.736, (p=<.001).

By using the same stride algorithms with actual step counts obtained through a video recording, the correlation to the reference distance increased to .852, (%Ht-Video Steps).    By substituting measured stride lengths, as determined by stepping off 10 strides walking and running with a measuring tape, the improvement in the distance estimates increased slightly from r=.852 to r=.853, (Stride-Video Steps).   It appears that using a percentage of the subject’s height to calculate stride length produces similar results to that of measuring stride length using a measuring tape, and involves less effort in programming by the user.

Figure 1: Methods of estimating distance as a percentage of height with actual step counts, percentage of height using video step counts, and using a measured stride length with video step counts.

Resting Caloric Expenditure:

Resting caloric expenditure using respiratory gas analysis was estimated (Estimated Cals) from measurements collected during the last five minutes of a two hour period of inactivity.   Calculations were made extrapolating oxygen consumption and RER values from the five minutes of resting data to the entire 2 hours of sedentary activity.  Step counts and algorithms contained within the pedometer (Pedometer Cals), were used to compute calories and were compared using a Paired-Sample T-Test, (Table 3).  The pedometer calorie estimates appeared to have a good correlation with respiratory gas analysis based estimates, (153 kcal versus 171 kcal), with a demonstrated correlation of r=.727

Exercise Caloric Expenditure:

Reliable calorie estimates are one of the most important variables in the prediction of the fluid loss.   In the study design, respiratory gas analysis was performed during the entire exercise session.  Each subjects average oxygen consumption and Respiratory Exchange Ratio (RER) was measured for each 15 minute stage.  To determine actual calories, each RER’s associated non-protein R calorie production per liter value was multiplied times the oxygen consumed, (Zuntz, 1901).  The associated actual calorie values (Actual Cals) were correlated with the caloric values computed by the pedometer during exercise, (Pedometer Cals).

A highly significant, (p=<.001), correlation of r=.841 was noted between actual calories (370 kcal) versus the pedometer calories (478 kcal).  This correlation was increased to r= .934, (397 kcal) when actual step counts taken from the video camera were used (Ped with Video).

Total Caloric Expenditure:

One of the difficulties faced in obtaining accurate total calorie estimates was extra time (mean= 27.6 minutes) between the pre weigh-in and post weigh-in that was not accounted for with the resting 2 hour assessment and exercise data alone.  These included such activities as the time taken to perform the initial BioZ, stride length by measurement of three trials of 10 steps both walking and running, trip to the restroom before the final weigh-in, travel to and from testing labs, time spent undressing and dressing for weigh-ins etc…  Estimates of caloric consumption of these extraneous activities were accounted for equally in all comparisons of Total Caloric Expenditure.

Resting Caloric Expenditures, Exercise Caloric Expenditures, and the calories resulting from extraneous activity using the data collected from the gas analysis cart (Cal Estimates), were compared to the pedometer outputs (Total Ped Cals), and to the pedometer outputs incorporating actual steps obtained off of video footage (Total Video Cals).   A correlation of r=.882 was achieved comparing indirect calorimetry estimates of 582 kcal, with the pedometer calorie values of 669 kcal, (Total Ped Cals & Cal Estimates).  The correlation increased to r=.918, using the step counts obtained from the video film (Total Video Cals & Cal Estimates), with a standard error of 13.1 calories, (552 kcal versus 582 calories).

Fluid Loss

A large factor in the prediction of fluid loss, not included in the computations used in the pedometer, is the initial state of the hydration of the subjects tested over the course of the evaluation.  All subjects Urine Specific Gravity (USG) measurements were taken both before and after the evaluation, and the average was taken to represent the average USG during the 3.5 to 4.0+ hours of testing.  The subjects were all asked to complete their urinations for both the pre and post USG sample collections, so that total urine production could be factored into the fluid loss of the subject throughout the assessment period (Figure 2).

Figure 2: Data showing the state of hydration during the assessment taken from urine specific gravities USG) has a correlation of r=.695, to the weight loss in milliliters per kilo per minute.

The fluid loss obtained, per kilo, and compensated for body dimension at -2/3 exponential ratio, was plotted against the subjects urine specific gravity and a correlation of -.695 was noted, with a two-tailed significance of p=<.001.  In essence the graph indicates that subjects testing in a more dehydrated state tended to lose less fluid during activity than those that came in with a more hydrated.  One subject before the assessment claimed to have consumed 32 oz of water right immediately prior to the assessment, and when this data was eliminated, the correlation increased to -.850, with an R-Square =0.72.  This infers there is a terrific range of normal fluid loss which may be individually dependant on the initial state of hydration of the individual.

The final factor to consider is that a base hydration value of 2200 ml of fluid loss was selected for the basal fluid loss state for the average female, (70.4 kilos) and 3000 ml of fluid loss was determined for the average male (83.5 kilos) for a 24 hour period.  This amount was to represent the beverage only contribution, (approximately 75-80%) of the total fluid recommendation for daily consumption as stated by the International Life Science Institute recommendations for fluid replacement.   Since the 75-80% of fluid loss calculations (the beverage portion of fluid loss), were programmed to display on the pedometer, a one-tailed significance measurement will be used.


Total fluid loss prediction using the pedometer in its current step counting and formulas had a significant correlation to actual fluid loss of p=<.001 with a correlation of r=.738 during exercise, and at r=.648 for the total fluid loss.  Actual total fluid loss measured from pre- to post-weight changes averaged 1017 ml, (Std Error 83.1 ml) while the average fluid loss estimates from the pedometer was 1023 ml. (Std Error 67.7 ml).  The ideal pedometer estimates would have been an estimate of approximately 75-80% of the fluid loss determined from the recorded weight changes.  This would represent the beverage portion of the subject’s fluid loss, so the values from the pedometer estimates appear inflated.  This was in a large part due to the over estimation of the step counts used in the calculation of fluid loss estimates.  Using the step counts from the video, the mean fluid loss estimate was decreased to 932 ml, with a Std Error of 51.4 ml.

The average urine specific gravity during the assessment period was determined at 1.0154 gm/liter for the 30 subjects tested, with a range of 1.0023 to 1.0288.  The mean urine specific gravity correlated to an average fluid loss of .056 ml of fluid loss per kilo per minute for the total activity session, with a range of 0.024 – 0.103 ml/kg-min-1 of fluid loss using the regression equation, (Figure 3).

Using the pedometer estimates, an average fluid loss of .055 ml per kilo, per minute for the activity was shown to be nearly identical.  The range of fluid loss for the pedometer estimates was 0.033 – 0.092 ml/kg-min-1, well within the range of fluid loss for the subjects tested.  Incorporating the video step counts into the computation showed a mean fluid loss of .050, with a range of fluid loss from 0.033-0.072 ml/kg*min-1.

Figure 3: Ranges of fluid loss comparing actual weight loss, with the fluid loss recommended by the pedometer, by actual step counts incorporated from video, and by subject’s perception of thirst. Boxes represent 25-75% of subjects fell in this range, bold line=median score, full quartile range represented by vertical lines extending from boxes. (* significance p =<.001).

At the conclusion of the testing session, each subject was asked to select one of the following answers that best typifies there current level of thirst.  13.3% said they were very thirsty, 53.3% said they were thirsty, 23.3% were not thirsty but could drink, and 10% said they were not thirsty.  Using the subjects own perception of thirst, a 16.9 oz bottle of water was used as a reference, and subjects were asked how much fluid they would consume to replenish themselves on their own volition.  The answer ranged from 2.0 oz to 33.8 oz, with a mean of 13.7 oz, (Std Dev. 6.0).   This was equivalent to 404 ml of fluid replacement and was compared to actual weight loss of 1.017 ml. and did not show significance with a correlation of r=.118.   Average replacement volume was .023 ml/kg-min-1, with a range of fluid of 0.004 to 0.045 ml/kg-min-1.  This was well below the range of fluid loss that was shown in the study of 0.024 to 0.103.  19 of 30 subjects (63.3%) were at or below the minimum value of fluid loss replacement if they were to use their own perception of thirst to dictate fluid replacement.  Predictions of fluid replacement recommended by the pedometer put all 30 subjects well within the range of fluid loss demonstrated within the study.   It seems apparent that the perception of thirst may not be a reliable indicator of adequate fluid replacement to maintain peak performance.

24 hour Assessment

Subjects were asked to wear a pedometer for 24 hours and record their fluid consumption for the 24 hour interval.  Urine Specific Gravities (USG) and body weight, (to the nearest .1 lbs) were collected prior to starting the pedometer, and after 24 hours had passed.  USG and weight measurements taken pre and post assessment and were identical at 1.0163.  Mean body weights for the 30 subjects was +0.1 kg heavier after the 24 hour assessment.

Predicted fluid loss displayed from the pedometer (mean = 3963 ml, Std Dev: 1054 ml) was compared to fluid intake from the subjects fluid log (mean = 3529 ml, Std Dev: 1400).  Mean step counts for the 30 subjects averaged 12,285, Std Dev: 8177 steps.   The fluid loss estimates from the pedometers appeared to be similar considering the inflated steps counts from the pedometers that were noted earlier in the study, but the correlation was non-significant at p=.147 with a correlation of  r=.271.  It appears this may have been due to the wide variations between individual subjects, as demonstrated by large standard deviations in the reported data.  This in part may be due to differences in fluid consumption resulting from foods that were consumed, in the altering the contribution from the beverage component.

Basal states of hydration are defined by the International Life Sciences Institute (Grandjean & Campbell, 2004) monograph, as the dietary fluid consumption that comes from the beverage portion of diet with no associated activity.  This represents between 75-80% of the total daily intake of fluid, with the remaining fluid intake coming from foods consumed in the diet.  The dietary recommendation for the basal beverage portion of fluid intake is 3000 ml per day for the average male, and 2200 for the average female.  The average weight and heights were programmed into pedometers for both males and females based on the Center for Disease Control (CDC) October 2004 statistics for the average American.  Estimates for basal fluid loss estimates over a 24 hour period, at all three pedometer settings for humidity are shown in Table 6.

These values are estimates of basal fluid rates representing 75-80% of the fluid requirements for a twenty four hour period without any physical activity.  It should be noted that the medium setting is similar to the International Life Sciences Institute (Grandjean & Campbell, 2004) recommendations for the average male and for the average female.

Impedance Cardiography Measurements

Baseline BioZ impedance cardiography data was collected on two different occasions scheduled on two separate days.  These were compared with a third BioZ assessment done 15 minutes post treadmill exercise on the day of the laboratory assessment.  There were no significance (p<.001) noted among all twenty variable assessed during the two baseline impedance cardiography trials.  Comparing the pre and post exercise values there was an expected significant difference noted in the subjects body weight, and body mass index (BMI), associated with the weight loss from the activity.  Hemodynamic measurements noted an increase in heart rate (+15 b/m) and decrease in stroke volume

(-17.0 ml), with no significant change in cardiac output.  This is consistent with the fluid loss associated with exercise, and the shunting of blood to the skin to cool the body.  The thoracic fluid content (TFC) verified there was a significant loss of fluid volume of 3.4% which contributed to the increase in heart rate and decrease in stroke volume.  An interesting finding was a noted drop in the Urine Specific Gravity measurements from 1.0160 to 1.0147, but it was not found to be significant (p=.161).

Other noted changes included a significant decrease in the acceleration index (AI) and the velocity index (VI) of 24% and 21% respectively.  This is due in part to the dilation of the blood vessels to accommodate the increase in blood flow needed for the exercise.  Lastly, a 9% decrease in the Left Ventricular Ejection Time was noted, indicating the increase in pre-load associated with exercise recovery.


It appears from the study that using a limited number of variables, namely: sex, weight, height and humidity, safe estimates for fluid replacement can be predicted utilizing a pedometer.  Furthermore these predictions can be accomplished in subjects while performing a single exercise session or throughout a twenty four hour period of normal daily activity.  It was noted that the subjects coming into the study showed a wide range in states of hydration that had a large effect on the fluid loss estimates while exercising.  Assuming that the subjects tested represent a normal range of hydration status, it was found that the pedometer recommendations for fluid replacement were nearly identical to the mean fluid loss of the group, at 0.055 versus .056 ml/kg*min-1 respectively.  It was also noted that the range of fluid replacement estimated by the pedometer was well within the range for the fluid losses reported from the subjects of the study, (Figure 3).   Subjects using their own perception of thirst and fluid replacement would have placed over half of the subjects of the study (63.3%) at risk of dehydration over time.

One concern is the error in the counting mechanism of the pedometer tested, and the observed over estimation of steps by 18.9%.  Since the goal of the pedometer was to display approximately 75-80% of the fluid loss on the scale, this was not achieved, as the pedometer estimates (1023 ml) were on par with the subjects weight change (1017 ml).  This can be remedied with a relatively simple adjustment to the equation or in the improvement in the technology of the step count technology.


Zuntz, N. (1901), Ueber die Bedeutung der verschiedenen Nahrstoffe als Erzeuger der
Muskelkraft. Arch. f.d. ges Physiol., Bonn, Ger.: LXXXIII, 557-571, 1901, Pflugers Acrh. Physiol., 83:557.

Jackson AS, Pollack ML, Ward A. (1980), Generalized equations for predicting body
density of women.  Med Sci Sport Exerc 12:175-182.

Jackson AS, Pollock ML. (1978), Generalized equations for predicting body density of
men. Br J Nutr 40:497-504.

Grandjean AC, Campbell SM. (2004), Hydration: Fluid for Life. International Life
Science Institute North American, Monograph Series. ILSI, Washington DC,
ISBN 1-57881-182-1.

Online, (June 2007), Center for Disease Control Center for Health Statistics, Advanced
Data No. 347, October 27, 2004,

Leave a Reply