Metabolic Costs of Military Load Carriage over Complex Terrain

Metabolic Costs of Military Load Carriage over Complex Terrain Abstract Introduction Dismounted military operations often involve prolonged load carriage over complex terrain, which can result in excessive metabolic costs that can directly impair soldiers’ performance. Although estimating these demands is a critical interest for mission planning purposes, it is unclear whether existing estimation equations developed from controlled laboratory- and field-based studies accurately account for energy costs of traveling over complex terrain. This study investigated the accuracy of the following equations for military populations when applied to data collected over complex terrain with two different levels of load carriage: American College of Sports Medicine (2002), Givoni and Goldman (1971), Jobe and White (2009), Minetti et al (2002), Pandolf et al (1977), and Santee et al (2003). Materials and Methods Nine active duty military personnel (age 21 ± 3 yr; height 1.72 ± 0.07 m; body mass 83.4 ± 12.9 kg; VO2 max 47.8 ± 3.9 mL/kg/min) were monitored during load carriage (with loads equal to 30% and 45% of body mass) over a 10-km mixed terrain course on two separate test days. The course was divided into four 2.5-km laps of 40 segments based on distance, grade, and/or surface factors. Timing gates and radio-frequency identification cards (SportIdent; Scarborough Orienteering, Huntington Beach, CA) were used to record completion times for each course segment. Breath-by-breath measures of energy expenditure were collected using portable oxygen exchange devices (COSMED Sri., Rome, Italy) and compared model estimates. Results The Santee et al equation performed best, demonstrating the smallest estimation bias (−13 ± 87 W) and lowest root mean square error (99 W). Conclusion Current predictive equations underestimate the metabolic cost of load carriage by military personnel over complex terrain. Applying the Santee et al correction factor to the Pandolf et al equation may be the most suitable approach for estimating metabolic demands in these circumstances. However, this work also outlines the need for improvements to these methods, new method development and validation, or the use of a multi-model approach to account for mixed terrain. Introduction Military operations often require dismounted personnel to carry heavy loads over long distances.1 Although loads of 30% body mass or less are recommended to avoid performance decrements,2 a previous study of combat operations in Afghanistan showed that the average light infantry rifleman carried over 43 kg of critical equipment during approach marches.3 Although additional equipment increases versatility, heavy load carriage significantly increases energy costs.4 Excessive energy expenditure can degrade soldier performance, contribute to thermal strain, as well as lead to physical and cognitive fatigue. Therefore, the ability to accurately predict energy costs is important for effectively planning soldier activities. Practitioners often rely on predictive equations to estimate the metabolic costs of military load carriage.5 Many of the well-established predictive equations were developed from laboratory steady-state exercise protocols that controlled for metabolic rate moderators such as walking speed, treadmill grade, and external load.6–9 An inherent assumption is that predictive equations developed from steady-state data can be used to predict energy costs under dynamic conditions where activities change frequently enough that equilibrium cannot be achieved. Military examples of dynamic conditions include dismounted patrols and approach marches that often require traversing landscapes with shifting gradients and surfaces.10 As these conditions cannot be realistically replicated in laboratory settings, field research data are necessary to determine whether these predictive equations accurately estimate the metabolic costs of military load carriage over complex terrain. However, there has not yet been a direct comparison between predictive equations using metabolic data collected over complex terrain. These data will provide a basis for realistic expectations regarding the accuracy of metabolic cost predictions for load carriage. This investigation sought to evaluate how well-established predictive equations estimate the metabolic costs of military load carriage over complex terrain and to identify the most suitable equations for these estimations. Ultimately, the findings of this investigation lay the ground work for future efforts to refine, hybridize, or apply multi-model approaches to improving the accuracy of metabolic cost estimation methods. Materials and Methods Study Design Test volunteers completed two marches over a complex terrain course carrying a different external load (either an additional 30% or 45% body mass load). The marches were separated by a minimum of 5 d and the order of the external load conditions was randomized. Six predictive equations6–8,11–13 (Table 1) were used to estimate metabolic rate over course segments and evaluated against measurements from a portable oxygen uptake monitor system (COSMED K4b2 (K4) COSMED Sri., Rome, Italy). The Santee et al equation used in this investigation includes a multiplication sign following the S2 term instead of the minus sign that was erroneously included in the original published equation. Table I. Predictive Equations from Previous Literature Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  G, grade (% for Givoni and Goldman, Pandolf, and Santee, decimal for rest); L, external load (kg); M, body mass (kg); S, speed (m/min for ACSM, km/h for Givoni and Goldman, m/s for rest). Table I. Predictive Equations from Previous Literature Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  G, grade (% for Givoni and Goldman, Pandolf, and Santee, decimal for rest); L, external load (kg); M, body mass (kg); S, speed (m/min for ACSM, km/h for Givoni and Goldman, m/s for rest). Volunteers Nine (eight males and one female) active duty military personnel (mean ± SD; age, 21 ± 3 yr; height, 1.72 ± 0.07 m; body mass, 83.4 ± 12.9 kg; VO2 max, 47.8 ± 3.9 ml/kg/min) volunteered to participate in this investigation. All volunteers were medically cleared before participation. Selection criteria included general good health as determined during clearance procedures, passing their most recent Army Physical Fitness Test, and no recent history of musculoskeletal injury or other pre-existing conditions that would preclude their participation in the study. The study was approved by both the Scientific Review Committee and Institutional Review Board at the U.S. Army Research Institute of Environmental Medicine (USARIEM) (Natick, MA). The USARIEM Institutional Review Board was responsible for regulatory oversight of the investigation. Procedures Complex Terrain Course All testing was conducted on the Natick Soldier System Center Fitness Trail in Natick, MA (Fig. 1). The complex terrain course (~2.5 km) was divided into 40 segments based on distance (60.1 ± 45.6 m), grade (0.4 ± 6.6%), and surface factors. Terrain factors of 1.0 and 1.2 were assigned for paved road (49.9% of total distance) and dirt road (50.1% of total distance) segments of the course in accordance with the recommendations of Richmond et al10 The altitude of the course ranged between 42 and 52 m above sea level. Two Remote Automated Weather Stations collected air temperature and other weather parameters during data collection (WeatherHawk 510; WeatherHawk, and Campbell CR3000; Campbell Scientific Instruments, Logan, UT). In addition, natural wet-bulb, dry-bulb, and black globe temperatures, used to calculate the wet bulb-globe temperature index, were collected with the CR3000 and a QuesTemp34 wet bulb-globe temperature monitor (Quest Technologies, Oconomwoc, WI). Average environmental data included the following: wet-bulb temperature (16.9 ± 4.4°C), dry-bulb temperature (21.1 ± 4.9°C), globe temperature (29.5 ± 7.5°C), wet-bulb globe temperature (19.8 ± 4.8°C), and relative humidity (52 ± 11%). Figure 1. View largeDownload slide Overview of complex terrain course (Natick Soldier System Center Fitness Trail, Natick, MA). Figure 1. View largeDownload slide Overview of complex terrain course (Natick Soldier System Center Fitness Trail, Natick, MA). Familiarization During initial pack and load fitting, volunteers were familiarized with carrying both the 30% and 45% loads over a short distance (50–150 m) on a level surface. Volunteers participated in a familiarization session for exposure to testing equipment, procedures, and the complex terrain course before data collection. Upon arrival, volunteers were fitted to a portable oxygen uptake (VO2) monitor system (COSMED K4b2 (K4) COSMED Sri., Rome, Italy). The body-mounted elements of the system consisted of a Portable Unit, a battery, a face mask covering the nose and mouth, mounting harness, and connecting tubes and cables. Each volunteer then walked one clockwise lap around the complex terrain course while accompanied by a research team member. Volunteers were not fitted with other instrumentation during this familiarization. Test Visits Each test visit began with preliminary measurements and instrumentation before a ~150 m warm-up walk without an external load. Subsequently, volunteers were fitted with the portable VO2 monitor and donned either the 30% or 45% load. The 30% load included the Army Combat Uniform uniform with the Army Combat Shirt, personal boots, with a non-functional replica rifle, body armor, water, and additional load to simulate munitions, but no rucksack. Additional weight was added to the vest pouches if the Clothing and Individual Equipment described above weighed less than 30% body mass. To be relevant to military training and operations, all volunteers carried at least the equivalent of a fighting or combat load, which includes body armor. The 45% load also included a Modular Lightweight Load Carrying Equipment frame rucksack; additional weight was placed into the rucksack to increase the total load to 45%. After standing for 10 min to collect baseline VO2 data, volunteers were then required to complete a minimum of two and a maximum of four laps around the complex terrain course as fast as possible without running, jogging, or shuffling. Volunteers were provided 5-min rest periods between laps during which they were allowed to sit down, temporarily remove their masks, and drink water. Each volunteer was provided a radio-frequency identification tag (SportIdent; Scarborough Orienteering, Huntington Beach, CA), which communicated with timing gates (SportIdent; Scarborough Orienteering, Huntington Beach, CA) placed between segments to record completion times and speeds along the course. Volunteers completed laps 1 and 3 in the clockwise direction and laps 2 and 4 in the counterclockwise direction. Volunteers completed an average of 3.6 ± 0.8 laps per test session with four trials terminated early due to threshold constraints on environmental exposures and in one case due to a significant increase in core temperature and other indications of heat strain. However, all volunteers met the minimum requirement and completed at least one lap in each direction during a given test session. Statistical Analyses All data are displayed as mean ± standard deviation (SD) and analyzed using RStudio (Version 0.98.1056; RStudio, Inc, Boston, MA). Metabolic rate estimations and observations were compared over the overall course as well as individual course segments. Predictive equation performance was evaluated by the bias, residual SD, root mean square error (RMSE), and Pearson’s correlation coefficient (r). The external load was added to body mass for the physical activity component of the equations that did not specifically account for external loading (ACSM, Jobe et al, Minetti et al) in order to account for the linear increase in metabolic rate with additional load.14 Resting metabolic rates (mL VO2/kg/min) were calculated for the Jobe et al and Minetti et al equations using the following equations from Schofield et al15:   RestingVO2(Male,18–30yr)=2.177+(100.055⋅M−1)RestingVO2(Female,18–30yr)=2.141+(70.343⋅M−1) Results Figure 2 displays mean VO2 per segment over the first two laps for the 30% and 45% body mass load tests. The mean VO2 over the entire course was 22.0 ± 4.4 mL O2/kg/min. Table 2 summarizes the performance of each predictive equation in estimating metabolic rates for load carriage over complex terrain. Nearly all predictive equations produced estimates that were lower compared with the observed mean metabolic rate of all segments (655 ± 176 W) as well as the observed mean across all courses completed (638 ± 143 W). The Pandolf et al equation had comparable accuracy when predicting metabolic rates per course with and without the corrective factor of Santee et al (Bias, −51 ± 88 W and −13 ± 87 W, respectively). However, the Santee et al equation had a lower RMSE than the Pandolf et al equation when estimating overall metabolic rates per course (99 W vs. 102 W). Table II. Predictive Equation Performance (W) Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  r, Pearson’s correlation coefficient. Table II. Predictive Equation Performance (W) Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  r, Pearson’s correlation coefficient. Figure 2. View largeDownload slide Mean oxygen consumption (VO2) per segment over the first two laps for the complex terrain marches with 30% and 45% body mass loads. Solid lines, mean VO2 per segment; dotted lines, mean ± SD. Figure 2. View largeDownload slide Mean oxygen consumption (VO2) per segment over the first two laps for the complex terrain marches with 30% and 45% body mass loads. Solid lines, mean VO2 per segment; dotted lines, mean ± SD. Discussion The most salient finding of this investigation was that each of the predictive equations underestimated the metabolic costs of military load carriage over complex terrain. The adjusted Pandolf et al equation had the most accurate metabolic rate estimates among the six equations evaluated. However, these results clearly demonstrate the need for improvements to these methods, replacement with new methods, or the use of a multi-model approach to account for mixed terrain. Metabolic rate estimates from the predictive equations were considerably lower than the metabolic rates measured by the COSMED K4b2. Although many studies have shown Pandolf et al to be an accurate predictor of metabolic costs in the lab,16 some have highlighted specific discrepancies linked to external load configurations (e.g., heavy ballistic suits).17 This brings into question some of the other significant contributors to individualized effects, such as different proportional loading (% body mass), fitness levels (% VO2max), hydration status, or sex. It is unlikely that metabolic rates were higher than estimated values due to unfamiliar or unfit test volunteers as they were experienced with heavy load carriage and were aerobically fit (VO2 max, 47.8 ± 3.9 mL/kg/min). It is also possible that certain course segments may have been too brief to approximate the steady-state metabolic rates predicted by the equations. Alternatively, just as heart rates remain elevated after cresting a steep uphill grades, it is highly possible that the more physically demanding segments caused persistent metabolic rate elevations over subsequent segments. Another complication to this is the fact that the length of individual breathes vary from ~3–10 sec and it can take less than or only slightly more than this time for an individual to traverse the shorter trail segments, it was as expected that some lag or overshooting would occur in metabolic responses. The metabolic rates measured by the COSMED K4b2 may have been higher than true values. McLaughlin et al18 found that the COSMED K4b2 overestimated VO2 consumption during cycle ergometer exercise by 50–200 W when compared with the Douglas bag method. Similarly, Duffield et al19 found that the COSMED K4b2 system overestimated VO2 during intermittent treadmill exercise compared with a laboratory metabolic cart but noted that the COSMED K4b2 system demonstrated satisfactory test–retest reliability. In contrast, Schrack et al20 found that VO2 measurements by the COSMED K4b2 were not significantly different (p = 0.25) but highly correlated (r = 0.96) to a traditional, stationary gas exchange during steady-state walking. A consistent offset by the COSMED K4b2 would call into question whether the Pandolf et al and Santee et al equations, which had the smallest bias and RMSE, were the best predictors of the true metabolic rate. The correlation between estimates and observations would be a better indicator of predictive equation performance in this case. However, the Pandolf et al and Santee et al equations also had the highest correlations with the observed metabolic rate for the course (r = 0.79 and 0.80, respectively) as well as uphill (r = 0.78 for each) and downhill segments (r = 0.80 and 0.81, respectively). Consequently, the findings of this investigation support the use of the Pandolf et al equation with the correction factor from Santee et al as most suitable for estimating metabolic demands when traversing complex terrain courses. Combinations of rational and empirical models have been used to produce more diverse and accurate predictions for physiological prediction modeling.21 This approach could conceivably be outlined from the current work by simple model selection based on grade alone, where methods are applied based on the best performing equation for level or uphill grades (≥ 0%) and for downhill (<0%). For example, applying the equation from Jobe et al for positive-grade segments (RMSE = 102 W) and Santee et al for negative-grade segments (RMSE = 92 W) should provide more accurate estimation for the overall energy expenditure for a known course. As the terrain features themselves have significant impact on the energy demands, grade alone may be a simplistic concept but could be extended to include several models in this multi-model approach specifically targeted to more defined terrain and surface features. The present investigation has a few notable limitations. For instance, the sample size was only nine volunteers with a single female volunteer. However, it should be noted that comparable samples were used in many of the original load carriage studies22 including the datasets used to validate the Minetti et al13 (n = 10), Pandolf et al6 (n = 6), and Santee et al8 (n = 16) equations. Although all volunteers were directed to attempt four laps, not all were able to complete all of the laps during each test session. This was due to threshold constraints on environmental exposures (n = 4) and significantly increased core temperature (n = 1). However, all volunteers completed at least one lap in each direction during a given test session. Conclusion Current predictive equations underestimate the metabolic costs of military load carriage over complex terrain. From this study, a combination of the Pandolf et al equation with the correction factor from Santee et al is the most suitable method for estimating metabolic demands over complex terrain. However, the disagreement between predicted and observed values reported in this article indicates a need for improvement using possible methods such as a multi-model approach. Acknowledgment In addition, we would like to acknowledge the contributions of a number of individuals affiliated with the U.S. Army Institute of Environmental Medicine (USARIEM). Ms. Leila A. Walker, Mr. Stephen P. Mullen, Mr. Julio A. Gonzalez, Mr. Timothy P. Rioux, SSG Glen M. Rossman, SSG John E. Camelo, and Dr. Gary P Zientara each contributed to the planning, setup, and collection of data during the study without direct compensation. Funding This research was supported in part by appointments to the Postgraduate Research Participation Program at the U.S. Army Research Institute of Environmental Medicine (USARIEM) administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy and USAMRMC. References 1 Henning PC, Park BS, Kim JS: Physiological decrements during sustained military operational stress. Mil Med  2011; 176( 9): 991– 7. Google Scholar CrossRef Search ADS PubMed  2 Field Manual F. FM 3-06.11 Combined arms operations in urban terrain. Washington; 2002. 3 Dean C, DuPont F:The modern warrior’s combat load. Dismounted Operations in Afghanistan Report from the US Army Center for Army Lessons Learned. 2003. 4 Crowder TA, Beekley MD, Sturdivant RX, Johnson CA, Lumpkin A: Metabolic effects of soldier performance on a simulated graded road march while wearing two functionally equivalent military ensembles. Mil Med  2007; 172( 6): 596– 602. Google Scholar CrossRef Search ADS PubMed  5 Drain J, Billing D, Neesham-Smith D, Aisbett B: Predicting physiological capacity of human load carriage – a review. Appl Ergon  2016; 52: 85– 94. Google Scholar CrossRef Search ADS PubMed  6 Pandolf KB, Givoni B, Goldman RF: Predicting energy expenditure with loads while standing or walking very slowly. J Appl Physiol Respir Environ Exerc Physiol  1977; 43( 4): 577– 81. Google Scholar PubMed  7 Givoni B, Goldman RF: Predicting metabolic energy cost. J Appl Physiol  1971; 30( 3): 429– 33. Google Scholar CrossRef Search ADS PubMed  8 Santee WR, Blanchard LA, Speckman KL, Gonzalez JA, Wallace RF: Load Carriage Model Development and Testing with Field Data. Technical Note . Natick, MA, U.S. Army Research Institute of Environmental Medicine, Report No.: ADA#415788, 2003. 9 Weyand PG, Smith BR, Schultz NS, Ludlow LW, Puyau MR, Butte NF: Predicting metabolic rate across walking speed: one fit for all body sizes? J Appl Physiol (1985)  2013; 115( 9): 1332– 42. Google Scholar CrossRef Search ADS PubMed  10 Richmond PW, Potter AW, Santee WR: Terrain Factors for predicting walking and load carriage energy costs: review and refinement. J Sport Hum Perf  2015; 3( 3): 1– 26. 11 American College of Sports Medicine: Guidelines for Graded Exercise Testing and Prescription , 7th Ed, Philadelphia, PA, Lippincott, Williams and Wilkins, 2006. 12 Jobe RT, White PS: A new cost-distance model for human accessibility and an evaluation of accessibility bias in permanent vegetation plots in Great Smoky Mountains National Park, USA. J Veg Sci  2009; 20( 6): 1099– 109. Google Scholar CrossRef Search ADS   13 Minetti AE, Moia C, Roi GS, Susta D, Ferretti G: Energy cost of walking and running at extreme uphill and downhill slopes. J Appl Physiol (1985)  2002; 93( 3): 1039– 46. Google Scholar CrossRef Search ADS PubMed  14 Beekley MD, Alt J, Buckley CM, Duffey M, Crowder TA: Effects of heavy load carriage during constant-speed, simulated, road marching. Mil Med  2007; 172( 6): 592– 5. Google Scholar CrossRef Search ADS PubMed  15 Schofield WN: Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr Clin Nutr  1985; 39( Suppl 1): 5– 41. Google Scholar PubMed  16 Hall C, Figueroa A, Fernhall B, Kanaley JA: Energy expenditure of walking and running: comparison with prediction equations. Med Sci Sports Exerc  2004; 36( 12): 2128– 34. Google Scholar CrossRef Search ADS PubMed  17 Bach AJ, Costello JT, Borg DN, Stewart IB: The Pandolf load carriage equation is a poor predictor of metabolic rate while wearing explosive ordnance disposal protective clothing. Ergonomics  2017; 60( 3): 430– 8. Google Scholar CrossRef Search ADS PubMed  18 McLaughlin JE, King GA, Howley ET, Bassett DR Jr, Ainsworth BE: Validation of the COSMED K4 b2 portable metabolic system. Int J Sports Med  2001; 22( 4): 280– 4. Google Scholar CrossRef Search ADS PubMed  19 Duffield R, Dawson B, Pinnington HC, Wong P: Accuracy and reliability of a Cosmed K4b2 portable gas analysis system. J Sci Med Sport  2004; 7( 1): 11– 22. Google Scholar CrossRef Search ADS PubMed  20 Schrack JA, Simonsick EM, Ferrucci L: Comparison of the Cosmed K4b 2 portable metabolic system in measuring steady-state walking energy expenditure. PLoS ONE  2010; 5( 2): e9292. Google Scholar CrossRef Search ADS PubMed  21 Xu X, Santee WR: Sweat loss prediction using a multi-model approach. Int J Biometeorol  2011; 55( 4): 501– 8. Google Scholar CrossRef Search ADS PubMed  22 Brainerd ST: A Compendium of Portage Studies . Aberdeen Research and Development Center, Aberdeen Proving Ground, MD, Human Engineering Lab, 1982. Published by Oxford University Press on behalf of Association of Military Surgeons of the United States 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Military Medicine Oxford University Press

Loading next page...
 
/lp/ou_press/metabolic-costs-of-military-load-carriage-over-complex-terrain-DSyD3gcaKl
Publisher
Association of Military Surgeons of the United States
Copyright
Published by Oxford University Press on behalf of Association of Military Surgeons of the United States 2018.
ISSN
0026-4075
eISSN
1930-613X
D.O.I.
10.1093/milmed/usx099
Publisher site
See Article on Publisher Site

Abstract

Abstract Introduction Dismounted military operations often involve prolonged load carriage over complex terrain, which can result in excessive metabolic costs that can directly impair soldiers’ performance. Although estimating these demands is a critical interest for mission planning purposes, it is unclear whether existing estimation equations developed from controlled laboratory- and field-based studies accurately account for energy costs of traveling over complex terrain. This study investigated the accuracy of the following equations for military populations when applied to data collected over complex terrain with two different levels of load carriage: American College of Sports Medicine (2002), Givoni and Goldman (1971), Jobe and White (2009), Minetti et al (2002), Pandolf et al (1977), and Santee et al (2003). Materials and Methods Nine active duty military personnel (age 21 ± 3 yr; height 1.72 ± 0.07 m; body mass 83.4 ± 12.9 kg; VO2 max 47.8 ± 3.9 mL/kg/min) were monitored during load carriage (with loads equal to 30% and 45% of body mass) over a 10-km mixed terrain course on two separate test days. The course was divided into four 2.5-km laps of 40 segments based on distance, grade, and/or surface factors. Timing gates and radio-frequency identification cards (SportIdent; Scarborough Orienteering, Huntington Beach, CA) were used to record completion times for each course segment. Breath-by-breath measures of energy expenditure were collected using portable oxygen exchange devices (COSMED Sri., Rome, Italy) and compared model estimates. Results The Santee et al equation performed best, demonstrating the smallest estimation bias (−13 ± 87 W) and lowest root mean square error (99 W). Conclusion Current predictive equations underestimate the metabolic cost of load carriage by military personnel over complex terrain. Applying the Santee et al correction factor to the Pandolf et al equation may be the most suitable approach for estimating metabolic demands in these circumstances. However, this work also outlines the need for improvements to these methods, new method development and validation, or the use of a multi-model approach to account for mixed terrain. Introduction Military operations often require dismounted personnel to carry heavy loads over long distances.1 Although loads of 30% body mass or less are recommended to avoid performance decrements,2 a previous study of combat operations in Afghanistan showed that the average light infantry rifleman carried over 43 kg of critical equipment during approach marches.3 Although additional equipment increases versatility, heavy load carriage significantly increases energy costs.4 Excessive energy expenditure can degrade soldier performance, contribute to thermal strain, as well as lead to physical and cognitive fatigue. Therefore, the ability to accurately predict energy costs is important for effectively planning soldier activities. Practitioners often rely on predictive equations to estimate the metabolic costs of military load carriage.5 Many of the well-established predictive equations were developed from laboratory steady-state exercise protocols that controlled for metabolic rate moderators such as walking speed, treadmill grade, and external load.6–9 An inherent assumption is that predictive equations developed from steady-state data can be used to predict energy costs under dynamic conditions where activities change frequently enough that equilibrium cannot be achieved. Military examples of dynamic conditions include dismounted patrols and approach marches that often require traversing landscapes with shifting gradients and surfaces.10 As these conditions cannot be realistically replicated in laboratory settings, field research data are necessary to determine whether these predictive equations accurately estimate the metabolic costs of military load carriage over complex terrain. However, there has not yet been a direct comparison between predictive equations using metabolic data collected over complex terrain. These data will provide a basis for realistic expectations regarding the accuracy of metabolic cost predictions for load carriage. This investigation sought to evaluate how well-established predictive equations estimate the metabolic costs of military load carriage over complex terrain and to identify the most suitable equations for these estimations. Ultimately, the findings of this investigation lay the ground work for future efforts to refine, hybridize, or apply multi-model approaches to improving the accuracy of metabolic cost estimation methods. Materials and Methods Study Design Test volunteers completed two marches over a complex terrain course carrying a different external load (either an additional 30% or 45% body mass load). The marches were separated by a minimum of 5 d and the order of the external load conditions was randomized. Six predictive equations6–8,11–13 (Table 1) were used to estimate metabolic rate over course segments and evaluated against measurements from a portable oxygen uptake monitor system (COSMED K4b2 (K4) COSMED Sri., Rome, Italy). The Santee et al equation used in this investigation includes a multiplication sign following the S2 term instead of the minus sign that was erroneously included in the original published equation. Table I. Predictive Equations from Previous Literature Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  G, grade (% for Givoni and Goldman, Pandolf, and Santee, decimal for rest); L, external load (kg); M, body mass (kg); S, speed (m/min for ACSM, km/h for Givoni and Goldman, m/s for rest). Table I. Predictive Equations from Previous Literature Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  Reference  Equation  ACSM11  (mlO2·kg−1·min−1)=0.1⋅S+1.8⋅S⋅G+3.5  Givoni and Goldman7  (kcal⋅hr−1)=η⋅(M+L)⋅[2.3+0.32⋅(S–2.5)1.65+G(0.2+0.07⋅(S−2.5))]  Jobe and White12  (J⋅kg−1⋅m−1)=20.9⋅tan(G)2+4.18⋅tan(G)+1.38 * −60 < G < −6  (J⋅kg−1⋅m−1)=52.1⋅10−3⋅tan(G)2+10.4⋅tan(G)+2.65 * −6 < G < 60  Minetti et al20  (J⋅kg−1⋅m−1)=280.5⋅G5−58.7⋅G4−76.8⋅G3+51.9⋅G2+19.6⋅G+2.5  Pandolf et al6  (W)=1.5⋅M+2.0⋅(M+L)+η⋅(M+L)⋅(1.5⋅S2+0.35⋅S⋅G)  Santee et al8  (W)=−η⋅[(G⋅(M+L)⋅S)/3.5−((M+L)(⋅(G+6)2)/M)+(25⋅S2)]  *Added to Pandolf et al.6 when G < 0  G, grade (% for Givoni and Goldman, Pandolf, and Santee, decimal for rest); L, external load (kg); M, body mass (kg); S, speed (m/min for ACSM, km/h for Givoni and Goldman, m/s for rest). Volunteers Nine (eight males and one female) active duty military personnel (mean ± SD; age, 21 ± 3 yr; height, 1.72 ± 0.07 m; body mass, 83.4 ± 12.9 kg; VO2 max, 47.8 ± 3.9 ml/kg/min) volunteered to participate in this investigation. All volunteers were medically cleared before participation. Selection criteria included general good health as determined during clearance procedures, passing their most recent Army Physical Fitness Test, and no recent history of musculoskeletal injury or other pre-existing conditions that would preclude their participation in the study. The study was approved by both the Scientific Review Committee and Institutional Review Board at the U.S. Army Research Institute of Environmental Medicine (USARIEM) (Natick, MA). The USARIEM Institutional Review Board was responsible for regulatory oversight of the investigation. Procedures Complex Terrain Course All testing was conducted on the Natick Soldier System Center Fitness Trail in Natick, MA (Fig. 1). The complex terrain course (~2.5 km) was divided into 40 segments based on distance (60.1 ± 45.6 m), grade (0.4 ± 6.6%), and surface factors. Terrain factors of 1.0 and 1.2 were assigned for paved road (49.9% of total distance) and dirt road (50.1% of total distance) segments of the course in accordance with the recommendations of Richmond et al10 The altitude of the course ranged between 42 and 52 m above sea level. Two Remote Automated Weather Stations collected air temperature and other weather parameters during data collection (WeatherHawk 510; WeatherHawk, and Campbell CR3000; Campbell Scientific Instruments, Logan, UT). In addition, natural wet-bulb, dry-bulb, and black globe temperatures, used to calculate the wet bulb-globe temperature index, were collected with the CR3000 and a QuesTemp34 wet bulb-globe temperature monitor (Quest Technologies, Oconomwoc, WI). Average environmental data included the following: wet-bulb temperature (16.9 ± 4.4°C), dry-bulb temperature (21.1 ± 4.9°C), globe temperature (29.5 ± 7.5°C), wet-bulb globe temperature (19.8 ± 4.8°C), and relative humidity (52 ± 11%). Figure 1. View largeDownload slide Overview of complex terrain course (Natick Soldier System Center Fitness Trail, Natick, MA). Figure 1. View largeDownload slide Overview of complex terrain course (Natick Soldier System Center Fitness Trail, Natick, MA). Familiarization During initial pack and load fitting, volunteers were familiarized with carrying both the 30% and 45% loads over a short distance (50–150 m) on a level surface. Volunteers participated in a familiarization session for exposure to testing equipment, procedures, and the complex terrain course before data collection. Upon arrival, volunteers were fitted to a portable oxygen uptake (VO2) monitor system (COSMED K4b2 (K4) COSMED Sri., Rome, Italy). The body-mounted elements of the system consisted of a Portable Unit, a battery, a face mask covering the nose and mouth, mounting harness, and connecting tubes and cables. Each volunteer then walked one clockwise lap around the complex terrain course while accompanied by a research team member. Volunteers were not fitted with other instrumentation during this familiarization. Test Visits Each test visit began with preliminary measurements and instrumentation before a ~150 m warm-up walk without an external load. Subsequently, volunteers were fitted with the portable VO2 monitor and donned either the 30% or 45% load. The 30% load included the Army Combat Uniform uniform with the Army Combat Shirt, personal boots, with a non-functional replica rifle, body armor, water, and additional load to simulate munitions, but no rucksack. Additional weight was added to the vest pouches if the Clothing and Individual Equipment described above weighed less than 30% body mass. To be relevant to military training and operations, all volunteers carried at least the equivalent of a fighting or combat load, which includes body armor. The 45% load also included a Modular Lightweight Load Carrying Equipment frame rucksack; additional weight was placed into the rucksack to increase the total load to 45%. After standing for 10 min to collect baseline VO2 data, volunteers were then required to complete a minimum of two and a maximum of four laps around the complex terrain course as fast as possible without running, jogging, or shuffling. Volunteers were provided 5-min rest periods between laps during which they were allowed to sit down, temporarily remove their masks, and drink water. Each volunteer was provided a radio-frequency identification tag (SportIdent; Scarborough Orienteering, Huntington Beach, CA), which communicated with timing gates (SportIdent; Scarborough Orienteering, Huntington Beach, CA) placed between segments to record completion times and speeds along the course. Volunteers completed laps 1 and 3 in the clockwise direction and laps 2 and 4 in the counterclockwise direction. Volunteers completed an average of 3.6 ± 0.8 laps per test session with four trials terminated early due to threshold constraints on environmental exposures and in one case due to a significant increase in core temperature and other indications of heat strain. However, all volunteers met the minimum requirement and completed at least one lap in each direction during a given test session. Statistical Analyses All data are displayed as mean ± standard deviation (SD) and analyzed using RStudio (Version 0.98.1056; RStudio, Inc, Boston, MA). Metabolic rate estimations and observations were compared over the overall course as well as individual course segments. Predictive equation performance was evaluated by the bias, residual SD, root mean square error (RMSE), and Pearson’s correlation coefficient (r). The external load was added to body mass for the physical activity component of the equations that did not specifically account for external loading (ACSM, Jobe et al, Minetti et al) in order to account for the linear increase in metabolic rate with additional load.14 Resting metabolic rates (mL VO2/kg/min) were calculated for the Jobe et al and Minetti et al equations using the following equations from Schofield et al15:   RestingVO2(Male,18–30yr)=2.177+(100.055⋅M−1)RestingVO2(Female,18–30yr)=2.141+(70.343⋅M−1) Results Figure 2 displays mean VO2 per segment over the first two laps for the 30% and 45% body mass load tests. The mean VO2 over the entire course was 22.0 ± 4.4 mL O2/kg/min. Table 2 summarizes the performance of each predictive equation in estimating metabolic rates for load carriage over complex terrain. Nearly all predictive equations produced estimates that were lower compared with the observed mean metabolic rate of all segments (655 ± 176 W) as well as the observed mean across all courses completed (638 ± 143 W). The Pandolf et al equation had comparable accuracy when predicting metabolic rates per course with and without the corrective factor of Santee et al (Bias, −51 ± 88 W and −13 ± 87 W, respectively). However, the Santee et al equation had a lower RMSE than the Pandolf et al equation when estimating overall metabolic rates per course (99 W vs. 102 W). Table II. Predictive Equation Performance (W) Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  r, Pearson’s correlation coefficient. Table II. Predictive Equation Performance (W) Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  Grade  Equation  Segment  Course  Bias ± SD  RMSE  r  Bias ± SD  RMSE  r  All  ACSM11  −163 ± 507  532  0.22  −140 ± 104  174  0.70  Givoni and Goldman7  −127 ± 359  381  0.32  −111 ± 90  143  0.78  Jobe et al12  −87 ± 181  201  0.46  −73 ± 100  124  0.73  Minetti et al20  −69 ± 262  271  0.39  −74 ± 100  124  0.73  Pandolf et al6  −72 ± 463  468  0.29  −51 ± 88  102  0.79  Santee et al8  −38 ± 313  315  0.44  −13 ± 87  99  0.80  >0%  ACSM11  112 ± 395  410  0.39  53 ± 107  120  0.69  Givoni and Goldman7  57 ± 295  300  0.47  15 ± 99  105  0.77  Jobe et al12  −27 ± 162  164  0.57  −19 ± 101  102  0.71  Minetti et al20  49 ± 283  288  0.44  20 ± 101  103  0.71  Pandolf et al6  −92 ± 138  166  0.65  102 ± 102  144  0.78  Santee et al8  −92 ± 138  166  0.65  175 ± 370  409  0.46  <0%  ACSM11  −435 ± 456  630  −0.03  −333 ± 115  352  0.68  Givoni and Goldman7  309 ± 322  447  0.11  −221 ± 85  237  0.80  Jobe et al12  −146 ± 181  232  0.32  −127 ± 101  162  0.73  Minetti et al20  −186 ± 173  254  0.34  −168 ± 103  197  0.74  Pandolf et al6  −314 ± 412  520  0.06  −202 ± 87  219  0.80  Santee et al8  −98 ± 147  177  0.56  −75 ± 83  92  0.81  r, Pearson’s correlation coefficient. Figure 2. View largeDownload slide Mean oxygen consumption (VO2) per segment over the first two laps for the complex terrain marches with 30% and 45% body mass loads. Solid lines, mean VO2 per segment; dotted lines, mean ± SD. Figure 2. View largeDownload slide Mean oxygen consumption (VO2) per segment over the first two laps for the complex terrain marches with 30% and 45% body mass loads. Solid lines, mean VO2 per segment; dotted lines, mean ± SD. Discussion The most salient finding of this investigation was that each of the predictive equations underestimated the metabolic costs of military load carriage over complex terrain. The adjusted Pandolf et al equation had the most accurate metabolic rate estimates among the six equations evaluated. However, these results clearly demonstrate the need for improvements to these methods, replacement with new methods, or the use of a multi-model approach to account for mixed terrain. Metabolic rate estimates from the predictive equations were considerably lower than the metabolic rates measured by the COSMED K4b2. Although many studies have shown Pandolf et al to be an accurate predictor of metabolic costs in the lab,16 some have highlighted specific discrepancies linked to external load configurations (e.g., heavy ballistic suits).17 This brings into question some of the other significant contributors to individualized effects, such as different proportional loading (% body mass), fitness levels (% VO2max), hydration status, or sex. It is unlikely that metabolic rates were higher than estimated values due to unfamiliar or unfit test volunteers as they were experienced with heavy load carriage and were aerobically fit (VO2 max, 47.8 ± 3.9 mL/kg/min). It is also possible that certain course segments may have been too brief to approximate the steady-state metabolic rates predicted by the equations. Alternatively, just as heart rates remain elevated after cresting a steep uphill grades, it is highly possible that the more physically demanding segments caused persistent metabolic rate elevations over subsequent segments. Another complication to this is the fact that the length of individual breathes vary from ~3–10 sec and it can take less than or only slightly more than this time for an individual to traverse the shorter trail segments, it was as expected that some lag or overshooting would occur in metabolic responses. The metabolic rates measured by the COSMED K4b2 may have been higher than true values. McLaughlin et al18 found that the COSMED K4b2 overestimated VO2 consumption during cycle ergometer exercise by 50–200 W when compared with the Douglas bag method. Similarly, Duffield et al19 found that the COSMED K4b2 system overestimated VO2 during intermittent treadmill exercise compared with a laboratory metabolic cart but noted that the COSMED K4b2 system demonstrated satisfactory test–retest reliability. In contrast, Schrack et al20 found that VO2 measurements by the COSMED K4b2 were not significantly different (p = 0.25) but highly correlated (r = 0.96) to a traditional, stationary gas exchange during steady-state walking. A consistent offset by the COSMED K4b2 would call into question whether the Pandolf et al and Santee et al equations, which had the smallest bias and RMSE, were the best predictors of the true metabolic rate. The correlation between estimates and observations would be a better indicator of predictive equation performance in this case. However, the Pandolf et al and Santee et al equations also had the highest correlations with the observed metabolic rate for the course (r = 0.79 and 0.80, respectively) as well as uphill (r = 0.78 for each) and downhill segments (r = 0.80 and 0.81, respectively). Consequently, the findings of this investigation support the use of the Pandolf et al equation with the correction factor from Santee et al as most suitable for estimating metabolic demands when traversing complex terrain courses. Combinations of rational and empirical models have been used to produce more diverse and accurate predictions for physiological prediction modeling.21 This approach could conceivably be outlined from the current work by simple model selection based on grade alone, where methods are applied based on the best performing equation for level or uphill grades (≥ 0%) and for downhill (<0%). For example, applying the equation from Jobe et al for positive-grade segments (RMSE = 102 W) and Santee et al for negative-grade segments (RMSE = 92 W) should provide more accurate estimation for the overall energy expenditure for a known course. As the terrain features themselves have significant impact on the energy demands, grade alone may be a simplistic concept but could be extended to include several models in this multi-model approach specifically targeted to more defined terrain and surface features. The present investigation has a few notable limitations. For instance, the sample size was only nine volunteers with a single female volunteer. However, it should be noted that comparable samples were used in many of the original load carriage studies22 including the datasets used to validate the Minetti et al13 (n = 10), Pandolf et al6 (n = 6), and Santee et al8 (n = 16) equations. Although all volunteers were directed to attempt four laps, not all were able to complete all of the laps during each test session. This was due to threshold constraints on environmental exposures (n = 4) and significantly increased core temperature (n = 1). However, all volunteers completed at least one lap in each direction during a given test session. Conclusion Current predictive equations underestimate the metabolic costs of military load carriage over complex terrain. From this study, a combination of the Pandolf et al equation with the correction factor from Santee et al is the most suitable method for estimating metabolic demands over complex terrain. However, the disagreement between predicted and observed values reported in this article indicates a need for improvement using possible methods such as a multi-model approach. Acknowledgment In addition, we would like to acknowledge the contributions of a number of individuals affiliated with the U.S. Army Institute of Environmental Medicine (USARIEM). Ms. Leila A. Walker, Mr. Stephen P. Mullen, Mr. Julio A. Gonzalez, Mr. Timothy P. Rioux, SSG Glen M. Rossman, SSG John E. Camelo, and Dr. Gary P Zientara each contributed to the planning, setup, and collection of data during the study without direct compensation. Funding This research was supported in part by appointments to the Postgraduate Research Participation Program at the U.S. Army Research Institute of Environmental Medicine (USARIEM) administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy and USAMRMC. References 1 Henning PC, Park BS, Kim JS: Physiological decrements during sustained military operational stress. Mil Med  2011; 176( 9): 991– 7. Google Scholar CrossRef Search ADS PubMed  2 Field Manual F. FM 3-06.11 Combined arms operations in urban terrain. Washington; 2002. 3 Dean C, DuPont F:The modern warrior’s combat load. Dismounted Operations in Afghanistan Report from the US Army Center for Army Lessons Learned. 2003. 4 Crowder TA, Beekley MD, Sturdivant RX, Johnson CA, Lumpkin A: Metabolic effects of soldier performance on a simulated graded road march while wearing two functionally equivalent military ensembles. Mil Med  2007; 172( 6): 596– 602. Google Scholar CrossRef Search ADS PubMed  5 Drain J, Billing D, Neesham-Smith D, Aisbett B: Predicting physiological capacity of human load carriage – a review. Appl Ergon  2016; 52: 85– 94. Google Scholar CrossRef Search ADS PubMed  6 Pandolf KB, Givoni B, Goldman RF: Predicting energy expenditure with loads while standing or walking very slowly. J Appl Physiol Respir Environ Exerc Physiol  1977; 43( 4): 577– 81. Google Scholar PubMed  7 Givoni B, Goldman RF: Predicting metabolic energy cost. J Appl Physiol  1971; 30( 3): 429– 33. Google Scholar CrossRef Search ADS PubMed  8 Santee WR, Blanchard LA, Speckman KL, Gonzalez JA, Wallace RF: Load Carriage Model Development and Testing with Field Data. Technical Note . Natick, MA, U.S. Army Research Institute of Environmental Medicine, Report No.: ADA#415788, 2003. 9 Weyand PG, Smith BR, Schultz NS, Ludlow LW, Puyau MR, Butte NF: Predicting metabolic rate across walking speed: one fit for all body sizes? J Appl Physiol (1985)  2013; 115( 9): 1332– 42. Google Scholar CrossRef Search ADS PubMed  10 Richmond PW, Potter AW, Santee WR: Terrain Factors for predicting walking and load carriage energy costs: review and refinement. J Sport Hum Perf  2015; 3( 3): 1– 26. 11 American College of Sports Medicine: Guidelines for Graded Exercise Testing and Prescription , 7th Ed, Philadelphia, PA, Lippincott, Williams and Wilkins, 2006. 12 Jobe RT, White PS: A new cost-distance model for human accessibility and an evaluation of accessibility bias in permanent vegetation plots in Great Smoky Mountains National Park, USA. J Veg Sci  2009; 20( 6): 1099– 109. Google Scholar CrossRef Search ADS   13 Minetti AE, Moia C, Roi GS, Susta D, Ferretti G: Energy cost of walking and running at extreme uphill and downhill slopes. J Appl Physiol (1985)  2002; 93( 3): 1039– 46. Google Scholar CrossRef Search ADS PubMed  14 Beekley MD, Alt J, Buckley CM, Duffey M, Crowder TA: Effects of heavy load carriage during constant-speed, simulated, road marching. Mil Med  2007; 172( 6): 592– 5. Google Scholar CrossRef Search ADS PubMed  15 Schofield WN: Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr Clin Nutr  1985; 39( Suppl 1): 5– 41. Google Scholar PubMed  16 Hall C, Figueroa A, Fernhall B, Kanaley JA: Energy expenditure of walking and running: comparison with prediction equations. Med Sci Sports Exerc  2004; 36( 12): 2128– 34. Google Scholar CrossRef Search ADS PubMed  17 Bach AJ, Costello JT, Borg DN, Stewart IB: The Pandolf load carriage equation is a poor predictor of metabolic rate while wearing explosive ordnance disposal protective clothing. Ergonomics  2017; 60( 3): 430– 8. Google Scholar CrossRef Search ADS PubMed  18 McLaughlin JE, King GA, Howley ET, Bassett DR Jr, Ainsworth BE: Validation of the COSMED K4 b2 portable metabolic system. Int J Sports Med  2001; 22( 4): 280– 4. Google Scholar CrossRef Search ADS PubMed  19 Duffield R, Dawson B, Pinnington HC, Wong P: Accuracy and reliability of a Cosmed K4b2 portable gas analysis system. J Sci Med Sport  2004; 7( 1): 11– 22. Google Scholar CrossRef Search ADS PubMed  20 Schrack JA, Simonsick EM, Ferrucci L: Comparison of the Cosmed K4b 2 portable metabolic system in measuring steady-state walking energy expenditure. PLoS ONE  2010; 5( 2): e9292. Google Scholar CrossRef Search ADS PubMed  21 Xu X, Santee WR: Sweat loss prediction using a multi-model approach. Int J Biometeorol  2011; 55( 4): 501– 8. Google Scholar CrossRef Search ADS PubMed  22 Brainerd ST: A Compendium of Portage Studies . Aberdeen Research and Development Center, Aberdeen Proving Ground, MD, Human Engineering Lab, 1982. Published by Oxford University Press on behalf of Association of Military Surgeons of the United States 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US.

Journal

Military MedicineOxford University Press

Published: May 31, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off