This randomised controlled trial assigned participants into one of the four dietary supplementation interventions: no nutritional supplement control (CON), carbohydrate placebo (PLA), moderate protein [20 g additional per day (MOD)] or high protein [60 g additional per day (HIGH)]. Data were collected at the start (weeks 1–2), middle (week 6) and end (week 12) of British Army BT. Ethical approval was granted by the United Kingdom Ministry of Defence Research Ethics Committee (1076/MODREC/20) and was conducted in accordance with the Declaration of Helsinki.

In total, 99 men and 23 women [mean ± standard deviation (SD): age: 21.3 ± 3.5 years, height: 174.8 ± 8.4 cm, body mass 75.4 ± 12.2 kg] from four separate intakes between May 2021 and May 2022 completed the study. Participants were randomly allocated into a study condition (CON: n = 26; PLA: n = 30; MOD: n = 32; and HIGH: n = 34) using a random number generator. Participants were verbally briefed and provided written-informed consent in week 1 of training at the Army training Centre (Pirbright, Surrey, United Kingdom). Participants were included if they were not taking any other nutritional supplement and were not lactose or dairy intolerant. All participants were passed medically fit to train. A copy of the training programme is not available but previous work in this population has provided a general overview of training activities (1, 4).

A priori power analysis was calculated using G*power (Dusseldorf, V 3.1) for a within-and-between factor difference in maximal lower body strength based previously collected data in soldiers receiving a mixed-macronutrient supplement (16). The pre-selected α was 0.05 and β was 0.95. From this calculation, it was determined that a total of 88 participants were required to replicate a partial eta squared (η_p²) of 0.05 (medium).

Participants were administered supplementation each weekday evening between 20:00 and 21:00 h in powder form, mixed fresh with ~500 mL water by a member of the research team. The supplements were isocaloric to isolate the effects of the additional protein intake. The HIGH protein group were provided with 80 g of protein powder, containing 60 g of protein in the form of a concentrate and isolate whey protein blend (Max whey, Maximuscle, Nantwich, United Kingdom). The MOD protein group were provided with the same whey protein blend with a 27 g fixed amount of protein powder which provided 20 g of protein (Maximuscle, max whey, Nantwich, United Kingdom), mixed with 55 g of maltodextrin (Myprotein, 100% maltodextrin carbohydrate, Nantwich, United Kingdom). The PLA group were provided with 35 g of maltodextrin (Myprotein, 100% maltodextrin carbohydrate, Nantwich, United Kingdom), which was bulked up with 35 g of corn flour (buywholefoodsonline, Ramsgate, United Kingdom) and 10 g of Nesquik powder (Nestle, Welwyn Garden city, United Kingdom) The PLA supplement was also flavoured with chocolate and strawberry flavour drops (Myprotein, Flavdrops, Nantwich, United Kingdom) The specific energy and macronutrient breakdown of each supplement was PLA: energy = 286 kcal, CHO = 67 g, protein = 2 g, fat = 1 g; MOD: energy = 286 kcal, CHO = 48 g, protein = 20 g, fat = 2 g; HIGH: energy = 296 kcal, CHO = 8 g, protein = 60 g, and fat = 3 g. The amino acid breakdown of each whey protein supplement was HIGH: leucine = 6.3 g; isoleucine = 4.0 g; valine = 3.8 g; glutamic acid = 10 g; MOD: leucine = 2.1 g; isoleucine = 1.3 g; valine = 1.3 g; and glutamic acid = 3.3 g.

Energy and macronutrient intake was collected using a self-reported food diary assessed over 4 consecutive days (3 weekdays and 1 weekend day) during weeks 1–2, week 6 and week 12. Participants were verbally instructed to estimate portion size using standardised measures (e.g., one cup, two handfuls, and one palm size) (21), alongside storing any snack or ration discards (i.e., wrappers and packets) to help improve the accuracy of this approach (16). Daily food diaries were checked by a member of the research team the following day to assure compliance and review any unclear information. Food diary records were entered into nutritional analysis software (version 5, Nutritics, Dublin, Ireland) to generate mean daily energy and nutrient intakes using the United Kingdom Scientific Advisory Committee on Nutrition database. Recipes that did not exist in the database (i.e., ration pack foods) were manually entered using the recipe or nutritional content information provided by the caterer.

Prior to the start of collection, participants were issued 3 L urine containers and verbally instructed to collect all urine during a specified 24-h period except the first morning void in line with previously reported research (6, 22–26). Urine was collected over 3 separate non-consecutive days during weeks 1–2, week 6 and week 12 of BT. Urine samples were stored at ambient temperature during the 24-h collection period. Once total urine volume had been recorded, 2 mL aliquots were frozen at −80°C until subsequent analysis.

Urinary urea nitrogen excretion was assessed to determine nitrogen balance using the following equation (6, 27) with protein intake (g) assumed to be 16% nitrogen and miscellaneous nitrogen excretion assumed to be 4 g (28): Nitrogen balance g = total nitrogen intake g : urinary urea nitrogen g + 4

All urinary urea analysis was conducted at the Core Biochemical Assay Laboratory, Addenbrookes Hospital, Cambridge, United Kingdom. Urinary nitrogen was determined enzymatically with a clinical chemistry system (Dimension Flex, Siemens Healthcare Diagnostics Ltd., Camberley, United Kingdom). The method employs a urase/glutamate dehydrogenase coupled technique. Urease specifically hydrolyses urea to form ammonia and carbon dioxide. The ammonia is used by the enzyme glutamate dehydrogenase to reductively aminate α-ketoglutarate with simultaneous oxidation of reduced nicotinamide-adenine dinucleotide. The change in absorption due to the disappearance of nicotinamide adenine dinucleotide (NADH) is directly proportional to the Blood Urea Nitrogen (BUN) concentration in the sample. The intra-and-inter-assay coefficient of variations were 1.2 and 8.8%, respectively.

Participants were provided with wrist-based Actigraph watches (WGT3X-BT, Actigraph, Pensacola, United States) on the evening prior to start of each data collection period. In accordance with previous military activity monitoring research, participants were asked to wear the watch on their non-dominant wrist for 24-h and on the same 4 days energy intake was assessed (29). Each participant’s age, stature, and mass (Seca, Hamburg, Germany) were entered into the ActiLife software (v6.13.4, Actigraph, Florida, United States). The Actigraph estimated daily physical activity and duration of time spent in moderate [2,090–6,166 counts per min (cpm)], vigorous (6,167–9,642 cpm), and very vigorous (>9,642 cpm) activity at a 30 Hz sampling rate (30). These activity categories were then be assigned a caloric conversion using the Cooper institutes MET (moderate = 3 METs, vigorous = 6 METs, and very vigorous = 9 METs) to estimate activity energy expenditure (31): Activity energy expenditure kcal = MET value × body mass kg × time min

Resting energy expenditure (REE) was estimated using the standardised modified Harris-Benedict equation (32) as used previously for military recruits (29): R E E kcal = 88.362 + 13.397 × body mass kg + 4.799 × height cm : 5.677 × age years

Overall daily expenditure was the sum of activity energy expenditure (AEE) and REE. This method has been used to assess the daily energy expenditure of Army recruits undergoing BT in the United States (29).

Saliva was collected from participants within 1 h upon awakening in the morning (06:00 h) prior to breakfast and cleaning teeth. Participants placed an Oral Fluid Collector (OFC) swab (Soma Bioscience, Wallingford, United Kingdom) in their mouth on top of the tongue for ~2 min whilst refraining from talking. The swab was then inserted into a swab storage buffer and stored at 4°C. The samples were then transported to Soma Bioscience, Wallingford for analysis using an Enzyme-Linked Immunosorbent Assay (ELISA) for total testosterone and cortisol. For cortisol and testosterone, a 96-well microlitre plate was coated with Rabbit anti-cortisol antibody (cortisol) or goat anti-mouse IgG (testosterone) at 5 μg·mL⁻¹ in phosphate buffered saline and incubated for 90 min at 37°C. The plate was then washed three times with the assay buffer (PBS + 5 mg/mL bovine serum albumin) and blocked with an assay buffer, with 300 μL added per well for 30 min. Duplicate samples (intra-sample CV < 10%) were then added at volumes of 100 μL per well. Horse Radish Peroxide was added (100 μL per well) at a 1:15,000 (cortisol) or 1:18,000 (testosterone) dilution and mixed for 15 min. For testosterone, a monoclonal anti-testosterone antibody was added at 50 ng·mL⁻¹ into the assay buffer at 50 μL per well and incubated for 90 min at room temperature (37°C) on a plate shaker. The plates were then washed three times with the assay buffer and Tetramethylbenzidine then added (100 μL) to each well. Following incubation in the dark for 40 min, stop colour development was completed by adding sulphuric acid (100 μL per well) and the resultant optical density read at 450 nm. The intra and inter assay CVs were 7.9 and 9.4% for testosterone and 7.9 and 13.1% for cortisol, respectively.

A whole-body scan was performed for the measurement of body composition using Dual X-ray Absorptiometry (DXA; Lunar iDXA, GE Healthcare, United Kingdom) in weeks 1 and 12. Each scan was performed with participants rested and at a similar time of day. Participants were scanned wearing t-shirt and shorts, after removing footwear and jewellery. A whole-body scan provided a three-component model of body composition: fat mass, fat-free soft tissue, and bone mineral content. From these, bone mineral density, fat free mass and percentage body fat were calculated for the determination of body composition on a whole-body level with specific body regions also calculated. For fat-free mass and fat-mass these were arms, legs, truck, android and gynoid. For bone mineral density (BMD) and content these were head, arms, legs, trunk, ribs, pelvis, and spine. Participants were instructed to lie in the supine position on the scanner, face-up and to refrain from moving and talking until the scan was complete. The DXA scanner was calibrated following manufacturer instructions each day prior to use.

Participants undertook the standardised Army performance tests in weeks 1 and 12 under the supervision of Army physical training instructors. All performance tests were completed with participants wearing Army standardised physical training kit (t-shirt, short, and trainers). Participants were highly motivated because their best effort was required for progression of their military careers (38). Prior to completing the tests, the participants completed a standardised warm-up which was 10 min in duration and followed a Raise, Activate, Mobilise and Potentiate (RAMP) protocol. The parameters measured included aerobic endurance, upper-body muscular endurance, upper-body explosive power, lower-body strength and lower-body explosive power.

To assess aerobic endurance, the time to complete a 2 km best effort run was measured to the nearest second. To assess upper body explosive power, participants completed a seated (with torso at 90°) 4 kg medicine ball throw with the distance thrown recorded in metres to the nearest centimetre using a tape measure. To assess upper-body muscular endurance, participants completed a 2-min maximal press-up test under the supervision of military staff and the research team. Lower-body strength was measured using a mid-thigh pull rig (Absolute performance, Cardiff, United Kingdom) with participants asked to stand in-front of the bar and pull the bar upwards for five seconds. The participant’s highest score from two attempts was taken and measured in kilograms. Explosive lower body power was assessed by a maximal countermovement jump using a jump mat (Takei Scientific Instruments, Tokyo, Japan). Participants were asked to jump as high as possible three times with their hands placed on their hips to prevent upper limb assistance with the highest score recorded (16). Each jump was completed with 30 seconds rest in between jumps.

Statistical analyses were carried out using the Statistical Package for Social Sciences (v26, IBM, Armock, New York, United States) with significance set at p ≤ 0.05. Data were visually assessed for normality and then quantified using a Shapiro–wilk test. Mixed-model analysis of covariance (ANCOVA) was employed to examine changes in performance outcomes and measures of body composition at week 12 with week 1 data used as the covariate to improve the sensitivity of detecting between group effects, which was the primary research interest (18). A series of ANCOVAs were also used to determine changes in dietary intake, energy expenditure, nitrogen balance and recovery (salivary cortisol and testosterone, subjective measures, and mood). In all instances the week 1 measures were used as covariates. All post hoc analyses were undertaken using an adjusted Bonferroni post hoc test. Partial eta squared effect sizes were classified as small (ƞ² = 0.01), medium (ƞ² = 0.0.06) and large (ƞ² = 0.14) (16, 39). Data are presented as mean ± SD.

Nutritional supplementation (in any form) increased absolute energy intake in PLA, MOD and HIGH compared to CON during week 12 of BT (F[3,93] = 3.940, p = 0.011, ƞ_p² = 0.113) but not during week 6 (F[1,93] = 18.260, p = 0.132, ƞ_p² = 0.058; Table 1). Specifically, post hoc analysis revealed nutritional supplementation increased energy intake by circa 500 kcal in PLA (p = 0.038), MOD (p = 0.022) and HIGH (p = 0.015), in comparison to CON. When calculated relative to body mass, supplementation increased energy intake in the supplement groups week 12 (F[3,91] = 3.078, p = 0.031, ƞ_p² = 0.092) but not week 6 (F[3,91] = 1.579, p = 0.200, ƞ_p² = 0.049; Table 2). Energy intake was greater but did not reach significance in PLA (p = 0.056), MOD (p = 0.055) and HIGH (p = 0.060) compared to CON at week 12 (Table 2).

For CHO intake, there was a significant group interaction at week 6 (F[3,93] = 5.795, p = 0.001, ƞ_p² = 0.157) and 12 (F[3,93] = 8.263, p < 0.001, ƞ_p² = 0.210; Table 1). Supplementation resulted in greater CHO intake in PLA compared to CON (p = 0.017) and HIGH (p = 0.002) at week 6 and in PLA compared to CON (p < 0.001) and HIGH (p = 0.006) at week 12. When intakes were calculated relative to body mass, there was also a significant group interaction at week 6 (F[3,91] = 4.439, p = 0.006, ƞ_p² = 0.128) and 12 (F[3,91] = 8.278, p < 0.001, ƞ_p² = 0.214; Table 2). CHO intake was greater at week 6 in PLA compared to CON (p = 0.024) and HIGH (p = 0.018). CHO intake was greater at week 12 in PLA compared to CON (p < 0.001) and in MOD compared to CON (p = 0.006) and HIGH (p = 0.047).

Supplementation resulted in a significant group interaction for self-reported protein intake at weeks 6 (F[3,93] = 31.416, p < 0.001, ƞ_p² = 0.503) and 12 (F[1,93] = 55.910, p < 0.001, ƞ_p² = 0.643; Table 1). Specifically, protein supplementation resulted in greater protein intake in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001). Protein supplementation also resulted in greater protein intake in MOD compared to CON (p = 0.010) and PLA (p = 0.005) at week 6. By week 12, protein supplementation resulted in greater total daily protein intake in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001). Protein intake was also greater at week 12 in MOD compared to CON (p < 0.001) and PLA (p < 0.001; Table 1).

When protein intakes were expressed relative to body mass, there was an interaction between groups at weeks 6 (F[3,91] = 20.98, p < 0.001, ƞ_p² = 0.404) and 12 (F[3,91] = 31.867, p < 0.001, ƞ_p² = 0.512; Table 2). Supplementation resulted in greater protein intake in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001). Protein intake was also greater in MOD compared to PLA (p = 0.042) and CON (p < 0.001). At week 12, protein intake was also greater in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001) and in MOD compared to CON (p < 0.001) and PLA (p = 0.002; Table 2).

There were no group interactions for absolute fat intake at week 6 (F[3,93] = 1.028, p = 0.384, ƞ_p² = 0.032) or 12 (F[1,93] = 1.002, p = 0.395, ƞ_p² = 0.031) indicating relative consistency between groups (Table 1). Similarly, there were no group interactions for fat intake when intakes were expressed relative to body mass at weeks 6 (F[3,91] = 0.952, p = 0.419, ƞ_p² = 0.030) or 12 (F[3,91] = 0.940, p = 0.425, ƞ_p² = 0.030; Table 2).

There was no group interactions for energy expenditure at week 6 (F[3,65] = 0.829, p = 0.483, ƞ_p² = 0.037) or week 12 (F[3,65] = 0.912, p = 0.440, ƞ_p² = 0.040; Table 3).

There was a group interaction for nitrogen excretion at week 6 (F[3,40] = 7.875, p < 0.001, ƞ_p² = 0.355) but not week 12 (F[1,40] = 3.102, p = 0.085, ƞ_p² = 0.141). Post hoc analysis revealed greater nitrogen excretion in HIGH compared to CON (16.9 ± 4.9 vs. 11.4 ± 3.5 g⸱day⁻¹, p = 0.011), PLA (10.6 ± 2.6 g⸱day⁻¹, p < 0.001) and MOD (13.2 ± 4.0 g⸱day⁻¹, p = 0.022) at week 6.

Following supplementation there was a significant group interaction for nitrogen intake at week 6 (F[1,43] = 34.434, p < 0.001, ƞ_p² = 0.706) and week 12 (F[1,43] = 62.197, p < 0.001, ƞ_p² = 0.813). Post hoc analysis revealed that nitrogen intake was greater in HIGH compared to CON (24.9 ± 4.1 vs. 15.0 ± 2.2 g⸱day⁻¹, p < 0.001), PLA (14.5 ± 2.2 g⸱day⁻¹, p < 0.001) and MOD (18.7 ± 2.2 g⸱day⁻¹, p < 0.001) and greater in MOD compared to PLA (p = 0.007) at week 6. Nitrogen intake was also greater in HIGH compared to CON (25.2 ± 2.9 vs. 14.1 ± 1.8 g⸱day⁻¹, p < 0.001), PLA (14.2 ± 1.8 g⸱day⁻¹, p < 0.001) and MOD (16.6 ± 2.9 g⸱day⁻¹, p < 0.001) at week 12. The greater nitrogen intake in the protein supplementation groups resulted in a group interaction for nitrogen balance at weeks 6 (F[3,43] = 2.200, p = 0.102, ƞ_p² = 0.133) and 12 (F[3,43] = 17.650, p < 0.001, ƞ_p² = 0.552; Figure 1). Nitrogen balance was greater in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001) at week 12 (Figure 1).

When expressed relative to body mass, there was no group interaction for nitrogen balance at week 6 (F[3,43] = 2.187, p = 0.103, ƞ_p² = 0.103) but there was at week 12 (F[3,43] = 15.738, p < 0.001, ƞ_p² = 0.523). Nitrogen balance was greater in HIGH compared to CON (p < 0.001), PLA (p < 0.001) and MOD (p < 0.001) at week 12 (Figure 2).

There was no effect of supplementation for mid-thigh pull (F[3,107] = 0.701, p = 0.554, ƞ_p² = 0.021), medicine ball throw (F[3,107] = 1.143, p = 0.335, ƞ_p² = 0.033), run time (F[3,95] = 0.214, p = 0.886, ƞ_p² = 0.007), press-up (F[3,92] = 1.180, p = 0.322, ƞ_p² = 0.040) or jump performance (F[3,90] = 1.165, p = 0.328, ƞ_p² = 0.040; Figures 3, 4; Table 4).

There was no impact of supplementation on body mass (F[3,121] = 0.400, p = 0.753, ƞ² = 0.010). The descriptive data for each group in week 1 and 12 were as follows: CON = 76.5 ± 11.8 vs. 76.2 ± 9.3 kg; HIGH = 75.4 ± 12.4 vs. 75.6 ± 11.1 kg; MOD = 74.8 ± 12.5 vs. 74.9 ± 10.5 kg; and PLA = 74.6 ± 11.9 vs. 74.7 ± 10.4 kg.

The salivary cortisol and testosterone concentrations for each group are shown in Figures 5, 6. There was no effect of supplementation on cortisol at week 6 (F[3,116] = 0.021, p = 0.996, ƞ_p² = 0.001) or 12 (F[3,116] = 1.514, p = 0.674, ƞ_p² = 0.013). There was also no effect of supplementation on testosterone at week 6 [F(3,113] = 0.812, p = 0.490, ƞ_p² = 0.021) and 12 (F[3,113] = 2.329, p = 0.078, ƞ_p² = 0.058) or on the T:C ratio in week 6 (F[3,110] = 0.261, p = 0.853, ƞ_p² = 0.007) or week 12 (F[3,110] = 0.438, p = 0.726, ƞ_p² = 0.012).

There was no effect of supplementation for muscle soreness at week 6 (F[3,43] = 0.520, p = 0.670, ƞ_p² = 0.035) and week 12 (F[3,43] = 0.313, p = 0.816, ƞ_p² = 0.021) or RPE at week 6 (F[3,36] = 2.450, p = 0.079, ƞ_p² = 0.170) and week 12 (F[3,36] = 1.602, p = 0.091, ƞ_p² = 0.065; Table 7).

Supplementation improved vigour at week 6 (F[3,39] = 3.787, p = 0.018, ƞ_p² = 0.226; Table 8). Vigour was greater and trending towards significance in MOD (p = 0.081) and HIGH (p = 0.076) compared to CON. There was no effect of supplementation at week 6 on anger (F[3,39] = 0.095, p = 0.962, ƞ_p² = 0.007), depression (F[3,39] = 1.540, p = 0.219, ƞ_p² = 0.106), fatigue (F[3,39] = 1.339, p = 0.276, ƞ_p² = 0.093), confusion (F[3,39] = 0.599, p = 0.619, ƞ_p² = 0.044), or global mood score (F[3,39] = 0.612, p = 0.611, ƞ_p² = 0.045; Table 8).

Supplementation improved tension at week 12 (F[3,39] = 3.284, p = 0.031, ƞ_p² = 0.202; Table 8). Tension was greater in CON compared to HIGH (4 ± 3 vs. 2 ± 3, p = 0.022). There was no effect of supplementation at week 12 on vigour (F[3,39] = 0.922, p = 0.439, ƞ_p² = 0.066), anger (F[3,39] = 0.958, p = 0.422, ƞ_p² = 0.069), depression (F[1,39] = 0.260, p = 0.854, ƞ_p² = 0.020), fatigue (F[3,39] = 1.518, p = 0.225, ƞ_p² = 0.105), confusion (F[3,39] = 0.877, p = 461, ƞ_p² = 0.063) or global mood score (F[3,39] = 2.520, p = 0.072, ƞ_p² = 0.162; Table 8).

There was no effect of supplementation at week 6 on general fatigue (F[3,13] = 1.349, p = 0.301, ƞ_p² = 0.237), physical fatigue (F[3,13] = 2.155, p = 0.142, ƞ_p² = 0.332), emotional fatigue (F[3,13] = 2.809, p = 0.081, ƞ_p² = 0.393), mental fatigue (F[3,13] = 1.230, p = 0.338, ƞ_p² = 0.221), vigour (F[3,13] = 2.052, p = 0.156, ƞ_p² = 0.321), or total fatigue (F[3,13] = 0.794, p = 0.519, ƞ_p² = 0.155; Table 9).

There was no effect of supplementation at week 12 on general fatigue (F[3,13] = 0.406, p = 0.751, ƞ_p² = 0.086), physical fatigue (F[3,13] = 0.673, p = 0.584, ƞ_p² = 0.134), emotional fatigue (F[3,13] = 0.861, p = 0.486, ƞ_p² = 0.166), mental fatigue (F[3,13] = 0.343, p = 0.795, ƞ_p² = 0.073), vigour (F[3,13] = 0.684, p = 0.578, ƞ_p² = 0.136) or total fatigue (F[3,13] = 0.369, p = 0.776, ƞ_p² = 0.079; Table 9).