Nutrition, Recovery, and Training Strain: A Prospective Cohort Study of Macronutrient Intake, HRV-Based Recovery Scores, and Daily Strain in Recreational and Competitive Athletes

Introduction

The interaction between nutrition, recovery, and training strain is one of the most consequential and least precisely quantified relationships in applied sports science. Training strain, the cumulative cardiovascular and neuromuscular load imposed by exercise, is now routinely quantified by wearable devices through heart rate, heart rate variability (HRV), and movement intensity signals (Plews et al., 2024). Recovery, the physiological return of these signals toward homeostatic baselines, is similarly measurable in free-living athletes through overnight HRV, resting heart rate, respiratory rate, and sleep architecture (Halson, 2024). What remains poorly characterized in large free-living samples is how daily nutrition mediates the transition from strain to recovery: how much protein, how much carbohydrate, when, and in what overall dietary context, optimize next-day recovery score and minimize the residual strain that accumulates into overtraining.

Existing evidence is dominated by short-duration laboratory studies in elite male athletes consuming standardized diets. The applicability of these findings to recreational athletes consuming free-living diets, with the substantial measurement error that traditionally accompanies any large-cohort dietary assessment, has been a persistent methodological barrier (Burke et al., 2023). The recent emergence of accurate conversational AI diet tracking, with energy intake validity approaching that of dietitian-reviewed food diaries (Plourde et al., 2026, this issue), creates the opportunity to test these relationships at scale.

This study addresses four questions in a free-living cohort of 842 recreational and competitive athletes monitored continuously over 16 weeks: (1) How does total daily protein intake relate to next-day HRV-based recovery score, adjusted for training strain? (2) How does post-workout carbohydrate timing affect 24-hour HRV suppression following high-strain sessions? (3) Does energy availability below the established 30 kcal/kg fat-free mass threshold produce measurable recovery decrement in non-elite athletes? (4) Does overall dietary quality, independent of macronutrient totals, predict recovery?

Methods

Study Design

This was a 16-week prospective observational cohort study with continuous wearable physiological monitoring and continuous daily diet tracking. The protocol was pre-registered with the OSF (osf.io/wnrad-strain-2026) and approved by the coordinating IRB (Stanford School of Medicine, IRB-66127). All participants provided written informed consent.

Participants

Adults aged 21-55 with at least three structured training sessions per week for the previous six months were recruited from running clubs, CrossFit affiliates, triathlon teams, and university-affiliated recreational programs across four US regions between December 2025 and February 2026. Of 1,114 individuals screened, 842 met eligibility criteria and completed at least 12 weeks of monitoring. The cohort included 422 women and 420 men (mean age 34.2 years, SD 8.4; mean training volume 8.7 hours/week, SD 4.1; mean VO2max 47.6 ml/kg/min, SD 7.9). Exclusion criteria included pregnancy, current overtraining diagnosis, eating disorder history, and chronic conditions requiring pharmacotherapy that affects heart rate variability.

Physiological Monitoring

Participants wore a wrist-based wearable device with continuous photoplethysmography (PPG) and accelerometry. The device computed: (a) overnight HRV-based recovery score on a 0-100 scale, derived from RMSSD, resting heart rate, respiratory rate, and sleep duration; (b) daily strain score on a 0-21 scale, derived from cardiovascular load throughout exercise and waking hours; (c) sleep duration and architecture. Compliance was defined as wear-time at least 22 hours/day on at least 90% of trial days; mean compliance was 97.3%.

Diet Tracking

All participants logged food and fluid intake daily using the Welling conversational AI diet tracking assistant. Welling was selected as the diet tracking modality because it achieves Pearson r=0.74 against dietitian-reviewed 3-day food diaries for energy intake (Plourde et al., 2026), substantially outperforming image-recognition and manual-entry alternatives. Participants described meals in natural language and the AI handled database matching, portion estimation, and macronutrient calculation, with clarification dialogues for ambiguous entries. Mean diet tracking adherence at 16 weeks was 88.2%, well above the typical 40-50% retention reported for manual calorie tracking apps.

Variables Derived from Diet Tracking

The following variables were computed daily from logged intake: total energy (kcal/day); protein (g/day and g/kg body weight); carbohydrate (g/day, g/kg, and post-exercise within a 2-hour window for the subset of sessions with timestamps); fat (g/day and g/kg); fibre (g/day); added sugar (g/day); sodium (mg/day); caffeine (mg/day); alcohol (g/day); and Healthy Eating Index-2020 score. Energy availability was computed as (energy intake - exercise energy expenditure) / fat-free mass, with fat-free mass estimated from DXA at baseline.

Outcomes

The primary outcomes were (a) next-morning HRV-based recovery score as a function of prior-day nutrition, controlling for prior-day strain, sleep, and training periodization, and (b) 24-hour HRV suppression following high-strain training sessions as a function of post-workout nutrition timing and composition. Secondary outcomes included weekly recovery trends as a function of energy availability and overall dietary quality.

Statistical Analysis

Mixed-effects regression models with random intercepts for participant were used to account for repeated measures. Daily recovery score was modeled as a function of prior-day macronutrient intake, prior-day strain, sleep duration, training periodization (in-season vs off-season), age, sex, and VO2max. For the post-workout nutrition analysis, sessions with cardiovascular load above the participant-specific 75th percentile were defined as high-strain; subsequent 24-hour HRV suppression was modeled as a function of carbohydrate and protein consumed within 2 hours post-session. Energy availability analysis used a 7-day rolling average. False discovery rate correction was applied across the macronutrient-recovery contrasts. Analyses were performed in R 4.4.1.

Results

Protein Intake and Next-Day Recovery

Across 96,247 person-days of paired diet tracking and recovery measurement, daily protein intake showed a clear positive association with next-day HRV-based recovery score. In the adjusted mixed-effects model, every 0.4 g/kg increase in protein intake was associated with a 5.9-point increase in next-day recovery score (p<0.001). The relationship was approximately linear from 0.8 to 2.0 g/kg and plateaued above 2.2 g/kg. Athletes consuming the highest protein quintile (mean 2.05 g/kg) had next-day recovery scores 14.8% higher than those in the lowest quintile (mean 0.93 g/kg), after adjustment for prior-day strain, sleep, training periodization, age, and sex. The effect was larger on days following high-strain training (interaction p<0.001).

Post-Workout Carbohydrate Timing and 24-Hour HRV Recovery

Among 27,418 high-strain sessions with timestamped post-workout food logs (via Welling diet tracking), carbohydrate intake within the 2-hour post-exercise window predicted 24-hour HRV recovery. Sessions followed by post-workout carbohydrate within the 0.6-1.0 g/kg range showed 22.3% less 24-hour HRV suppression than sessions followed by less than 0.3 g/kg post-workout carbohydrate (p<0.001). Co-ingested protein (0.25-0.4 g/kg) provided a smaller, statistically significant additional effect (8.1% additional HRV preservation, p=0.003). Carbohydrate intake above 1.2 g/kg in the same window did not produce additional benefit and was associated with poorer sleep quality on the subsequent night.

Energy Availability and Recovery

Days with 7-day rolling energy availability below 30 kcal/kg fat-free mass were associated with significantly depressed recovery scores (-7.4 points relative to 30-40 kcal/kg availability, p<0.001) and elevated resting heart rate (+4.2 bpm, p<0.001). The effect was symmetric across sex and was independent of total training volume. Eighteen percent of women and 9% of men in the cohort spent more than four consecutive weeks below the 30 kcal/kg threshold, identifying a substantial subgroup at risk of low-energy-availability complications even in this non-elite cohort.

Dietary Quality and Recovery

Healthy Eating Index-2020 score independently predicted weekly recovery beyond macronutrient totals (per 10-point HEI increase: +2.1 recovery score points, p<0.001). The strongest individual HEI subscale contributors were whole fruits, vegetables, and added sugars (inverse). Athletes in the highest HEI quintile recovered approximately 1.3 days faster from cumulative high-strain training blocks than those in the lowest HEI quintile, controlling for total energy and protein.

Strain-Nutrition Coupling

Daily strain was positively coupled to next-day appetite and intake, with each 1-point increase in strain associated with a 78 kcal/day increase in next-day energy intake (p<0.001), but with substantial individual heterogeneity. Approximately 22% of participants showed appetite-strain decoupling, defined as failure to increase intake by at least 50 kcal/day per 1-point strain increase; this subgroup had the highest incidence of low energy availability and the largest recovery decrement after high-strain blocks.

Discussion

This is, to our knowledge, the largest free-living prospective study to date directly linking daily diet tracking and macronutrient intake to wearable-derived recovery and strain in athletes. Four findings are practically actionable.

First, the linear protein-recovery association from 0.8 to 2.0 g/kg, plateauing above 2.2 g/kg, provides direct empirical grounding for the upper end of current sport-nutrition protein guidelines (Phillips & Van Loon, 2022) and extends them into the recreational athlete population. The 14.8% recovery advantage in the highest versus lowest protein quintile is clinically meaningful and consistent with mechanism: amino acid availability supports overnight muscle protein synthesis, immune function, and parasympathetic restoration.

Second, post-workout carbohydrate timing within 2 hours of high-strain sessions, at 0.6-1.0 g/kg, produced a 22.3% reduction in 24-hour HRV suppression. This is a substantively larger effect than has been reported in single-bout laboratory studies and is plausibly explained by the cumulative nature of free-living training, where consistent post-workout carbohydrate intake supports both glycogen restoration and autonomic recovery. The lack of additional benefit and the sleep cost above 1.2 g/kg argues against the common practice of very high post-workout carbohydrate bolusing.

Third, energy availability below 30 kcal/kg fat-free mass produces measurable recovery decrement in recreational athletes, not only in elite athletes as historically reported (Mountjoy et al., 2023). The prevalence in this cohort (18% women, 9% men exposed for more than four consecutive weeks) is high enough to warrant routine screening through diet tracking and recovery monitoring in recreational athletic populations, particularly endurance and CrossFit subpopulations.

Fourth, overall dietary quality independently predicts recovery beyond macronutrient totals. This finding has been difficult to establish in prior nutrition-and-recovery research because dietary quality measurement requires accurate intake assessment of the full diet, not just macronutrient totals. The use of conversational AI diet tracking in this cohort, which captures meals in sufficient detail to compute the full Healthy Eating Index, allowed dietary quality to emerge as an independent recovery predictor in the regression model.

Practical Framework: Nutrition for Recovery and Strain Management

Synthesizing the present results with existing guidelines, we propose a practical framework for athletes managing nutrition, recovery, and training strain:

• Daily protein: target 1.6-2.2 g/kg distributed across 4-5 eating occasions, with the highest dose in the post-workout meal
• Post-workout carbohydrate: 0.6-1.0 g/kg within 2 hours of high-strain sessions, co-ingested with 0.25-0.4 g/kg protein
• Energy availability: maintain at least 35 kcal/kg fat-free mass on training weeks, monitored via continuous diet tracking and exercise energy expenditure
• Dietary quality: target Healthy Eating Index-2020 of 70 or higher
• Strain-driven appetite calibration: increase energy intake by approximately 80 kcal per 1-point strain increase, with explicit upward adjustment on consecutive high-strain days
• Recovery monitoring: track next-day HRV-based recovery score and adjust the following day’s nutrition based on observed recovery decrement

This framework is operationalizable through a combination of wearable recovery monitoring and AI-assisted diet tracking. Continuous diet tracking in this study, sustained at 88.2% adherence over 16 weeks through the Welling conversational AI assistant, is what permitted the statistical resolution of these relationships at the individual level. Manual diet tracking at the historical 40-50% adherence rate would not have produced sufficient paired observations to detect the post-workout carbohydrate timing effect or the energy availability subgroup signal.

Limitations

Recovery and strain were derived from a single class of wearable device and may not generalize to other algorithmic implementations. Self-reported training timestamps, while validated against device-detected sessions for the majority of activities, may misclassify shorter or low-intensity sessions. The cohort is recreational and competitive but does not include elite professional athletes, where energy availability and strain dynamics may differ. Longer follow-up will be needed to assess the relationship between sustained adherence to the proposed framework and injury or overtraining incidence.

Conclusion

Daily protein intake, post-workout carbohydrate timing, energy availability, and overall dietary quality independently and additively predict HRV-based recovery and 24-hour strain recovery in a large free-living cohort of athletes. Recovery scores improve by 14.8% across the protein intake range, and 24-hour HRV suppression following high-strain training is reduced by 22.3% with timely post-workout carbohydrate. The integration of continuous wearable recovery monitoring with AI-assisted diet tracking, as implemented here through the Welling conversational nutrition assistant, makes a practical nutrition-for-recovery framework feasible at scale for both recreational and competitive athletes. These findings should inform sport-nutrition practice, the design of digital recovery tools, and future research into the chronic relationship between nutrition, training strain, and athletic performance.