Do marathoners who do more quality sessions run faster?
Among 917 Boston 2022 runners, each extra weekly quality session 12-4 months out went with a ~16-min faster race, adjusted for volume; recalled training, association only. B.
Main caveat: Recall and response. Weekly averages over 8- and 4-month windows recalled a month before the race; training logs were not required.
DeJong Lempke et al. surveyed 2022 Boston Marathon registrants in the month before the race about their training over the past year (12-4 months out) and the last 4 months (4-0 months): weekly running hours, distance, number of runs, number of quality ("hard") sessions, cross-training hours and sessions.
Survey answers were linked to official chip times and World Athletics points. Of 1063 enrolled, 917 with complete data were analysed (495 women; 4% of the field). Runners reported 4.95 runs and 1.65 quality sessions a week in the 12-4 month window and 5.26 and 1.83 in the last 4 months, and 66.1 km a week.
In regressions with age, sex, years of marathon training and previous marathons as covariates and all training variables entered together, more quality sessions went with faster times: about 16.2 min per extra weekly session in the 12-4 month window and about 17 min in the last 4 months.
More runs a week (about 3.6 and 3 min per run), more distance and running more than 10 h a week also went with faster times, and more cross-training sessions in the last 4 months with about 6 min per session. The change in quality sessions between windows was not related to performance (p = 0.720).
The authors conclude that habitually high training exposure with a relative reduction in the last months goes with better performance.
Our reading: the only large data set in the base that counts quality sessions in amateurs; faster Boston runners did more of them, but this is cross-sectional recall in which ability, training history and session count travel together, and it measured session counts where REF-197 measured minutes in zone.
Protocol
Design: prospective observational study: cross-sectional pre-race survey (REDCap) in the month before the 2022 Boston Marathon, recruitment through the organiser's weekly emails, chip time obtained after the race. IRB Harvard Medical School / Boston Children's Hospital (IRB-P00041933).
Participants: 1063 enrolled; n = 917 with complete survey and race data (495 women, 422 men); age 43 ± 13 (women) and 52 ± 13 y (men); 9 ± 8 years of dedicated marathon training; 16 ± 26 previous marathons. Non-finishers and incomplete responses excluded.
Exposures (recalled, per week, for 12-4 and 4-0 months before the race): running hours (bins), running distance, total runs, quality sessions ("hard sessions (e.g., interval sessions, tempo runs, 'fartlek' runs, or similar workouts by cross-training)"), cross-training hours (bins) and sessions. Training-frequency change = 4-0 month average minus 12-4 month average.
Outcomes: official race time (hours) and World Athletics points (from time, sex and age); McKay tiers assigned from WA points.
Analysis: three multivariable linear regressions (12-4 months, 4-0 months, change between windows) with all training variables and a runs × cross-training interaction; covariates age, sex, years of marathon training and number of previous marathons; α 0.05; R and Jamovi.
Funding and interests: "Joe and Clara Tsai Foundation as part of the Wu Tsai Human Performance Alliance.Eleanor and Miles Shore Faculty Development Awards Program, Harvard Medical School: Boston Children's Hospital Department of Orthopaedic Surgery and Sports Medicine Award." Conflicts: DeJong Lempke consults for research projects with Boston Children's Hospital and receives research support from the BIRCWH programme of the NIH's National Institute of Arthritis and Musculoskeletal and Skin Diseases (5K12AR084233); others none.The Boston Athletic Association and the Boston Marathon Medical Program are thanked for their support.
Key Results
| Predictor (per week), adjusted | 12-4 months before: β, h (95% CI) | ≈ minutes | 4-0 months before: β, h (95% CI) | ≈ minutes |
|---|---|---|---|---|
| Quality sessions (+1) | −0.27 (−0.42 to −0.10), p < 0.001 | 16.2 faster | −0.29 (−0.43 to −0.16), p < 0.001 | ~17 faster |
| Runs (+1) | −0.06 (−0.13 to −0.02), p = 0.010 | 3.6 faster | −0.05 (−0.10 to −0.01), p = 0.023 | 3 faster |
| Distance (+1 km) | −0.01 (−0.02 to −0.005), p < 0.001 | 0.6 faster | printed −0.22 (−0.48, −0.34), p < 0.001 | printed "~13" |
| Cross-training sessions (+1) | −0.01 (−0.10 to 0.01), p = 0.329 | — | −0.10 (−0.16 to −0.04), p < 0.001 | ~6 faster |
| Model R² (race time) | 0.485 | 0.490 |
Training reported (mean ± SD): runs 4.95 ± 1.42 (12-4 months) and 5.26 ± 1.42 (4-0 months); quality sessions 1.65 ± 1.07 and 1.83 ± 1.00 (women 1.63 → 1.84, men 1.68 → 1.83); distance 66.1 ± 25.2 km/week (12-4 months); cross-training sessions 2.68 ± 2.05 and 2.70 ± 1.99.
In World Athletics points: +73.6 (24.6 to 122.6) per extra quality session (12-4 months) and +76.6 (34.5 to 115.0) in the last 4 months.
Change between windows: Δ quality sessions β −0.01 (−0.03 to 0.04), p = 0.720 (no evidence of a relation); Δ runs β −0.05 (−0.09 to −0.003), p = 0.035, which the authors read as "decreased running sessions → about 3 min faster".
Running hours: > 10 h/week went with faster times than 5-7.5 h (β −0.19, −0.30 to −0.08) and 2.5-5 h (−0.22, −0.37 to −0.08) in the 12-4 month window.
Covariates: age and sex significant; years of marathon training and previous marathons not.
Sample: 95% Tier 2 (3:32 ± 0:32 three-year average personal best), 5% Tier 3 (3:00 ± 0:14).
Mechanisms
Measured: none; the study links recalled training to race time.
Authors' interpretation: quality sessions stand for intensity and training variety, and interval-based training induces adaptations (aerobic capacity, muscular power) that support running performance; volume builds economy; a relative reduction in the last months reflects a taper.
Inference: faster runners may tolerate and choose more hard sessions because they are fitter, and a runner's count of "hard sessions" grows with experience and goals; the association cannot say that adding a session makes a given runner 16 min faster. The quality-session coefficient is adjusted for distance and runs, which are correlated with it, so its size is unstable.
Practical Application
Do marathoners who do more quality sessions run faster? Among 917 Boston 2022 runners, each extra weekly quality session 12-4 months out went with a ~16-min faster race, adjusted for volume; recalled training, association only. B.
Read it beside [REF-197], not against it: in 119,452 Strava marathoners faster runners had a bigger easy layer and similar minutes above critical speed; here faster runners recalled more hard sessions. Counting sessions and counting minutes are different measures, and both studies are cross-sectional (inference).
Not from this paper; the base's own practice: for a recreational marathoner on 4-6 runs a week, two quality sessions at most, the long run carrying marathon pace late counting as one of them — on 4 runs one quality day + the marathon-pace long run + two easy runs, if recovery holds (SYNTH-64, on SYNTH-26's table); this survey shows that faster runners recalled more quality sessions; it cannot say whether adding one would make a given runner faster.
Limitations
Recall and response. Weekly averages over 8- and 4-month windows recalled a month before the race; training logs were not required; 4% of the field responded (5% of women, 3% of men).
Direction and confounding. Cross-sectional: fitter runners can do more hard sessions; training variables are entered together although correlated.
Crude exposure. "Hard session" was self-defined and could include hard cross-training; no intensity, duration or distribution of the sessions; cross-training not separated into strength and other sports; running and cross-training hours in bins.
Estimates that do not add up. In the 4-0 month model the distance coefficient is printed as −0.22 h per km with a 95% CI of −0.48 to −0.34 (the estimate lies outside its own interval) and translated as "~13 min" per km, against 0.6 min per km in the 12-4 month model; the age covariate is printed 0.02 (−0.02, 0.03) with p < 0.001; in the change model the age and sex coefficients look swapped (age −0.58, sex 0.03).The change model's Δ-runs coefficient is negative (−0.05 h) with positive WA points (+14.6) under the Methods' definition Δ = 4-0 minus 12-4 months, which reads as more runs in the last 4 months going with faster times; the abstract and conclusion say runners who decreased their runs were faster.Mean runs and quality sessions also rose, not fell, from the 12-4 to the 4-0 month window (4.95 → 5.26; 1.65 → 1.83), while the conclusion speaks of a "relative taper".
A 4-month window is not a taper. The authors acknowledge that 4-0 months mixes specific preparation with the taper.
Who. 2022 Boston Marathon registrants (the paper does not split qualifiers from other entrants), 92% white, 92-93% North American; men older than women (52 vs 43 y).
Funding and interests: Joe and Clara Tsai Foundation (Wu Tsai Human Performance Alliance) and a Harvard / Boston Children's Hospital faculty award; the first author's consulting for research projects with Boston Children's Hospital and research support from the BIRCWH programme (NIH NIAMS, 5K12AR084233) are listed as conflicts; recruitment ran through the race organiser's emails, and the Boston Athletic Association is thanked for its support.
How this entry was checked and changed
Why this level
correlational study with n = 917 analysed, read from the full text (open access, CC BY). Large enough for B+ by size; graded B because the exposure is recalled weekly averages over 8- and 4-month windows, only 4% of the field took part, faster runners may simply train harder (reverse direction), and several table entries are internally inconsistent
Full text (checked 2026-10-27)
In total, 917 athletes were included (female: n = 495, 3:53 ± 0:37 h race times; male: n = 422; 3:35 ± 0:39 h race times); 4% of the 2022 Marathon field (full text)
Weekly quality sessions (n): 12–4 months: 1.65 ± 1.07; 4–0 months: 1.83 ± 1.00; weekly running sessions 12–4 months: 4.95 ± 1.42; 4–0 months: 5.26 ± 1.42 (full text)
With each additional quality session/week in the 12–4 months pre-race, participants improved their race time by ~ 16.2 min (β [95% CI] − 0.27 [− 0.42, − 0.10]; p < 0.001) or 73.6 points (full text)
For each additional running session/week in the 12–4 months pre-race, participants improved their race time by ~ 3.6 min (β [95% CI] − 0.06 [− 0.13, − 0.02]; p = 0.010) (full text)
For every 1 km increase in running distance/week 4–0 months pre-race, race time improved by ~ 13 min (β [95% CI] − 0.22 [− 0.48, − 0.34]; p < 0.001) (full text)
Δ Weekly quality sessions − 0.01 (− 0.03, 0.04), p 0.720; Δ Weekly running sessions − 0.05 (− 0.09, − 0.003), p 0.035 (full text)
Funding: Joe and Clara Tsai Foundation as part of the Wu Tsai Human Performance Alliance; Eleanor and Miles Shore Faculty Development Awards Program, Harvard Medical School (full text)