Cycling for a marathon: transfer, substitutions and limits

On this page

1. Aerobic transfer and specificity
2. What the 2026 synthesis adds
3. What this cannot establish for a marathon
4. Replacing is not adding
5. A practical framework, not a formula
6. Sources

Aerobic transfer and specificity

Tanaka’s 1994 review describes transfer of some VO₂max effects between modes, with specificity more important in highly trained athletes. It remains a historical source on oxygen uptake, not a rule for how much cycling replaces a marathon long run.¹

Millet and colleagues’ 2009 review discusses running-cycling differences, including in triathletes: muscle recruitment, neuromuscular characteristics and physiological responses do not fully match. Shared aerobic work does not automatically reproduce the running movement, foot strikes or mechanical demands.²

What the 2026 synthesis adds

Menges and colleagues included seven randomized studies with interventions of at least four weeks and mixed populations, from untrained people to competitive runners. Comparisons involved running only, cycling only or combined training. Outcomes were treadmill- or cycle-assessed VO₂max and running tests over one mile, 3000 and 5000 metres, not marathons.³

“No difference detected” does not mean “equivalent modes”. Confidence intervals are wide and cross zero. Authors note few studies, older protocols, heterogeneity and risk-of-bias concerns. The characteristics section initially describes studies from 1974–2003, but the review also includes a 2018 trial: the important issue is the limited, varied evidence, not assigning all studies to one era.³

What this cannot establish for a marathon

• One-mile to 5-km outcomes do not establish preparation for 42.195 km.
• VO₂max alone does not measure marathon economy, mechanical tolerance, fuelling or fatigue resistance.
• These sources provide no optimal substitution share for every marathon runner.
• They do not prove cycling prevents injuries or suits every painful condition: position and load change, but risks do not disappear.
• These sources do not establish a rule allowing “recovery cycling only in the final 8–10 weeks”. Its role depends on the program and person.

Replacing is not adding

An illustrative load example

Adding 60 cycling minutes to a 240-minute running week gives 300 total minutes: time rises 25%. Replacing 40 running minutes with 60 cycling minutes instead leaves 200 running minutes and 60 cycling minutes, 260 in total. Time rises 20 minutes, but mechanical and physiological load cannot be calculated from this sum alone. Neither operation converts cycling into running kilometres.

A long or hard ride may compromise recovery despite lacking the same running impacts. Calling a ride “recovery” does not make it so. Actual intensity, residual fatigue and quality of subsequent sessions still need assessment.

A practical framework, not a formula

• State the purpose: variety, alternative aerobic work or reduced impacts. Do not automatically promise performance maintenance.
• Record time and intensity separately by mode; adding durations does not make stimuli equivalent.
• Consider race-specific sessions before choosing what to replace, without a universal percentage.
• If running is reduced because of pain or injury, agree tolerated activity and progression with whoever assesses the problem.
• When returning to running, perceived cardiovascular fitness does not by itself justify immediately resuming old volume.

Sources

1. Tanaka H. Effects of cross-training: transfer of training effects on VO2max between cycling, running and swimming,1994. PubMed, abstract.
2. Millet GP,Vleck VE,Bentley DJ. Physiological differences between cycling and running: lessons from triathletes,2009. PubMed, abstract.
3. Menges T et al. Cross-training between running and cycling: effects on VO2max and running performance - a systematic review and meta-analysis,2026. Frontiers.

Sources consulted on 29 October 2026. Illustrative time examples, not prescriptions.

All articles