SYNTH-61 · Guide

Trail and mountain races

Quadriceps fatigue resistance sits beside VO2max, so train uphill and downhill as separate skills, precondition the quads for descents weeks out, and pace a mountain ultra by effort and stops, not by heart rate

Blimited evidenceMostly from recreational to highly trained athletesWritten before October 2026, being re-checked

The question it answers

"I have a trail race, short trail to mountain ultra — what must my training contain that road training does not?"

Default (what to prescribe first)

For a healthy Tier 1-3 runner more than ~8 weeks from a trail race:

Keep the engine. VO2max correlated with 27-km trail race time (r = -0.76) in nine highly trained trail runners (REF-780, B, correlational). Inference: aerobic volume and VO2max work stay in the plan (SYNTH-42, C+; SYNTH-33, B).

One uphill quality session a week. In 20 well-trained runners, six weeks of uphill intervals improved 5-km time trial by 2.0% on average with no clear best intensity; the hardest programme was best for economy (2.4%) (REF-781, B; no flat control; transfer to trail is inference). Not from a study; the base's own practice: the weekly slot and shapes such as 6-10 × 1-3 min hard uphill, jog down.

Graded downhill exposure, finished before the taper (house practice for dose and timing; below).

Long runs on the race's terrain, with climbing late in the run (house practice; below).

Heavy strength twice a week (SYNTH-44, B+ for direction; the dose is house practice).

Rehearse race fuelling on long runs (below).

Decision (who → what)

Situation	What changes	Source and how sure
Short trail race (27 km in the one study; ~20-40 km is house extension)	Quadriceps local endurance alongside VO2max (the two significant correlates); uphill economy only inside a 9-runner model the entry calls likely overfit; test on a gradient	REF-780 B, n=9, correlational
Mountain marathon (~4000 m D+)	Descent preconditioning; up to 120 g/h carbohydrate only if the gut is trained	REF-660 B+; REF-331 REF-332 B+, elite
Mountain ultra (~75 km and longer)	Aerobic speed and the share of it held for hours, strong quadriceps; conservative early pace, even effort, fewer stops, night-section safety, ~2 weeks' recovery after	REF-827 B (24 men, 75 km: fraction of MAS held r = −0.89, MAS −0.74, knee-extensor force −0.51; economy not correlated); REF-829 B (161 km winners varied speed least); REF-664 B (fewer stops; its faster runners lost more speed, so 'even' rests on REF-829); REF-665 B; REF-662 B
Runner new to descending or to ultras	Slower build of downhill and vertical; the least-accustomed take the most damage, and less experienced ultrarunners reported more injury	REF-660 B+; REF-779 B+, self-reported (REF-783 found the opposite for experience)
Woman ultrarunner with focal foot bone pain	Suspect stress injury, stop and refer	REF-779 B+, association

What decides trail performance

REF-780 (B, n=9 highly trained trail runners, Tier 3, one 27-km race): a knee-extensor fatigue index (r = 0.91) and VO2max (r = -0.76) correlated with race time; maximal concentric and eccentric knee-extension torque did not.

The classic model (VO2max, %VO2max at VT, level economy) predicted poorly (R = 0.48); fatigue index plus economy on a +10% slope raised it to R = 0.98 — which the entry itself calls likely overfit in nine runners. Uphill costs more work at every joint, energy cost rising linearly with grade; downhill is braking, cheapest near −20% (REF-546, B+, narrative review).

Up and down are different skills: in 10 highly trained men, 5-km uphill and downhill time trials (±8%) both depended on vVO2max and maximal leg-extensor strength; uphill added BMI (r² = 0.94), downhill leg musculotendinous stiffness (r² = 0.84); HR was the same up and down (94-95% of max) while VO2 was lower downhill (REF-828, B, correlational) — HR over-reads a fast descent (the entry's inference).

At ultra length the picture shifts: in 24 men over a 75 km mountain ultra (3 930 m up), the fraction of maximal aerobic speed held (r = −0.89), maximal aerobic speed (r = −0.74) and knee-extensor force (r = −0.51) tracked race time, but the energy cost of level or uphill running did not (REF-827, B; the fraction held is partly the race outcome itself).So "uphill economy" is a short-trail finding (REF-780); for ultras, weight the engine, its sustainable fraction and quadriceps strength.

Uphill work

Evidence: REF-781 above — the hardest programme helped economy and all neuromuscular measures most, and the authors conclude any high-intensity uphill format can be assumed to help the 5-km. The abstract gives no durations, gradients or recoveries, so any session is the base's practice, not the study's.

Walk the steep pitches. At +45% on a treadmill, walking cost 17.33 vs running 18.93 J/kg/m (REF-547, B, n=10); the switch point is individual. Not from a study; the base's own practice: train power-hiking on steep ground.

Poles: in 14 mountain runners, poles cut vertical cost only ~2-3% at 25.4-35.5° but RPE 14.4-18.5% at 15.5-38.9° (REF-661, B, 5-min fresh treadmill walking bouts); no benefit shown near 10°. Not from a study; the base's own practice: practise with poles before racing with them.

Downhill and eccentric preparation — the repeated-bout protection

Why: over long efforts, hilly running fatigues the legs much like flat running of the same duration, but pure downhill running causes severe muscle damage and low-frequency fatigue from mechanical stress, and hard climbing low-frequency fatigue too (REF-830, B+, narrative review) — duration first, descents as a separate damaging load.Downhill running causes muscle damage that can cut performance; the least-accustomed suffer most; prior downhill exposure is the best-supported prevention, compression and footwear unconfirmed (REF-660, B+, narrative review). One eccentric bout protects against damage from later ones (REF-549, B+, mechanism review; no dose in the abstract).

Dose that exists: in 12 untrained young adults, 10 downhill sessions over 4 weeks at a speed equal to 60-65% VO2max raised knee-extensor strength 9.7-15.2%, mostly neurally, with muscle size up by week 2 and more by week 4; VO2max unchanged (REF-782, B, uncontrolled, Tier 0-1; trained runners may adapt less, and protection from race damage was not measured).

Cost: 40 min of downhill running kept leg-muscle stiffness raised up to 96 h in 18 recreational men (REF-317, B+).

Not from a study; the base's own practice: start with short, moderate descents 8-12 weeks out, add gradient and length every 1-2 weeks, reach race-like descending in long runs 3-5 weeks out, and put the last big descent session outside the taper so its stiffness is gone. REF-660 says the optimal dose and timing are undefined.

Vertical per week and technical terrain

No entry sets a weekly vertical target. Not from a study; the base's own practice: build weekly vertical like distance, and by the final build weeks make some long runs match the race's metres of climb per km. Inference from REF-565 (B+, review of 4 studies on distance, not vertical): 10% and 24% weekly increases carried the same risk, >30% vs <10% gave HR 1.59 (CI crossing 1) — avoid big jumps, no fixed percentage.

Technical terrain has no entry. Not from a study; the base's own practice: run the race's surface weekly, easy, and learn descending footwork slowly first.

Long runs on terrain

In 23 trained trail runners (11 women), 180 min with an hourly +12% uphill time trial cost 6.6% of uphill TT distance while energy cost and leg stiffness held and carbohydrate oxidation fell (REF-548, B, treadmill, no descents). Not from a study; the base's own practice: long runs on hilly terrain with the hardest climbing late, and back-to-back long days only for ultra goals. Durability logic is SYNTH-32 (B, mostly cyclists).

Strength

Heavy lower-body strength twice a week for at least 8 weeks, then a lower dose in season (SYNTH-44, B+; the dose is house practice). REF-780's correlate was quadriceps fatigue resistance, not peak torque, and the entry's own house note is that local endurance adds to strength work, it does not replace it. Inference: downhill running may do part of the eccentric job (REF-782 compares it to eccentric gym work but did not test it).

Fuelling — the ultra carbohydrate conflict, stated

The base holds four numbers that do not agree:

30-50 g/h carbohydrate in racing, within 150-400 kcal/h from calorie-dense foods, protein 5-10 g/h (REF-528, A, ISSN stand, single-stage ultra).

~90 g/h multiple-transportable carbohydrate supported for runs ≥3 h, with tolerability in ultra competition "questionable"; LCHF not supported for performance (REF-663, B+, narrative review).

71 ± 20 g/h in three elite male 100-mile runners (REF-068, B, case series).

120 g/h gave lower 24-h muscle-damage markers and internal load than 60 or 90 g/h in gut-trained elite male mountain-marathon runners (REF-331, B+, n=20), and was the only arm without a significant 24-h drop in jump and half-squat 1RM (REF-332, B+, n=26 elite trail runners; "might limit" in the authors' words).

Not from a study; the base's own practice (SYNTH-02, B-): up to ~6 h use its 3-6 h row; beyond 6 h start at 60-80 g/h and let rehearsal set the ceiling; 120 g/h only gut-trained and at mountain-marathon durations, where it was measured.

Fluid: REF-663 says euhydration is generally maintained by drinking to thirst, while REF-528 gives 450-750 mL/h with sodium — a second disagreement no entry settles (SYNTH-68 holds the drinking decision); on hyponatraemia the consensus (REF-776, A-) attributes it to drinking beyond losses and recommends drinking to thirst (Grade 1C), so thirst governs and 450-750 mL/h serves to plan what to carry.

Injury risk

Trail injury was associated with neglecting the warm-up, no specialised plan, training on asphalt, double sessions a day, physically demanding jobs, and — unexpectedly — more experience; foot and toe blisters were the most reported injury (REF-783, A-, living SR of 10 observational risk-factor studies, 2 785 runners; no effect sizes).

In longer ultras ~50-60% of runners have musculoskeletal problems, mostly ankle and knee, and experience is the strongest performance predictor (REF-826, C+, narrative review, approximate figures).

Among 1,212 ultrarunners, 64.6% reported an injury costing training days in the past year, knee most common, stress fractures 5.5% (44.5% foot), more in women; injured runners were younger and less experienced (REF-779, B+, self-reported). The two disagree on experience; neither entry resolves it.

Taper

No trail-specific taper entry. The general rule: volume down 41-60% over ~2 weeks with intensity and frequency kept (SYNTH-34, B+); 10-14 days for 4-6 h events is house practice (SYNTH-34 §5). Not from a study; the base's own practice: no new or large descent load in the last ~10 days (the 96-h stiffness of REF-317 is the reason, not a tested cut-off).

Mountain ultra pacing and after

In 15 runners over a 106 km mountain ultra, speed fell at every gradient until 70-90% of the race, HR fell while RPE rose, and faster finishers stopped less (r = −.772) but lost more speed, not less (r = .540; speed variability not significant) (REF-664, B, observational). Inference, in the entry: a low late HR is not spare capacity.

Across 24 editions of the 161 km Western States, winners varied their speed less than the rest of the top 5 (CV 12% vs 14-15%), and among the 10 fastest times steadier pacing went with faster times (r = .80); winners sat close behind the leaders early, led in the middle half and slowed least late; variation rose with heat (REF-829, B, elite, one course).

Millet's flush model: pace is set by perceived exertion held under a security reserve, shifted by sleep loss, heat, pain-killers, fuelling (REF-777, B+, review; untested). In a 24-h road race the better performers started slower relative to their own 24-h average (r = -.58 for the first 2 h) (REF-778, B+, 398 men and 103 women, observational; road/loop format, transfer to mountain courses is inference).

At UTMB, sleep loss plus effort impaired cognition up to visual hallucinations (REF-665, B, n=17): a stop-and-assess signal.

After a 166 km mountain ultra, knee-extensor and plantar-flexor force fell 35-39% and recovered by 16 days, mostly by 9 (REF-662, B, 22 experienced runners, 11 followed through recovery); in 50 runners of a 330 km mountain ultra, quadriceps shear modulus fell ~7% and was only partly back at 48 h (REF-313, B).

Not from a study; the base's own practice: pace by effort and grade-adjusted pace (REF-547 is the basis of GAP), cap the first hours, plan aid-station time, and no hard or heavy training for ~2 weeks after.

How to progress / periodise

Not from a study; the base's own practice: base = volume, strength, easy hills; build (8-12 weeks out) = uphill session, downhill progression, terrain long runs with rehearsed fuelling; specific (3-6 weeks out) = race-profile long runs, poles; taper. Raise one load (distance, vertical or descending) at a time.

When NOT to use it / red flags

Not a first downhill block inside the last ~10 days before the race (house practice).

Focal bone pain, especially foot pain in a woman (REF-779), stops running — refer, do not taper through it.

Confusion, hallucinations, vomiting or worsening headache in a race: stop and get medical help (house practice; REF-665; the consensus REF-776, A-, attributes hyponatraemia to overdrinking and advises drinking to thirst).

Do not read a falling HR late in a mountain ultra as a green light (REF-664).

No ibuprofen during an ultra: in 89 runners over 80 km, 400 mg every 4 h went with acute kidney injury in 52% vs 34% on placebo (number needed to harm 5.5; harm not statistically proven) (REF-892, A-, desert races).

What the evidence does and does not show

Shows (small studies): uphill and downhill are different tasks with partly different determinants (REF-828); quadriceps endurance and VO2max go with short-trail time, aerobic speed, its sustainable fraction and quadriceps strength with 75 km time (REF-827); winners of a 161 km race paced most evenly (REF-829); uphill intervals were followed by better 5-km and economy; downhill running builds strength in novices and prior exposure is the best-supported protection.

Does not show: downhill preconditioning with a race outcome in trained runners, a vertical dose, a terrain method, a trail taper, a controlled >6 h fuelling rate.

Evidence gaps

Determinants at ultra-trail distance rest on 24 men in one 75 km race (REF-827) and 10 men over 5 km (REF-828); no women, no training test.

Downhill preconditioning in trained runners with a race or damage outcome; downhill running vs eccentric gym work.

Uphill session details (Barnes 2013 full text); weekly vertical dose; technical terrain; trail taper.

Mountain-ultra pacing below the elite front (REF-829 is top-5 finishers of one race).

Trail data in women.

How this entry was checked and changed

Why this level

(determinants and training rest on small correlational or uncontrolled studies, n=9-24; fuelling on an A stand and B+ trials; downhill progression, long-run design, vertical, terrain, taper and pacing caps are house practice)

Content correction 2026-09-27

Revision 2026-09-29: folded in REF-827 (Balducci 2017: 75 km mountain ultra — MAS, its sustainable fraction and quadriceps force track time, economy does not) and REF-828 (Lemire 2021: uphill vs downhill determinants) into "what decides" and the mountain-ultra row — the uphill-economy claim is now limited to short trail races; REF-830 (Giandolini 2016) in downhill preparation; REF-829 (Hoffman 2014: even pacing by 161 km winners) in pacing; REF-826 (Knechtle & Nikolaidis 2018, C+) in injury; REF-892 (Lipman 2017, A-) as a red flag.Correction: the fluid line said the hyponatraemia consensus REF-776 was a record only (C) with no recommendation — live it is A- and recommends drinking to thirst; fixed to match SYNTH-64/65. Two candidate gaps closed. Level unchanged.

Verification 2026-10-01: all 31 cited REF entries opened in the live snapshot; every level/design/n written next to a refId matches its head block, every number checked against the entry. No cited REF is unverified or without DOI/PMID. One cited framework (SYNTH-05) is a superseded pointer.Findings: 3 STRETCH (2 medium), 1 SOURCE, 1 CONTRADICTION, 1 LEVEL, 1 UNLABELLED, 1 NUMBER, 1 LEVEL-OVERALL. Not flagged: REF-661 RPE range '15.5-38.9°' skips 25.4° (as the entry's own lead does).

2026-10-01: level tags next to cited frameworks updated to their current levels after the independent audit.

2026-10-02: aligned with the rewritten SYNTH-42 (missing stimulus, one lever per block, level C+): level written next to SYNTH-42 B- -> C+; cross-reference label "limiter" -> "block lever".