Why does my lower back hurt on the bike?
A back history is the strongest marker: in 1,274 Italian amateurs, back pain before cycling went with 4.2× odds of back pain in the past year; coached riders had lower odds. Association.
Main caveat: Cross-sectional and self-reported. Exposures and back pain were recalled together; causes cannot be inferred, and the authors say so.
Battista et al. (physiotherapy, University of Genoa) surveyed Italian amateur cyclists about low back pain: how common it is, what it is like, what goes with it, how riders manage it and what they believe causes it.
A 56-question web questionnaire built by physiotherapists, not announced as being about back pain, was completed in full by 1,274 riders between 15 and 30 July 2020 (73.3% of the 1,738 reached; 198 excluded for back surgery, vertebral fracture, severe spondylolisthesis or rheumatic disease).
Low back pain had occurred in 55.1% in their lifetime, 26.5% in the last 12 months and 10.8% in the last 4 weeks; the last episode usually lasted some days (80.4%) at a median intensity of 5 of 10.
A backward-stepwise logistic model for back pain in the last 12 months kept sex, work type, years of cycling, being coached or following a scheduled plan, other sports, specific pedalling-technique training and back pain before cycling: back pain before cycling went with higher odds (OR 4.20, 3.21-5.40), coaching or a scheduled plan with lower odds (OR 0.53, 0.37-0.74), and pedalling-technique training (OR 0.69, 0.51-0.94) and 2-5 vs under 2 years of cycling (OR 0.48, 0.26-0.89) with lower odds at p = 0.02 — which the authors call significant, though they had set p < 0.01 by Bonferroni.
A past "biomechanical visit" (a bike fit; 54.2% had had one), weekly sessions, stretching, competitions, age, BMI and sleep dropped out of the model.
Our reading: in amateurs as in rowers (REF-641), the back you bring is the strongest marker of the back pain you get — 74.2% of riders who had had back pain on the bike had had it before they started cycling; structured training went with less back pain, which fits a load explanation but is not proof of one; and a past bike fit did not separate riders with and without back pain.
Protocol
Design: web-based cross-sectional survey developed by physiotherapists following the International Handbook of Survey Methodology; reported to STROBE; ethics: University of Genoa (CERA2020.13, 25 June 2020); digital informed consent.
Questionnaire (56 questions, 6 sections): exclusion criteria; demographics and lifestyle; cycling activity (years, membership, sessions and hours, coaching or scheduled training, devices, pedalling-technique training, races, other sports, stretching, past biomechanical visits); low back pain prevalence by a standard definition (lifetime, 12 months, 4 weeks; last episode; sciatica); back pain and cycling; management.Participants were not told in advance that the focus was back pain.
Recruitment: instant-messaging platforms, social networks and mailing lists of cycling companies contacted by telephone; 15-30 July 2020; no incentive. Sent to 1,738; 1,274 complete (73.3%), 266 incomplete, 198 excluded (89 vertebral fracture or back surgery, 23 spondylolisthesis grade III or more, 106 rheumatic disease; overlapping).
Who: age 43.18 ± 12.21 y; 146 women (11.5%) and 1,128 men (88.5%); BMI 23.20 ± 2.80; 44.8% sedentary jobs; cycling over 5 years 74.4%; 3-4 sessions a week 47.0%; coached or on a scheduled plan 26.3%; specific pedalling-technique training 64.4%; past biomechanical visit 54.2% (for pain or discomfort 37.4%, performance 33.7%, injury prevention 28.9%).
Analysis: binomial logistic regression, Wald backward stepwise, 14 candidate covariates (sex, age, BMI, sleep, work type, years cycling, sessions a week, stretching, coach or scheduled plan, other sports, competitions a year, past biomechanical visit, pedalling-technique training, back pain before cycling); Box-Tidwell for linearity; Bonferroni over 20 terms, significance p < 0.01.
Funding and interests: no external funding; the acknowledgements place the work within the DINOGMI Department of Excellence framework of the Italian ministry (MIUR 2018-2022); each author declares no competing interests. No author is printed with a bike-fit or industry affiliation.
Key Results
Prevalence (Table 4) and the final model for low back pain in the last 12 months (Table 6):
| Measure | Result |
|---|---|
| Lifetime low back pain | 702 (55.1%); episodes 6.6 ± 10.24 |
| Last 12 months | 337 (26.5%) |
| Last 4 weeks | 137 (10.8%) |
| Last episode (n of 702) | some days 80.4%, one month 10.4%, over 3 months 4.8%; NRS median 5 [4-7]; radiating to the leg 27.4% |
| Back pain before cycling (yes) | OR 4.20 (95% CI 3.21-5.40), p < 0.01 |
| Coached or scheduled training (yes) | OR 0.53 (0.37-0.74), p < 0.01 |
| Specific pedalling-technique training (yes) | OR 0.69 (0.51-0.94), p = 0.02 |
| Cycling 2-5 vs under 2 years | OR 0.48 (0.26-0.89), p = 0.02 |
| Cycling over 5 vs under 2 years | OR 0.93 (0.54-1.61), p = 0.80 |
| Sex (coding direction not stated) | OR 0.67 (0.42-1.07), p = 0.09 |
| Other sports regularly (yes) | OR 1.28 (0.97-1.69), p = 0.08 |
| Dropped from the model | age, BMI, sleep, sessions a week, stretching, competitions a year, past biomechanical visit |
Back pain on the bike: 581 riders (45.6%) had had back pain during or after cycling at least once; 431 of them (74.2%) had had back pain before taking up cycling. In 32.1% it came only under load, in 54.9% after a few hours whatever the effort.
Training at the time (only for the 17.4% whose back pain coincided with a training period): long threshold climbs (20-40 min) 32.6%, low-cadence strength climbs 22.7%, long easy climbs 14.2%, flat aerobic riding 14.2%; descriptive, no comparison group.
Management: while riding, 54.4% eased it by extending the back; 19.5% said back pain had affected their training; 47.3% of those with lifetime back pain had seen a professional (physiotherapist 63.3%, osteopath 50.0%, physician 38.1%, bike fitter 30.8%); 35.8% had taken medication (NSAIDs 79.3%; 27.8% on their own initiative).
After a bike-fit visit for pain (258 riders, recalled, no control): pain unchanged 3.5%, partly improved 31.4%, sharply improved 38.8%, gone 26.3%.
Beliefs: 31.3% said their back pain began after a change of set-up or a component; 34.5% of those with back pain on the bike thought their position might be the cause.
Numbers that do not add up: the authors set significance at p < 0.01 (Bonferroni over 20 terms, although 0.05/20 is 0.0025; our arithmetic) but call the p = 0.02 results significant, which fail either threshold, while the two at p < 0.01 would pass 0.0025 by their intervals (our arithmetic); the text says five predictors were significant while Table 6 marks four odds ratios (plus the overall years-of-cycling term); the model is reported as χ2(4) = 159.81, but a model with these seven covariates has at least 11 parameters (our arithmetic); and the "exclusive" prevalences in the text (35%, 14%, 6%) do not follow from the nested 55.1%, 26.5% and 10.8% (that would give 28.6%, 15.7% and 10.8%; our arithmetic).Sciatica (10.5% of the sample) appears only in the Discussion.
Mechanisms
Measured: associations only, between self-reported characteristics and self-reported back pain in the last 12 months.
Authors' interpretation: pedalling-technique training that spreads force more evenly round the stroke lowers peak lower-limb and lumbo-pelvic muscle activation and the fatigue that changes trunk and pelvic posture; coaching or a scheduled plan means planned, progressive loads and varied stimuli rather than long low-intensity riding; back pain before cycling fits previous back pain as a known marker of future episodes.
Inference: "back pain before cycling" is a history marker: most riders who get back pain on the bike already had back pain before cycling (74.2%), so the bike explains less of their back pain than they may think.
Inference: coached riders may differ in more than coaching (experience, motivation, the rest of their training), and riders with back pain may drop a structured plan; the cross-sectional odds ratio cannot separate these.
Inference: riders go to a bike fitter because of pain (37.4% of visits), which biases a fit towards looking harmful; its dropping out of the model is not evidence that a fit does not help, only that a past fit did not separate riders with and without back pain.
Practical Application
Why does my lower back hurt on the bike? A back history is the strongest marker: in 1,274 Italian amateurs, back pain before cycling went with 4.2× odds of back pain in the past year; coached riders had lower odds. Association.
Do not promise a fix from the bike or from drills: a past bike-fit visit did not stay in the model, and riders' recalled improvement after a fit for pain (96.5% some improvement, our arithmetic from the four categories) has no control. Pedalling-technique training (OR 0.69) missed the authors' own threshold; the base already says never to promise efficiency from pedalling drills (SYNTH-55). For position and back mechanics see REF-1266.
Painkillers: 35.8% had taken medication for back pain, 27.8% of them on their own initiative (printed as "the remaining" after 70.7% prescribed and 4.5% advised by friends or family, which would leave 24.8%; our arithmetic). Not from this paper; the base's own practice: never mask pain to train (SYNTH-73).
Red flags: pain travelling down a leg, numbness, night pain → clinician (SYNTH-73).
Limitations
Cross-sectional and self-reported. Exposures and back pain were recalled together; causes cannot be inferred, and the authors say so.
Model building. Backward stepwise selection on 14 candidates inflates apparent associations; two of the four "significant" odds ratios miss the stated Bonferroni threshold; the degrees of freedom reported do not match the model (our arithmetic).
Selection. Recruitment through messaging groups, social networks and cycling-company mailing lists; the 73.3% response rate is of people the link reached, so the denominator is not a defined population; 88.5% men; one country.
Timing. Collected in the second half of July 2020, a few weeks after Italy's spring 2020 COVID-19 restrictions; recent training may have been atypical (inference).
Exposure definitions. "Specific pedal training", "biomechanical visit" and "coach or scheduled training" are single yes/no questions with no detail on content or timing relative to the pain.
Reporting. Inconsistent prevalence breakdown (Fig. 3 text) and a sciatica figure not in the results tables.
Interests. No external funding; no competing interests declared; no bike-fit or industry affiliation.
How this entry was checked and changed
Why this level
cross-sectional survey with n = 1274 (rubric: cross-sectional measurement study ≥100), read from the full text (open access, CC BY 4.0). The prevalence figures carry that level; the odds ratios carry less: they are cross-sectional and self-reported, come from a backward-stepwise model, and two of the four called significant miss the authors' own Bonferroni threshold (p < 0.01)
Full text (checked 2026-10-27)
1274 amateur cyclists completed the questionnaire (response rate: 73.3% of 1738 sent); 198 (11.4%) were excluded by exclusion criteria (full text)
Mean age 43.18 (12.21); 146 women (11.5) and 1128 men (88.5) (full text)
Low back pain prevalence 55.1%, 26.5% and 10.8% in life, in the last 12 months and in the last 4 weeks (full text)
Final model: LBP before cycling OR 4.20 (3.21-5.40); being supervised by a coach or following a scheduled training OR 0.53 (0.37-0.74); specific pedal training OR 0.69 (0.51-0.94), p 0.02; cycling 2-5 vs less than 2 years OR 0.48 (0.26-0.89), p 0.02 (full text)
Bonferroni correction using all twenty terms: significance accepted when p < 0.01; the model reported as χ2(4) = 159.81, p < 0.01 (full text)
581 people (45.6%) reported LBP during or after cycling; 431 (74.2%) had LBP before starting cycling (full text)
Biomechanical visit 691 (54.2); reasons pain/discomfort 258 (37.4), improve performance 233 (33.7), injury prevention 200 (28.9); after the visit unchanged 9 (3.5), partially improved 81 (31.4), sharply improved 100 (38.8), completely disappeared 68 (26.3) (full text)
Among those with lifetime LBP, 47.3% turned to a professional; 35.8% took medications, NSAIDs 79.3% (full text)
Funding: this research did not receive external funding; each of the three authors declares no competing interests (full text)