Prospective Study of Bicycling and Risk of Coronary Heart Disease in Danish Men and Women

This was a prospective observational study that tested whether cycling (for commuting or leisure) was associated with a lower risk of developing coronary heart disease (CHD). The researchers compared:

**Cyclists vs. non-cyclists** at the start of the study (baseline)

**Different amounts of cycling** per week (0 hours, 0–1 hours, 1–2.5 hours, 2.5+ hours)

**Changes in cycling habits** over time (people who never cycled, stopped cycling, started cycling, or cycled consistently)

**Commuter cycling vs. leisure-time cycling** separately

The main outcome was incident CHD — either fatal or non-fatal heart attacks (classified using ICD-10 codes I21.0–I21.9 and I46.0–I46.9).

**53,723 Danish adults** (25,329 men and 28,394 women)

Aged **50–65 years** at recruitment (1993–1997)

Recruited from the "Diet, Cancer, and Health" prospective cohort study in Copenhagen and Aarhus, Denmark

Participants were free of stroke, CHD, and cancer at baseline

A subset of **45,264 participants** completed a second examination between 1999–2003 (about 6 years later)

For commuter cycling analyses, only **16,181 employed participants** were included (retired and unemployed excluded)

For change-in-habits analyses, **30,440 participants** had complete data at both time points

**Cycling exposure:** Self-reported via questionnaire at baseline and again at the second examination. Participants reported hours per week of:

- Overall cycling (commuting + leisure combined)

- Commuter cycling (cycling to/from work)

- Leisure-time cycling (recreational cycling)

**CHD incidence:** Identified through Danish national registries (Danish Civil Registration System and national patient registers) using ICD-10 codes. This is a major strength — registry data is complete and objective, not self-reported.

**Confounders:** Measured via questionnaire at baseline and second examination, including:

- Demographics: age, sex, years of school, educational level

- Lifestyle: smoking, alcohol intake, coffee intake, diet (total energy, whole grains, fruit, vegetables, glycemic load, fat ratio)

- Physical activity: occupational physical activity, leisure-time physical activity (other than cycling)

- Medical history: family history of CHD

- At baseline only: objectively measured body mass index (BMI), self-reported diabetes, use of hypertension medication, use of hypercholesterolemia medication

**Study design:** Prospective cohort study with repeated measures. This is NOT a randomized controlled trial — it is an observational study where researchers followed people over time and compared those who chose to cycle with those who did not.

**Duration:** 20 years of follow-up (1993–2013), with a second exposure measurement approximately 6 years after baseline (1999–2003).

**Statistical approach:** Cox proportional hazards regression models were used to estimate hazard ratios (HR) with 95% confidence intervals (CI). Three levels of adjustment were used:

1. Basic adjustment for age and sex

2. Multivariable adjustment for all measured confounders (age, sex, education, smoking, alcohol, coffee, diet, occupational and leisure physical activity, family history of CHD)

3. Additional adjustment for BMI, diabetes, hypertension medication, and cholesterol medication (to test whether these might be mediators or confounders)

**Population attributable fraction** was calculated to estimate what proportion of CHD cases could theoretically be prevented if everyone cycled.

**What this design can prove:**

A prospective cohort with 20 years of follow-up can establish **temporal sequence** (cycling came before CHD)

The large sample size (53,723) and 2,892 CHD events provide good statistical power

Repeated measurements allow analysis of **changes in behavior** over time

Registry-based CHD diagnosis is objective and complete (no loss to follow-up)

Extensive adjustment for confounders reduces (but does not eliminate) the risk of spurious associations

**What this design cannot prove:**

**Causation.** People who choose to cycle may differ from non-cyclists in ways that were not measured (residual confounding). For example, cyclists might be more health-conscious in general, have lower stress levels, live in more walkable neighborhoods, or have higher socioeconomic status — all of which could independently lower CHD risk.

**No randomisation.** The researchers did not assign people to cycle or not cycle. This means self-selection bias is a major concern.

**Self-reported cycling** is subject to recall bias and social desirability bias (people may over-report how much they cycle)

**No blinding** — participants knew they were in a health study, which could influence behavior

**Major methodological weaknesses:**

Cycling was measured at only two time points, but people's habits likely changed over 20 years

The second examination had substantial dropout (only 45,264 of 53,723 completed it)

Commuter cycling analysis had limited statistical power (only 463 CHD events among 16,181 employed participants)

No objective measurement of cycling (no GPS, no accelerometers, no bike odometers)

The population is relatively homogeneous (Danish, mostly white, aged 50–65 at baseline), limiting generalizability

**Primary outcome: Overall cycling at baseline and CHD risk (n=53,723, 2,892 CHD events over 846,487 person-years)**

Non-cyclists: CHD event rate of 4.31 per 1,000 person-years (reference group)

0–1 hour/week cycling (mean 0.8 h/w): HR 0.87 (95% CI 0.79–0.97) — **13% lower risk**

1–2.5 hours/week cycling (mean 1.8 h/w): HR 0.92 (95% CI 0.79–1.04) — **8% lower risk (not statistically significant)**

2.5+ hours/week cycling (mean 5.5 h/w): HR 0.86 (95% CI 0.78–0.94) — **14% lower risk**

Note: The risk reduction ranged from 11–18% depending on the adjustment model. The fully adjusted model (including BMI, diabetes, medications) showed slightly attenuated but still significant associations.

**Secondary outcome: Commuter cycling at second examination (n=16,181 employed, 463 CHD events)**

0 hours/week: reference

0–1.5 hours/week (mean 0.7 h/w): HR 0.81 (95% CI 0.60–1.07) — **not statistically significant**

1.5+ hours/week (mean 3.6 h/w): HR 0.94 (95% CI 0.72–1.22) — **not statistically significant**

**Secondary outcome: Leisure-time cycling at second examination (n=31,632, 1,233 CHD events)**

0 hours/week: reference

0–1 hours/week (mean 0.5 h/w): HR 0.84 (95% CI 0.73–0.97) — **16% lower risk**

1–2.5 hours/week (mean 1.7 h/w): HR 0.80 (95% CI 0.67–0.96) — **20% lower risk**

2.5+ hours/week (mean 5.7 h/w): HR 0.80 (95% CI 0.68–0.94) — **20% lower risk**

**Secondary outcome: Changes in cycling habits (n=30,440, 1,169 CHD events)**

Never cycled (at either baseline or second exam): reference

Cycled at baseline, stopped by second exam: HR 0.88 (95% CI 0.72–1.08) — **not statistically significant**

Did not cycle at baseline, started by second exam: HR 0.76 (95% CI 0.61–0.95) — **24% lower risk**

Cycled at both time points: HR 0.84 (95% CI 0.73–0.97) — **16% lower risk**

**Population attributable fraction:** 7.4% (95% CI 3.6–11.1%) of all CHD cases in this population could theoretically have been prevented if everyone cycled or continued cycling.

The risk reduction associated with cycling was modest but meaningful at the population level:

**13–14% lower CHD risk** for any cycling vs. none (about 1 in 7 fewer heart attacks)

**24–26% lower CHD risk** for people who started cycling during the study (about 1 in 4 fewer heart attacks)

The effect was roughly **constant across all cycling amounts** — even 0–1 hour per week showed similar benefit to 2.5+ hours per week

To put this in context: the absolute risk difference was about 1.4 fewer CHD cases per 1,000 person-years (from 4.31 to 2.87 per 1,000 person-years). For an individual, this means if 100 people cycled for 10 years, about 1–2 heart attacks would be prevented compared to 100 non-cyclists.

**Acknowledged by authors:**

Selection bias (people who cycle may be healthier in unmeasured ways)

Recall bias (self-reported cycling)

Social desirability bias (over-reporting cycling)

Limited generalizability to non-Danish populations

Limited statistical power for commuter cycling analyses

Possible inaccuracy in CHD case identification (though registry data is generally excellent in Denmark)

**Additional critical limitations:**

**No randomisation** — this is the single biggest limitation. The observed association could be entirely due to confounding by health-conscious behavior, socioeconomic status, or other unmeasured factors

**Self-reported cycling** is imprecise — people may round up or forget

**Only two measurement points** over 20 years — cycling habits likely fluctuated

**No dose-response relationship** — the benefit was similar for 1 hour/week and 5.5 hours/week, which is unusual for a true causal effect (though it could reflect a threshold effect)

**No objective physical activity measurement** — the researchers could not verify cycling intensity or duration

**Healthy volunteer bias** — participants in the "Diet, Cancer, and Health" study were likely healthier than the general Danish population

**No data on cycling intensity** (e.g., leisurely vs. vigorous cycling) — intensity may matter for cardiovascular benefit

**No data on cycling terrain** (flat vs. hilly) which affects energy expenditure

**The "started cycling" group** may have been motivated by health concerns or lifestyle changes that also included other healthy behaviors not fully captured by the confounders

For someone running their own n=1 experiment:

### What to test

**Intervention:** Regular cycling, either as commuting or leisure activity. Aim for at least 1 hour per week, but the data suggest even small amounts (0–1 hour/week) may be beneficial.

**Dose options to test:**

- Option A: 15 minutes of cycling, 4 days per week (1 hour total)

- Option B: 30 minutes of cycling, 5 days per week (2.5 hours total)

- Option C: Cycle commute (e.g., 20 minutes each way, 5 days = 3.3 hours total)

**Comparator:** Your baseline period (e.g., 4 weeks of no cycling, followed by 4–8 weeks of cycling)

### Minimum meaningful duration

**At least 8–12 weeks** of consistent cycling to see any measurable change in cardiovascular risk markers

The original study looked at 20-year outcomes, but intermediate biomarkers (blood pressure, resting heart rate, cholesterol, VO2 max) can change in 4–12 weeks

For a true n=1 experiment, aim for 4 weeks baseline + 8 weeks intervention + 4 weeks washout (if testing multiple doses)

### What to measure (specific metrics)

**Primary outcome (choose one or more):**

Resting heart rate (measure upon waking, before getting out of bed, average over 5 mornings)

Blood pressure (systolic and diastolic, same time each day, after 5 minutes seated rest)

Heart rate variability (HRV) — use a chest strap monitor, measure upon waking

VO2 max estimate (using a fitness test like Cooper 12-minute run or a cycling ergometer test)

Fasting blood lipids (LDL cholesterol, HDL cholesterol, triglycerides) — requires blood draw, but home test kits are available

**Secondary outcomes:**

Body weight and waist circumference

Perceived energy levels (daily 1–10 scale)

Sleep quality (e.g., using a sleep tracker or daily diary)

Mood (e.g., PHQ-9 or daily mood rating)

### Key confounds to control for

**Other physical activity:** Keep non-cycling exercise constant during the experiment. If you normally run 3x/week, don't change that.

**Diet:** Track your diet (use a food diary app) and keep it consistent. The cycling benefit could be confounded by dietary changes.

**Sleep:** Track sleep duration and quality. Poor sleep increases cardiovascular risk.

**Stress:** Log daily stress levels (1–10). Major life events can confound results.

**Alcohol and smoking:** Keep these constant. Even small changes affect cardiovascular risk.

**Seasonal effects:** Cycling in winter vs. summer involves different temperatures, daylight, and effort. Run your experiment in one season or control for weather.

**Medication:** If you take blood pressure or cholesterol medication, do not change dosage during the experiment without consulting your doctor.

### What a positive result would look like

**Resting heart rate:** Decrease of 3–5 beats per minute (from baseline average)

**Blood pressure:** Decrease of 3–5 mmHg systolic (from baseline average)

**HRV:** Increase of 10–20% in RMSSD or SDNN (from baseline average)

**VO2 max:** Increase of 3–5% (from baseline)

**LDL cholesterol:** Decrease of 5–10 mg/dL (from baseline)

**Subjective energy:** Consistent increase of 1–2 points on a 10-point scale

**Body weight:** No change expected unless cycling is combined with dietary changes

**Important caveat:** Even if you see no change in these short-term biomarkers, the long-term cardiovascular benefit from regular cycling (as shown in this study) may still be real. The biomarkers are proxies, not guarantees. The Danish study found benefit even with modest cycling (1 hour/week) — so don't assume more is always better. Start with a sustainable dose and track adherence, not just outcomes.