Design of a RCT evaluating the (cost-) effectiveness of a lifestyle intervention for male construction workers at risk for cardiovascular disease: The Health under Construction study

This is a study protocol, meaning the authors describe what they *plan* to test, not what they found. The planned intervention is a 6-month lifestyle programme delivered through an Occupational Health Service (OHS). The intervention has three components:

**Physical activity (PA) counselling:** Individual sessions to increase daily physical activity (e.g., walking, cycling, work-related movement).

**Dietary behaviour counselling:** Advice to improve eating habits (e.g., reduce saturated fat, increase fruit/vegetable intake).

**Smoking cessation counselling:** Support to quit smoking, using motivational interviewing techniques.

The intervention is delivered through a combination of:

3 face-to-face sessions at the OHS (likely at months 0, 2, and 4 or similar spacing)

4 telephone contacts (likely between face-to-face sessions)

Written materials about healthy lifestyle

The comparator group receives **usual care** — meaning whatever standard health advice or medical check-ups construction workers normally get through their occupational health service. No active placebo or alternative intervention is given to the control group.

**Primary outcomes** (the main things they plan to measure):

Daily physical activity (minutes per week or MET-minutes)

Dietary intake (specific nutrients or food groups)

Smoking status (current smoker vs. non-smoker, or cigarettes per day)

**Secondary outcomes** (additional health markers):

Body mass index (BMI, kg/m²)

Systolic and diastolic blood pressure (mmHg)

Total and HDL blood cholesterol (mmol/L)

HbA1c (glycated haemoglobin, a marker of average blood sugar over ~3 months)

Cardiorespiratory fitness (CRF, likely measured by a submaximal exercise test)

Sickness absenteeism (days off work)

Cost-effectiveness (cost per unit of health improvement)

The study plans to recruit **692 male construction workers** from the Dutch construction industry. Inclusion criteria are:

Male (the construction workforce in the Netherlands is predominantly male)

Employed in the construction industry

Identified as having an elevated risk of cardiovascular disease (CVD) based on a health risk assessment

The authors note that approximately 20% of Dutch construction workers have elevated CVD risk, so the sample is drawn from that higher-risk subgroup. No specific age range is given, but construction workers typically range from ~20 to 65 years old. The setting is occupational health services within construction companies.

The protocol specifies these measurement tools (though exact instruments are not fully detailed in the abstract):

**Physical activity:** Likely measured by a validated questionnaire (e.g., the International Physical Activity Questionnaire, IPAQ, or the Short QUestionnaire to ASsess Health-enhancing physical activity, SQUASH). These ask about minutes per week of moderate and vigorous activity, walking, and sitting.

**Dietary intake:** Likely a food frequency questionnaire (FFQ) or 24-hour dietary recall, assessing intake of saturated fat, fibre, fruits, vegetables, and total energy.

**Smoking status:** Self-reported current smoking (yes/no) and number of cigarettes per day. May be biochemically verified with cotinine (a nicotine metabolite) in saliva or urine.

**BMI:** Measured height and weight (kg/m²).

**Blood pressure:** Measured with a standardised sphygmomanometer, likely after 5 minutes of seated rest, with multiple readings averaged.

**Blood cholesterol and HbA1c:** Fasting blood sample analysed in a laboratory.

**Cardiorespiratory fitness:** Likely a submaximal cycle ergometer test (e.g., the Åstrand-Ryhming test) to estimate VO₂max.

**Sickness absenteeism:** Company records or self-reported days off work due to illness.

**Cost-effectiveness:** Direct costs (intervention delivery, healthcare use) and indirect costs (productivity loss) measured via questionnaires and registries.

### Study design

This is a **randomised controlled trial (RCT)** — the gold standard for testing whether an intervention causes a change in outcomes. Participants are randomly assigned to either the intervention group (lifestyle counselling) or the control group (usual care).

### Randomisation

The authors state that 692 participants will be randomised. Randomisation is typically done using a computer-generated random sequence, with allocation concealed from the researchers enrolling participants (so they cannot influence who gets which treatment). The abstract does not specify the randomisation method in detail, but for a trial of this size, block randomisation (e.g., blocks of 4 or 6) is common to ensure equal group sizes.

### Blinding

This is an **open-label** trial — participants know whether they are receiving the lifestyle counselling or usual care, because it is impossible to blind someone to the fact that they are attending counselling sessions. The researchers measuring outcomes (e.g., blood pressure, blood samples) may be blinded to group assignment, but the abstract does not confirm this. Lack of participant blinding is a weakness because it can lead to **expectation bias** (people in the intervention group may report better behaviours because they feel they should, not because they actually changed).

### Duration

**Intervention period:** 6 months (3 face-to-face + 4 telephone contacts)

**Follow-up:** Measurements at baseline (0 months), post-intervention (6 months), and maintenance (12 months)

This means the study can assess both immediate effects (at 6 months) and whether any changes are sustained 6 months after the intervention ends.

### Statistical approach

The authors plan to use **multilevel analysis** (also called hierarchical linear modelling or mixed-effects models). This is appropriate because participants are nested within companies or occupational health services, and repeated measurements (baseline, 6 months, 12 months) are nested within individuals. Multilevel analysis accounts for this clustering and can handle missing data more robustly than traditional methods.

### What this design can and cannot prove

**Can prove:**

Whether the lifestyle intervention causes changes in physical activity, diet, smoking, and health markers compared to usual care (causality, because of randomisation)

Whether effects are sustained at 12 months (because of the follow-up)

Whether the intervention is cost-effective from a societal perspective

**Cannot prove:**

Which component of the intervention (PA counselling vs. diet vs. smoking) is most effective — because it is a multi-component package, not a factorial design

Whether the intervention works in women, other industries, or non-occupational settings (limited generalisability)

Long-term effects beyond 12 months (no longer follow-up)

Whether the effects are due to the specific counselling technique (motivational interviewing) or just the attention/contact time (no attention-placebo control group)

### Major methodological weaknesses (from the protocol)

**No blinding of participants** — expectation bias is possible

**No active control group** — usual care may be minimal, so any attention could produce a placebo effect

**Self-reported primary outcomes** — physical activity, diet, and smoking are all self-reported, which is prone to social desirability bias (people say they are healthier than they are)

**No biochemical verification of smoking** (unless added later) — self-reported quitting may be unreliable

**The protocol does not describe how they will handle dropouts** — construction workers may change jobs or be lost to follow-up

**This is a protocol paper — no results are reported.** The authors describe only the planned study design. There are no effect sizes, confidence intervals, or p-values because the trial has not yet been completed or analysed.

However, the paper provides useful context:

20% of Dutch construction workers have elevated CVD risk

The intervention is designed based on interviews with workers and current literature (so it is tailored to the population)

The trial is registered (ISRCTN60545588), meaning results should eventually be published

If you want the actual results, you would need to find the follow-up papers reporting the outcomes of this trial. A quick search suggests that the results were published in later papers (e.g., Groeneveld et al., 2010, 2011 in journals like *Occupational and Environmental Medicine* or *BMC Public Health*).

Not applicable — no results are reported in this protocol paper.

Since this is a protocol, the limitations are about the *planned* study design, not the findings:

**No blinding:** Participants know they are in the intervention group, which can inflate self-reported improvements.

**Self-reported outcomes:** Physical activity, diet, and smoking are all measured by questionnaire, not objective devices (e.g., accelerometers for PA, cotinine for smoking). This is a major source of bias.

**Single industry, single gender:** Results may not generalise to female workers, other industries, or non-occupational settings.

**Multi-component intervention:** Cannot isolate which part (PA, diet, smoking) drives any effects.

**No active control:** Usual care may be very minimal, so any attention from counsellors could produce a placebo effect.

**Attrition risk:** Construction workers have high job turnover, and 12-month follow-up may lose many participants.

**Cost-effectiveness analysis limitations:** Productivity costs are hard to measure accurately in construction (seasonal work, contract work).

**The protocol does not describe power calculations** — though 692 participants is a large sample, it is unclear whether it is sufficient to detect small effects on blood pressure or cholesterol.

For someone running their own n=1 experiment based on this study design:

### What to test

Test a **multi-behaviour lifestyle intervention** combining:

**Physical activity:** Aim for 150+ minutes of moderate activity per week (e.g., brisk walking, cycling, bodyweight exercises)

**Diet:** Reduce saturated fat (swap butter for olive oil, choose lean meats), increase fibre (vegetables, whole grains, legumes), and eat 5+ servings of fruits/vegetables daily

**Smoking cessation:** If you smoke, use a structured quit plan (e.g., nicotine replacement therapy or a quit app) combined with counselling

You could test all three simultaneously (like the study) or pick one behaviour to isolate. For a cleaner n=1, test one behaviour at a time.

### Minimum meaningful duration

**Intervention period:** 6 months (the study uses 6 months of counselling)

**Minimum to see a change in behaviour:** 4–8 weeks for habit formation

**Minimum to see a change in biomarkers:** 8–12 weeks for blood pressure and cholesterol; 3 months for HbA1c

**Maintenance check:** Measure again 3–6 months after stopping the intervention to see if changes stick

For a self-experiment, run the intervention for **12 weeks** (3 months) with weekly check-ins, then measure outcomes. If you want to match the study, extend to 6 months.

### What to measure (specific metrics)

**Primary (behavioural):**

**Physical activity:** Minutes per week of moderate-to-vigorous activity (use a pedometer, smartwatch, or log). Aim for ≥150 min/week.

**Diet:** Track servings of fruits/vegetables per day (target ≥5) and grams of saturated fat per day (target <10% of calories). Use a food diary app (e.g., MyFitnessPal).

**Smoking:** Cigarettes per day (if smoking). Target: 0.

**Secondary (health markers):**

**Body weight and BMI:** Weigh yourself weekly at the same time of day (morning, after bathroom, before eating).

**Blood pressure:** Measure at home with a validated monitor (e.g., Omron). Take 3 readings after 5 minutes seated rest, average them. Measure at the same time each day.

**Resting heart rate:** A proxy for fitness. Measure first thing in the morning before getting out of bed.

**Waist circumference:** A better marker of visceral fat than BMI. Measure at the narrowest point between ribs and hips.

**Blood lipids and HbA1c:** Get a blood test at baseline and after 12 weeks (your doctor can order this, or use a home finger-prick kit).

**Optional:**

**Cardiorespiratory fitness:** Do a submaximal test (e.g., 1-mile walk test or 3-minute step test) to estimate VO₂max.

**Sickness days:** Log any days you miss work or feel too unwell to exercise.

### Key confounds to control for

**Seasonal effects:** Physical activity and diet change with seasons (more walking in summer, more comfort food in winter). Run your experiment in a consistent season, or control for it by noting weather.

**Stress and sleep:** Both affect blood pressure, weight, and cravings. Track your sleep quality (e.g., hours slept, wake-ups) and daily stress level (1–10 scale) as covariates.

**Alcohol intake:** Alcohol affects blood pressure, weight, and motivation. Keep it constant or track it.

**Medication changes:** If you start or stop any medication (blood pressure, cholesterol, antidepressants), this will confound results. Document any changes.

**Work schedule:** Construction workers have variable hours. If you work shifts or have a physically demanding job, note your work type each day.

**Social support:** If you involve a partner or friend in your experiment, their influence is a confound. Decide whether to include them or go solo.

**Expectation bias:** You know you are testing an intervention, so you may unconsciously try harder. Use objective measures (e.g., step count from a device, not self-reported minutes) to reduce this.

### What a positive result would look like

A positive result in your n=1 experiment would mean:

**Behavioural changes:**

Physical activity increases by ≥30 minutes per week of moderate activity (e.g., from 90 to 120 min/week)

Fruit/vegetable intake increases by ≥2 servings per day (e.g., from 3 to 5)

Saturated fat intake drops below 10% of total calories

Smoking reduces to 0 cigarettes per day (or a ≥50% reduction if quitting is not the goal)

**Health marker changes (after 12 weeks):**

Body weight: Loss of 2–5 kg (if overweight)

Waist circumference: Reduction of 3–6 cm

Systolic blood pressure: Drop of 5–10 mmHg (if elevated at baseline)

Diastolic blood pressure: Drop of 3–5 mmHg

Resting heart rate: Drop of 5–10 bpm

Total cholesterol: Drop of 0.5–1.0 mmol/L (if elevated)

HDL cholesterol: Increase of 0.1–0.3 mmol/L

HbA1c: Drop of 0.2–0.5% (if pre-diabetic)

**What a negative result looks like:**

No change in any metric, or changes that are within normal day-to-day variation (e.g., weight fluctuates ±1 kg, blood pressure varies ±5 mmHg). This would suggest the intervention is not effective for you, or that you need a longer duration, a higher dose, or a different approach.

**How to interpret:**

Because this is an n=1, you cannot calculate p-values. Instead, look for **consistent trends** over time. Plot your data weekly (e.g., weight, blood pressure, steps). A positive result is a clear, sustained shift in the direction you want, not just a one-off good week. If you see a trend after 4–6 weeks, continue for the full 12 weeks to confirm. If you see no trend by 8 weeks, consider modifying the intervention (e.g., increase exercise intensity, change diet composition, add a quit-sm