Effects of a community gardening intervention on diet, physical activity, and anthropometry outcomes in the USA (CAPS): an observer-blind, randomised controlled trial.

The intervention was a community gardening programme — participants received a free plot in a community garden, seeds, seedlings, gardening tools, and a series of introductory gardening workshops. The comparator was a wait-list control group — these participants were told they would receive a garden plot after the one-year study ended, but during the study they continued their normal lives.

The primary outcome was dietary intake of fruits and vegetables (measured in servings per day). Secondary outcomes included:

• Total vegetable intake (servings/day)
• Total fruit intake (servings/day)
• Fibre intake (grams/day)
• Physical activity (minutes/week of moderate-to-vigorous physical activity, and minutes/week of sedentary time)
• Body weight (kg)
• Waist circumference (cm)
• Blood pressure (systolic and diastolic, mmHg)
• Glycated haemoglobin (HbA1c, %)
• Fasting glucose (mg/dL)
• Fasting insulin (μU/mL)
• C-reactive protein (mg/L)
• Total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides (all mg/dL)

Sample size: 291 adults (145 intervention, 146 control) who completed the study. Originally 371 were randomised, but 80 dropped out or were lost to follow-up.

Population: Adults aged 18–85 years living in the Denver, Colorado metropolitan area. All participants were:

• Not currently gardening (or had not gardened in the past two years)
• Willing to be randomly assigned to either start gardening immediately or wait one year
• Able to attend gardening workshops and maintain a garden plot
• English- or Spanish-speaking
• Not pregnant or planning to become pregnant during the study

Setting: Community gardens in low-to-moderate income neighbourhoods in Denver, Colorado, USA. Gardens were managed by a non-profit organisation (Denver Urban Gardens) that provided the plots, tools, and educational support.

Demographics: Mean age was 41 years (range 18–85). 85% were female. 44% were non-Hispanic White, 36% Hispanic/Latino, 10% Black, 4% Asian, and 6% other or mixed race. About 40% had a household income below $50,000 per year.

• Dietary intake: Measured using the National Cancer Institute's Automated Self-Administered 24-Hour Dietary Recall (ASA24) — participants completed three 24-hour dietary recalls (two weekdays and one weekend day) at baseline and again at the end of the study. This is a validated web-based tool that asks participants to recall everything they ate and drank in the previous 24 hours. The recalls were self-administered but guided by a computerised interviewer.
• Physical activity and sedentary time: Measured using ActiGraph GT3X+ accelerometers worn on the right hip for seven consecutive days at baseline and again at the end of the study. Participants wore the device during waking hours except when swimming or bathing. Data were processed using standard cut-points: moderate-to-vigorous physical activity (MVPA) was defined as ≥1952 counts per minute; sedentary time was defined as <100 counts per minute.
• Anthropometry: Body weight was measured using a calibrated digital scale (to 0.1 kg). Waist circumference was measured at the iliac crest using a non-stretchable tape (to 0.1 cm). Both were measured in duplicate and averaged.
• Blood pressure: Measured using an automated oscillometric device (Omron HEM-907XL) after five minutes of seated rest. Three readings were taken one minute apart, and the average of the last two was used.
• Biomarkers: Fasting blood samples were collected by venipuncture after a 12-hour fast. Samples were analysed for HbA1c, glucose, insulin, C-reactive protein, and lipid panel using standard clinical laboratory methods.
• Gardening participation: Self-reported via monthly online surveys asking how many hours per week participants spent gardening, what they grew, and how much they harvested.

Study design: This was a randomised controlled trial (RCT) with two parallel arms: immediate gardening intervention versus wait-list control. Randomisation was done at the individual level (not by garden site) using a computer-generated random sequence with random permuted blocks of sizes 2, 4, and 6, stratified by sex and age group (18–44 vs 45–85 years).

Blinding: This was an observer-blind trial. The outcome assessors (the research staff who measured weight, waist circumference, blood pressure, and collected blood samples) were blinded to group assignment. The participants and the gardening instructors were not blinded — obviously, participants knew whether they were gardening or not. The statisticians were also blinded until after the primary analyses were completed.

Duration: The intervention lasted one year (12 months). Baseline measurements were taken before randomisation, and follow-up measurements were taken at 12 months post-randomisation. There was no interim measurement at 6 months.

Statistical approach: The primary analysis used intention-to-treat — meaning all participants were analysed in the group they were randomised to, regardless of how much they actually gardened. The primary outcome (total fruit and vegetable intake) was analysed using a linear mixed model with adjustment for baseline values, age, sex, and race/ethnicity. Secondary outcomes were analysed similarly, with a Bonferroni correction for multiple comparisons applied to the secondary outcomes (but not the primary outcome). Sensitivity analyses included per-protocol analyses (excluding participants who gardened less than one hour per week) and analyses adjusting for seasonality.

What this design can prove:

• Because this is an RCT with random assignment, it can establish causality — the observed differences in diet and activity can be attributed to the gardening intervention, not to pre-existing differences between groups.
• The observer blinding reduces the risk of measurement bias (staff might unconsciously measure control participants differently if they knew their group).
• The one-year duration is long enough to capture seasonal effects (gardening is seasonal in Colorado) and to allow for habituation to the intervention.

What this design cannot prove:

• It cannot tell us why gardening improved diet and activity — was it the increased access to fresh produce, the physical activity of gardening itself, the social aspects of community gardening, or some combination? The study was not designed to test mechanisms.
• It cannot tell us whether the effects persist beyond one year — the wait-list control group received gardens after the study ended, so there was no long-term follow-up of the control group.
• The lack of participant blinding means that placebo effects or demand characteristics could have influenced self-reported dietary recalls (people who know they are in the gardening group might over-report vegetable intake).
• The study was conducted in one city (Denver) with a specific non-profit gardening programme, so results may not generalise to other settings or other types of gardening (e.g., home gardening, school gardening).

Major methodological weaknesses:

• High dropout rate: 80 of 371 randomised participants (22%) dropped out or were lost to follow-up. If dropouts differed systematically between groups (e.g., if less motivated people dropped out more from the gardening group), this could bias results.
• Self-reported diet: Despite using a validated tool, 24-hour dietary recalls rely on memory and honesty. People in the gardening group might have been more motivated to report healthy eating.
• Seasonal effects: Baseline and follow-up measurements were taken at the same time of year for each participant (one year apart), but different participants started at different times of year. This means some participants had their follow-up measurement in winter (when garden harvests are minimal) and others in summer (peak harvest season). The authors adjusted for seasonality in sensitivity analyses, but this is a limitation.
• Single geographic location: Denver has a specific climate (semi-arid, with cold winters and hot summers) and a specific community gardening infrastructure. Results might differ in other climates or with different gardening support systems.

Primary outcome — total fruit and vegetable intake:

• At baseline, both groups consumed about 3.5 servings/day of fruits and vegetables combined.
• At 12 months, the gardening group consumed 4.8 servings/day (adjusted mean), while the control group consumed 4.0 servings/day.
• Difference: +0.8 servings/day (95% CI: 0.3 to 1.3, p=0.001) — this was statistically significant and met the pre-specified primary outcome.

Secondary outcomes — dietary:

• Total vegetable intake: Gardening group increased by 0.6 servings/day more than control (95% CI: 0.2 to 1.0, p=0.002).
• Total fruit intake: Gardening group increased by 0.2 servings/day more than control (95% CI: 0.0 to 0.5, p=0.07) — not statistically significant after adjustment for multiple comparisons.
• Fibre intake: Gardening group increased by 1.6 grams/day more than control (95% CI: 0.3 to 2.9, p=0.02) — not significant after multiple comparison adjustment.

Secondary outcomes — physical activity:

• Moderate-to-vigorous physical activity (MVPA): Gardening group increased by 5.8 minutes/week more than control (95% CI: -8.2 to 19.8, p=0.42) — not significant.
• Sedentary time: Gardening group decreased by 38.6 minutes/week more than control (95% CI: -73.6 to -3.6, p=0.03) — this was statistically significant, meaning the gardening group sat about 40 fewer minutes per week.

Secondary outcomes — anthropometry and biomarkers:

• Body weight: Gardening group lost 0.3 kg more than control (95% CI: -1.2 to 0.6, p=0.50) — not significant.
• Waist circumference: Gardening group decreased by 0.3 cm more than control (95% CI: -1.2 to 0.6, p=0.50) — not significant.
• Systolic blood pressure: Gardening group decreased by 0.9 mmHg more than control (95% CI: -2.8 to 1.0, p=0.35) — not significant.
• Diastolic blood pressure: Gardening group decreased by 0.6 mmHg more than control (95% CI: -1.8 to 0.6, p=0.32) — not significant.
• HbA1c: Gardening group decreased by 0.02% more than control (95% CI: -0.06 to 0.02, p=0.37) — not significant.
• Fasting glucose: Gardening group decreased by 1.5 mg/dL more than control (95% CI: -4.2 to 1.2, p=0.27) — not significant.
• Fasting insulin: Gardening group decreased by 0.8 μU/mL more than control (95% CI: -2.5 to 0.9, p=0.35) — not significant.
• C-reactive protein: Gardening group decreased by 0.2 mg/L more than control (95% CI: -0.6 to 0.2, p=0.35) — not significant.
• Total cholesterol: Gardening group decreased by 1.5 mg/dL more than control (95% CI: -6.5 to 3.5, p=0.56) — not significant.
• LDL cholesterol: Gardening group decreased by 1.0 mg/dL more than control (95% CI: -5.4 to 3.4, p=0.66) — not significant.
• HDL cholesterol: Gardening group increased by 0.3 mg/dL more than control (95% CI: -1.2 to 1.8, p=0.70) — not significant.
• Triglycerides: Gardening group decreased by 3.0 mg/dL more than control (95% CI: -12.0 to 6.0, p=0.51) — not significant.

Gardening participation:

• Among those in the gardening group, the median time spent gardening was 2.5 hours per week (interquartile range: 1.0 to 4.5 hours/week).
• About 75% of gardening group participants gardened at least one hour per week throughout the growing season (April–October).
• During winter months (November–March), gardening participation dropped to near zero for most participants.

• Diet: The increase of 0.8 servings/day of fruits and vegetables is roughly equivalent to adding one medium apple and a handful of baby carrots to your daily diet. This is a modest but meaningful increase — public health guidelines recommend 5–9 servings/day, and most Americans eat only 2–3 servings/day. An increase of 0.8 servings/day would move someone from "below average" to "average" intake.
• Sedentary time: The reduction of 38.6 minutes/week of sedentary time is about 5.5 minutes per day — roughly the time it takes to walk to the garden, check on your plants, and walk back. This is a small reduction, but even small reductions in sedentary time have been linked to improved metabolic health in observational studies.
• Weight and waist: The lack of significant change in body weight (0.3 kg, or about 0.7 lbs) and waist circumference (0.3 cm, or about 0.1 inches) suggests that gardening alone, without calorie restriction or other lifestyle changes, is unlikely to produce meaningful weight loss. The increased vegetable intake might have been offset by increased calorie intake from other sources, or the physical activity of gardening (2.5 hours/week) may not have been intense enough to create a calorie deficit.
• Biomarkers: None of the blood markers changed significantly. This is not surprising given the modest changes in diet and activity — it typically takes larger changes in diet or weight loss to see improvements in blood pressure, cholesterol, or blood sugar.

Acknowledged by authors:

• High dropout rate (22%) and potential for differential dropout between groups.
• Self-reported dietary intake (recall bias, social desirability bias).
• Single geographic location (Denver, Colorado) limits generalisability.
• Seasonal variation in gardening participation (winter months had minimal gardening).
• Lack of blinding of participants (could affect self-reported outcomes).
• The wait-list control design means the control group knew they would eventually get gardens, which might have reduced their motivation to change behaviour on their own.

Additional critical observations:

• No measurement of total calorie intake: The study measured fruit and vegetable servings and fibre, but not total calories. It's possible that the gardening group increased their vegetable intake but also increased their overall calorie intake (e.g., by eating more salad dressing, cooking with oil, or snacking more). Without total calorie data, we cannot determine whether the dietary change was truly beneficial or just a substitution.
• No measurement of gardening intensity: The accelerometer measured overall physical activity, not specifically gardening activity. It's possible that gardening replaced other forms of physical activity (e.g., people who gardened might have gone to the gym less), which would explain why MVPA didn't increase significantly.
• No measurement of social or psychological outcomes: Community gardening might improve mental health, social connectedness, or stress levels, but these were not measured. The study focused only on physical health outcomes.
• No cost-effectiveness analysis: The intervention included free garden plots, tools, seeds, and workshops. It's unclear whether