School-based gardening, cooking and nutrition intervention increased vegetable intake but did not reduce BMI: Texas sprouts - a cluster randomized controlled trial

The researchers tested a comprehensive school-based program called Texas Sprouts, which combined three components: (1) building a 0.25-acre outdoor teaching garden at each school, (2) delivering 18 one-hour student lessons on gardening, nutrition, and cooking during the school day across one academic year (9 months), and (3) offering nine monthly parent lessons. The comparator was a delayed intervention control group — schools that received the same program the following academic year but served as a no-treatment control during the study period. The primary outcomes were changes in vegetable intake, fruit intake, sugar-sweetened beverage consumption, BMI z-scores (a standardized measure of body mass index adjusted for age and sex), waist circumference, body fat percentage, and blood pressure.

The study enrolled 3,135 children in 3rd through 5th grade across 16 elementary schools in central Texas. The average age was 9.2 years. The sample was 64% Hispanic, 47% male, and 69% eligible for free and reduced-price lunch (a marker of low-income status). All schools had >50% Hispanic enrollment and >50% of students eligible for free/reduced lunch. Schools were located within 60 miles of central Austin and had no existing garden or gardening program at baseline.

Dietary intake was measured using a self-reported survey that asked children how frequently they consumed vegetables, fruits, and sugar-sweetened beverages. The survey asked about frequency per day (e.g., "How many times did you eat vegetables yesterday?"). This is a simple recall method, not a detailed food diary or 24-hour recall interview. Obesity outcomes were measured directly: height and weight were measured by trained staff to calculate BMI and BMI z-scores (using CDC growth charts); waist circumference was measured with a tape measure; body fat percentage was estimated using bioelectrical impedance analysis (a device that sends a weak electrical current through the body to estimate fat mass). Blood pressure was measured using an automated monitor. All measurements were taken at baseline (before the school year started) and at follow-up (at the end of the 9-month school year).

Study design: This was a cluster randomized controlled trial (cluster RCT). Sixteen schools were randomly assigned to either the Texas Sprouts intervention (8 schools) or a delayed intervention control (8 schools). Randomization was done by a biostatistician who was blinded to school identities, using a block randomization procedure. The intervention was implemented in three waves over three years (2016–2019), with 6 schools per wave in waves 1 and 2, and 4 schools in wave 3.

Why cluster randomization matters: Schools, not individual children, were the unit of randomization. This is critical because the intervention was delivered at the school level — all children in a given school received the same program. If they had randomized individual children within the same school, children in the control group might have been exposed to the garden or learned from friends in the intervention group (contamination). Cluster randomization accounts for the fact that children within the same school are more similar to each other than to children in other schools (the "intra-cluster correlation"). The statistical analysis used generalized weighted linear mixed models to account for this clustering effect.

Blinding: The biostatistician who performed randomization was blinded, but the study was not blinded for participants, teachers, or outcome assessors. Children, parents, teachers, and the educators delivering the lessons all knew which schools were receiving the intervention. Outcome assessors (staff measuring height, weight, waist circumference, and blood pressure) were not explicitly stated to be blinded, though they were likely aware of school assignment since gardens were visible. This is a significant limitation — knowledge of group assignment can influence behavior and reporting.

Duration: The intervention lasted one full academic year (9 months). Measurements were taken at baseline (fall) and follow-up (spring). There was no long-term follow-up after the intervention ended, so we don't know if effects persisted.

Statistical approach: The primary analysis used complete cases (only children with both baseline and follow-up data). They also performed analyses with multiple imputation for missing data to check robustness. They reported results as adjusted mean changes from baseline, with p-values and 95% confidence intervals. They controlled for sex, age, ethnicity, and free/reduced lunch status as covariates.

What this design can and cannot prove: A cluster RCT is the gold standard for establishing causality when the intervention is delivered at the group level. Because schools were randomly assigned, any differences between groups at follow-up can be attributed to the intervention (assuming no major confounds). However, this design cannot tell us which specific component of the intervention (gardening, cooking, nutrition lessons, parent lessons, or the garden itself) caused the effects — it tests the whole package. It also cannot tell us whether the effects would persist beyond the 9-month intervention period, or whether they would generalize to other populations (e.g., non-Hispanic, higher-income, or older children).

Major methodological weaknesses: (1) No blinding of participants or outcome assessors — this introduces potential bias, especially for self-reported dietary intake. (2) Dietary intake was measured by a simple frequency survey, not a validated 24-hour recall or food diary, which is less accurate. (3) Attrition — not all children completed follow-up measurements, and while they used imputation, missing data can still bias results. (4) The control group received the intervention the following year, which is ethical but means the control group knew they would eventually get the program, potentially reducing motivation to change behavior on their own. (5) Only 16 schools were randomized, which is a small number for a cluster RCT — with only 8 schools per arm, a single unusual school could skew results.

Primary outcome — Vegetable intake:

• Intervention group: increased by +0.48 frequency/day from baseline
• Control group: increased by +0.04 frequency/day from baseline
• Difference between groups: +0.44 frequency/day (p = 0.02)
• This was statistically significant — the intervention caused a meaningful increase in vegetable consumption.

Secondary outcomes — No significant effects:

• Fruit intake: No significant difference between groups (p > 0.05)
• Sugar-sweetened beverage intake: No significant difference between groups (p > 0.05)
• BMI z-score: No significant difference between groups (p > 0.05)
• Waist circumference: No significant difference between groups (p > 0.05)
• Body fat percentage: No significant difference between groups (p > 0.05)
• Systolic blood pressure: No significant difference between groups (p > 0.05)
• Diastolic blood pressure: No significant difference between groups (p > 0.05)

Subgroup analyses (not pre-specified, so interpret cautiously):

• The authors reported that among children who were overweight or obese at baseline (BMI ≥85th percentile), there was a trend toward reduced BMI z-score in the intervention group compared to control, but this did not reach statistical significance (p = 0.07).
• No significant interactions by sex, ethnicity, or free/reduced lunch status were found.

Attrition and missing data:

• Of 3,135 enrolled children, approximately 2,400 had complete data at follow-up (about 77% retention).
• Results were similar when using multiple imputation for missing data, suggesting attrition did not substantially bias the findings.

The increase in vegetable intake was about half a serving per day (0.48 frequency/day in the intervention group vs. 0.04 in controls). To put this in perspective: if a child was eating vegetables once per day at baseline, they increased to about 1.5 times per day after the program, while control children stayed at about 1 time per day. This is a modest but meaningful increase — roughly equivalent to adding one additional serving of vegetables every two days. For comparison, the U.S. Dietary Guidelines recommend children aged 9–13 eat 2–3 cups of vegetables per day, so this intervention closed about 15–25% of the gap between typical intake and recommendations.

The lack of effect on BMI, waist circumference, body fat, and blood pressure means that even though children ate more vegetables, this did not translate into measurable changes in body composition or cardiovascular risk factors over 9 months. This could be because: (1) the increase in vegetables was too small to affect energy balance; (2) children may have compensated by eating less of other healthy foods or more of unhealthy foods; (3) 9 months is too short to see changes in body composition from dietary changes alone; or (4) the intervention did not reduce total calorie intake or increase physical activity enough to shift energy balance.

What the authors acknowledge:

• The dietary intake measure was a simple frequency survey, not a validated 24-hour recall or food diary, which limits accuracy.
• The study was not powered to detect subgroup effects (e.g., by sex, ethnicity, or baseline weight status).
• The intervention was only 9 months — longer exposure may be needed to see effects on BMI.
• The control group received delayed intervention, which may have reduced contrast between groups if control schools were motivated to improve on their own.
• The study was conducted in low-income, predominantly Hispanic schools in central Texas, limiting generalizability to other populations.

What a critical reader would note:

• No blinding of participants, teachers, or outcome assessors — this is a major source of potential bias, especially for self-reported dietary intake. Children in the intervention group knew they were supposed to eat more vegetables and may have over-reported.
• The vegetable intake increase (0.44 servings/day) is small and may not be clinically meaningful for weight or metabolic outcomes.
• The study did not measure total energy intake, physical activity, or other dietary changes (e.g., did children eat fewer calories from other sources, or did they just add vegetables on top of their usual diet?).
• The lack of effect on fruit intake is puzzling — if the program improved overall dietary habits, you'd expect fruit intake to increase too. This suggests the effect was specific to vegetables, or that the vegetable finding could be a chance result.
• Only 16 schools were randomized — with such a small number of clusters, the study may have been underpowered to detect small-to-moderate effects on BMI, especially if there was high variability between schools.
• The study did not measure long-term follow-up, so we don't know if the vegetable intake increase was maintained after the program ended.
• The parent component (9 monthly lessons) had low attendance — the authors noted that parent engagement was challenging, which may have limited the intervention's reach into the home environment.

For someone running their own n=1 experiment (or a small group experiment):

What to test:

• A combined gardening, cooking, and nutrition education program. The specific dose was 18 one-hour lessons over 9 months (about 2 lessons per month) plus hands-on gardening and cooking activities. For a self-experiment, you could test: "If I spend 2 hours per week gardening and cooking with vegetables, and attend a nutrition class once per month, will my vegetable intake increase?"
• The key active ingredients appeared to be: (1) hands-on exposure to vegetables (growing and cooking them), (2) repeated tasting opportunities (7 garden taste-tests and 11 cooking activities), and (3) culturally tailored recipes and content.

Minimum meaningful duration:

• 9 months was the intervention length in this study. For a self-experiment, a minimum of 3 months would be reasonable to see if vegetable intake changes, but 6–9 months would be better to assess whether changes are sustained. For body composition changes (BMI, waist circumference), this study suggests 9 months may not be enough — you might need 12–18 months of sustained dietary change.

What to measure (specific metrics):

• Primary: Vegetable intake (servings per day) — use a food diary or a validated app like MyFitnessPal or Cronometer. Record all vegetables eaten, including fresh, frozen, canned, and cooked. Be specific about portion sizes (e.g., 1 cup raw leafy greens = 1 serving, ½ cup cooked vegetables = 1 serving).
• Secondary: Fruit intake (servings per day), sugar-sweetened beverage intake (ounces per day), total calorie intake, and fiber intake (grams per day).
• Body composition: Weight, waist circumference (measured at the narrowest point between ribs and hip bones), and body fat percentage (if you have access to bioelectrical impedance scales or calipers). Measure at the same time of day, under the same conditions (e.g., morning, after bathroom, before eating).
• Blood pressure: If you have a home blood pressure monitor, measure at the same time each day (e.g., morning before breakfast, after sitting quietly for 5 minutes).

Key confounds to control for:

• Seasonality: Vegetable intake naturally varies with seasons (more fresh produce in summer/fall). Run your experiment across at least one full season, or compare same-season periods.
• Other dietary changes: If you start eating more vegetables, you might also change other eating habits (e.g., eating out less, cooking more at home). Track total diet, not just vegetables.
• Physical activity: Changes in exercise can affect body composition independently of diet. Keep activity levels constant or track them.
• Stress and sleep: Both affect appetite and food choices. Track sleep quality and stress levels (e.g., using a daily 1–10 rating).
• Social environment: If you're cooking with family or friends, their habits may influence yours. Note whether others in your household are also changing their diet.
• Expectation bias: If you know you're supposed to eat more vegetables, you might over-report intake. Use objective measures when possible (e.g., photograph all meals, weigh vegetables before cooking).

What a positive result would look like:

• Vegetable intake: An increase of at least 0.5 servings per day (the effect size in this study) sustained for 3+ months. A meaningful clinical target would be reaching 2–3 servings per day (the recommended amount for adults).
• Body composition: A decrease in waist circumference of 1–2 cm over 6 months, or a decrease in body fat percentage of 1–2% over 6–12 months. Note that this study found no effect on these measures, so don't expect large changes from vegetable intake alone.
• Blood pressure: A decrease of 2–5 mmHg in systolic blood pressure over 6–12 months (if you were borderline high at baseline). Again, this study found no effect.
• What to watch for: If you increase vegetable intake but also increase total calories (e.g., by adding high-calorie dressings, oils, or cheese), you might not see body composition changes. The goal is to replace less healthy foods with vegetables, not just add vegetables on top of your usual diet.