Promoting Water Consumption for Prevention of Overweight in School Children in a Controlled Intervention Trial: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Promoting Water Consumption for Prevention of Overweight in School Children in a Controlled Intervention Trial (NCT00554294), a randomized na study. The primary outcome is Overweight.

2. Clinical Question

The clinical question is framed around a binary outcome: each participant either met the specified overweight criterion or did not. The registry identifies prevalence of overweight as the primary outcome and gives the timeframe as one school year. The posted counts are 385 of 1641 in the intervention group and 683 of 1309 in the control group. These are group-level counts, not individual trajectories; they do not show which participants changed status during the study.

The intervention combined environmental resources with lessons, so the comparison addresses the school program as a package. It does not isolate the contribution of a water dispenser, a drinking bottle, or a lesson. The registry lists intake of drinks, physical activity and inactivity, dispenser water flow, and process-evaluation parameters such as acceptance and feasibility among secondary outcomes. Those topics add context about implementation and behavior, but they do not change the definition of the primary comparison.

The record does not report the age distribution, baseline overweight status, or other eligibility details for the analyzed participants. It therefore supports a description of the comparison in the enrolled school population, but not a more detailed characterization of which children were represented.

3. Trial Design

The registry describes a completed, randomized, parallel-group intervention with no masking. It lists the enrollment as 2950 and the study dates as 2006-04 through primary completion in 2007-06. The stated study period covered one school year. The brief description places the study in 32 elementary schools in two German cities.

The intervention group received water dispensers, drinking bottles and lessons, while control schools continued the usual curriculum without the environmental intervention. Thus, the contrast is between a package of school-based environmental and behavioral changes and the usual school curriculum. The record does not describe any additional intervention components beyond those named.

Although the registry identifies random allocation, the unit of allocation is not reported in the registry record. The description refers to intervention schools and control schools, but the record does not specify the allocation procedure or whether assignment was carried out at the school or participant level. The record also does not describe how analyses accounted for the possibility that children in the same school may have related outcomes. This matters when interpreting uncertainty because participants within a school may not contribute wholly independent information; whether and how that issue was addressed cannot be established from the posted record.

Masking is listed as none. The record does not explain which study staff, outcome assessors, or participants knew group assignment, nor does it report blinding of outcome measurement. The primary outcome was a count of participants meeting an overweight definition, with the timeframe specified as one school year. The registry lists five outcome measures and one posted analysis, alongside four secondary outcome topics.

The protocol’s sample-size assumptions, allocation ratio, planned analysis population, missing-data procedures, interim-analysis plans, and any prespecified adjustment variables are not reported in the registry record. A separate planning illustration based on the observed effect is available, but it is not a statement of the original trial design.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Overweight, measured over one school year.

The primary endpoint is the number of participants classified as overweight after one school year. The registry defines overweight according to criteria from the International Obesity Task Force. The endpoint is binary at the participant level: a participant either meets the stated classification or does not. The posted results give 683 participants meeting the definition among 1309 in the control group and 385 among 1641 in the intervention group.

Prevalence is used to describe how common a defined condition is in a group at a specified time or over a specified period. A binary endpoint makes it possible to compare the proportion meeting the criterion in each group. The record posts an unadjusted comparison expressed as a risk difference, as well as a sponsor analysis using logistic regression and reporting an odds ratio.

The endpoint describes group status, not necessarily change in status for each child. The record does not provide participant-level baseline and follow-up classifications or explain whether the reported counts represent only participants with complete measurements. It also does not report details of outcome ascertainment, such as who measured or classified participants. Those omissions limit what can be inferred about individual change and measurement procedures.

Using a recognized classification gives the outcome a defined threshold, but the registry does not include further detail about how the criteria were applied in this study. The result should therefore be read as a comparison of the posted counts under the registry’s stated definition, rather than as a description of all dimensions of children’s health or of water consumption itself.

7. Results Posted in the Registry

• Intervention Group: 23.5% (385/1641) versus 52.2% (683/1309) on control; risk difference -28.7 percentage points (95% CI -32.1 to -25.3), risk ratio 0.45 (0.41–0.50); unadjusted two-proportion z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is binary: participants either met the stated overweight criterion or did not. The posted counts let the outcome be summarized as a proportion in each group. A comparison can then be expressed on several statistical scales. The registry’s sponsor analysis uses logistic regression and reports an odds ratio. The recalculation from the posted counts gives a risk difference, a risk ratio, and an odds ratio. Each measure answers a related but distinct question, and numerical estimates on different scales should not be treated as interchangeable.

What does a risk difference describe?

The risk difference subtracts the control-group proportion from the intervention-group proportion. It is an absolute comparison on the proportion scale. The recalculated estimate is -0.287, with a 95% confidence interval from -0.321 to -0.253. A negative value indicates that the observed proportion was lower in the intervention group under this ordering of the comparison. The confidence interval describes a range of values compatible with the data and the calculation’s assumptions. It is not a range containing every participant’s outcome and does not resolve questions about bias or the study design.

How do relative measures differ?

A risk ratio compares the two group proportions by division, whereas an odds ratio compares the odds of the event between groups. Odds are a transformation of the event proportion, so an odds ratio is not the same quantity as a risk ratio or a risk difference. The unadjusted recalculation reports a risk ratio of 0.450, with a 95% confidence interval from 0.406 to 0.498, and an odds ratio of 0.281, with a 95% confidence interval from 0.24 to 0.329. The sponsor’s logistic-regression analysis reports an odds ratio of 0.69, with a 95% confidence interval from 0.48 to 0.98. The record does not provide enough model detail to establish why the sponsor’s odds-ratio estimate differs from the unadjusted calculation.

What does logistic regression contribute?

In general, logistic regression models the probability of a binary outcome through its odds. The resulting odds ratio can describe a comparison between groups, and regression can incorporate specified explanatory variables when they are included in the model. The registry reports the method as logistic regression but does not identify variables in the model, the precise model form, or whether school-level clustering was incorporated. It would therefore be inaccurate to label the posted estimate as adjusted for baseline characteristics or for the school structure.

In general, over-dispersion or dependence can matter when outcomes are more variable than a model assumes, or when observations are correlated. In a school-based study, outcomes among children in the same school may be related. If that dependence is ignored, standard errors and confidence intervals may not reflect the structure of the data. The registry does not report whether this issue arose or how it was handled; this is a general consideration, not a claim that the sponsor’s analysis omitted a necessary method.

How should p-values be read?

A p-value measures how unusual data at least as inconsistent with a specified null model would be if that model and its assumptions held. It does not give the probability that the null hypothesis is true, nor does it measure the probability that a finding is due to chance in a general sense. The unadjusted recalculation reports p < 0.001. The sponsor reports p < 0.05 for the logistic-regression result. These are results from different analyses and should be interpreted in relation to their respective estimates and assumptions.

Statistical significance is not a measure of practical importance. The risk difference provides an absolute-scale estimate, while relative measures provide comparisons on ratio scales. Assessing importance also requires context about the outcome, setting, implementation, and consequences, none of which can be reduced to a p-value. No clinical recommendation follows from these statistical summaries alone.

What is the role of recalculation?

The recalculated estimates are simple, unadjusted comparisons based on the posted group counts. They do not reproduce all possible details of the sponsor’s analysis and cannot account for information absent from the registry, such as covariates or clustering. The sponsor’s logistic-regression result is the analysis reported by the registry. Because the discrepancy indicator is false, the record does not flag a discrepancy requiring one estimate to take precedence over the other. The different effect scales and unavailable regression details should nonetheless remain explicit when comparing them.

The registry lists five outcome measures, four secondary outcome topics, and one posted analysis. It does not report whether the analysis was prespecified, whether additional analyses were conducted but not posted, or how multiplicity was handled. When many outcomes or analyses are considered, the chance of finding at least one apparently strong result can increase. Whether that concern applies to the trial’s full analysis program cannot be determined from the registry record.

10. Confidence Intervals and Precision

A confidence interval communicates the uncertainty around an estimated group contrast under the statistical method used. It does not show the range of individual participant outcomes, and its nominal coverage depends on the method’s assumptions. A narrower interval indicates more precision on the scale being reported; it does not guarantee that the estimate is unbiased or that the analysis accounts for every feature of the design.

For the unadjusted risk difference, the estimate is -0.287 and the 95% confidence interval runs from -0.321 to -0.253. The interval is entirely on the negative side of zero, so the posted unadjusted comparison is inconsistent with no difference under this calculation. The interval gives a range of plausible risk-difference values according to the calculation, but it cannot address unreported missingness, outcome measurement, or school-level dependence.

The recalculated risk ratio is 0.450, with a 95% confidence interval from 0.406 to 0.498. The recalculated odds ratio is 0.281, with a 95% confidence interval from 0.24 to 0.329. The sponsor’s logistic-regression odds ratio is 0.69, with a 95% confidence interval from 0.48 to 0.98. The ratio-scale intervals are interpreted relative to one, the value corresponding to no ratio difference. Estimates and intervals from different analyses should not be compared as though they share the same adjustment or assumptions.

The sponsor’s interval is less far from the no-difference value than the recalculated odds-ratio interval. The record does not state the regression variables or other model details, so the difference cannot be attributed to a particular adjustment. The point estimates and confidence intervals show statistical uncertainty conditional on the analyses; they do not include uncertainty about unreported design decisions or the extent to which the recorded outcome represents the intended comparison.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are illustrations based on fractions of the observed effect, not the trial’s original design assumptions. For this binary endpoint, the reported figures are participants per arm, rather than total events. The assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is the stated fraction of the observed effect. The trial’s own original sample-size justification is not reported in the registry record.

The 80% power column gives the number of participants per arm in the planning illustration for a design with an 80% chance of detecting the specified effect if that effect and the assumptions used for the calculation were true. The 90% power column gives the corresponding number per arm for a 90% chance. These are planning properties, not guarantees about what any particular study will detect. They are not evidence about the probability that the observed result is true.

For planning for 100% of the observed effect, the figures are 44 participants per arm for 80% power and 58 participants per arm for 90% power. For planning for 75% of the observed effect, they are 81 and 108 participants per arm, respectively. For planning for 50% of the observed effect, they are 188 and 251 participants per arm, respectively. Smaller effects are harder to distinguish from random variation, so the illustrations require more participants per arm as the target effect becomes smaller. The displayed planning calculations are not observed power.

These figures should not be read as a retrospective assessment of whether the trial was adequately designed. They do not reproduce the trial’s original assumptions, and they do not incorporate any unreported allowance for school-level dependence, missing data, or other design features. The trial’s original sample-size justification and planning assumptions are not reported in the registry record.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.522; treatment proportion p₁ = 0.235; average p̄ = 0.378.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6858 = 1.3441.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6551 = 0.5513.
4. n = (1.3441 + 0.5513)² / (-0.2872)² = 43.6, rounded up to 44 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.521772345301757                          # control proportion
p1 <- 0.234613040828763                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(554294)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 44 (80% power), 58 (90% power)
 75% of observed effect: n per arm = 81 (80% power), 108 (90% power)
 50% of observed effect: n per arm = 188 (80% power), 251 (90% power)
power.prop.test: 43.6 
Simulated power at n = 44 per arm: 0.817 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The registry reports the sponsor’s analysis as logistic regression, with an odds ratio of 0.69, a 95% confidence interval from 0.48 to 0.98, and p < 0.05. This is the sponsor-reported analysis of the comparison. The registry does not report which explanatory variables were included, whether the model accounted for school-level grouping, or other details needed to reproduce the regression.

The separate recalculation uses the posted counts and is unadjusted. It reports a risk difference of -0.287, a risk ratio of 0.450, and an odds ratio of 0.281. These summaries are calculated on different scales from one another, and the regression estimate may reflect model features that are not described in the registry. The record does not permit a determination of which model choices explain the difference between the two odds-ratio estimates.

The discrepancy indicator is false, so the registry does not flag a discrepancy between the posted analysis and the recalculation. This does not make the estimates identical or establish that they use equivalent methods. The recalculation is useful as a transparent summary of the group counts, while the sponsor’s estimate remains the registry’s own reported logistic-regression result. Neither analysis should be described as accounting for covariates or school clustering unless the registry documents that feature.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists five outcome measures, including overweight as the primary outcome and four secondary outcome topics: intake of drinks, physical activity and inactivity, water flow of the dispensers, and process-evaluation parameters such as acceptance and feasibility. It also lists one posted analysis. The record does not describe whether other analyses were conducted or whether the listed outcomes represent the complete set considered by investigators.

Multiplicity refers, in general, to the interpretive issue that arises when several outcomes, comparisons, or analyses are examined. With more opportunities to test a hypothesis, the chance of at least one result appearing statistically notable can rise. Some studies address this with prespecified testing hierarchies or adjustments, but the registry does not report a multiplicity plan or correction for this trial. It is therefore not possible to determine from the record whether the posted p-values account for other outcomes or analyses.

The record does not report how missing outcome data were handled, the number of participants with outcome measurements in each analysis, or whether participants were excluded after enrollment. The posted group counts have denominators of 1309 and 1641, but the registry does not explain their relationship to the enrolled population or describe any missing-data assumptions. No imputation method or sensitivity analysis is reported.

The analysis population is also not defined in the registry record. It does not say whether the analysis followed an intention-to-treat principle, included only participants with complete outcome data, or used another population. The record gives a randomized allocation designation, but does not describe adherence, crossovers, or how participants were analyzed in relation to their assigned group.

Finally, the randomization unit and treatment of school-level clustering are not reported. These are relevant design and analysis details for a school-based comparison. Their absence does not establish that the analysis was inappropriate, but it limits independent assessment of the reported uncertainty and of the safeguards against selective analysis.

15. What the Result Does, and Does Not, Mean

The posted results show that fewer participants in the intervention group met the registry’s overweight definition than in the control group at the one-school-year assessment: 385 of 1641 compared with 683 of 1309. The unadjusted risk-difference estimate is -0.287, with a 95% confidence interval from -0.321 to -0.253 and p < 0.001. The sponsor’s logistic-regression analysis reports an odds ratio of 0.69, with a 95% confidence interval from 0.48 to 0.98 and p < 0.05.

These statistics describe the observed comparison and its uncertainty under the respective analyses. They do not, by themselves, show that the water-promotion program caused the difference. The registry identifies random allocation, but does not report the allocation unit, whether the analysis accounted for school-level dependence, missing-data handling, or the analysis population. Without those details, the strength of the causal interpretation cannot be fully evaluated from the registry record.

The comparison concerns a combined environmental and behavioral program: water dispensers, drinking bottles and lessons. It cannot isolate the effect of any one component. Nor does the primary endpoint directly measure water consumption; the registry lists intake of drinks as a secondary outcome. The posted primary result is about classification under the stated overweight definition.

The result also does not establish whether any observed group difference persisted beyond one school year, how participants’ individual status changed from baseline, or how the program affected outcomes not included in the posted analysis. Such information is not reported in the registry record. The statistics should not be extended to populations, settings, or time periods not described by the trial.

In general, statistical evidence and practical or clinical importance are related but separate questions. The estimates quantify a group contrast, while judgment about importance requires context that the registry record does not fully provide. The result is not a treatment recommendation and should not be interpreted as proof that a particular intervention component is effective or ineffective.

16. Why This Trial Matters Statistically

This trial is statistically informative because it makes a group comparison using a binary outcome in a school-based intervention. The counts, denominators, absolute contrast, and ratio measures illustrate how the same underlying outcome can be described on different scales. The absolute risk difference is useful for understanding the size and direction of the observed proportion contrast, while ratios express a relative comparison. Both forms need their own uncertainty intervals and careful interpretation.

The example also highlights the distinction between a registry-reported model and a simple unadjusted recalculation. The sponsor reports logistic regression and an odds ratio, while the recalculation from group counts yields unadjusted estimates. Without model details, readers cannot assume these are estimates of exactly the same quantity under identical assumptions. Transparent interpretation means identifying the method behind each estimate rather than treating the values as interchangeable.

The school setting makes design structure important. Participants may be grouped within schools, while the registry does not report the randomization unit or how the analysis handled potential clustering. In general, these features can influence standard errors and confidence intervals. Recognizing that uncertainty is distinct from claiming the analysis did or did not handle it.

Finally, the record demonstrates why statistical evidence must be read alongside reporting completeness. The posted counts and estimates are interpretable, but missing details about sample-size justification, outcome completeness, analysis populations, multiplicity, and model specification constrain independent appraisal. A clear result is not the same as a fully documented analysis plan. Separating what is reported from what remains unknown is central to sound statistical reading.

17. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several limitations are specific to the information available for this trial. The registry does not state whether randomization occurred at the school or participant level, although the arms are described as intervention schools and control schools. It also does not report how the analysis accounted for the school setting. The posted confidence intervals and p-values therefore cannot be independently assessed for their handling of potential within-school dependence.

The sponsor’s logistic-regression model is not described beyond its method and reported odds ratio. Covariates, model specification, any clustering adjustment, and model diagnostics are not reported in the registry record. The difference between the sponsor’s odds ratio and the unadjusted odds ratio recalculated from counts cannot be explained from the record alone.

The record does not describe participant flow, baseline group characteristics, missing outcome data, or the analysis population. It does not state whether all randomized participants were included, how incomplete measurements were treated, or whether the posted denominators reflect the full enrolled groups. Outcome ascertainment procedures and the individuals responsible for measurement are also not described.

Although the registry lists secondary outcome topics, it does not report a multiplicity strategy or a complete analysis plan. The original sample-size assumptions and justification are not reported. The planning illustration based on the observed effect is not a substitute for those design details.

The brief description places the study in 32 elementary schools in two German cities, but the registry record does not provide enough information to characterize participant demographics or the broader school context. The primary outcome is reported for one school year; longer-term outcomes are not described in the posted results. These limits affect interpretation of generalizability, durability, and the completeness of the statistical assessment.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Promoting Water Consumption for Prevention of Overweight in School Children in a Controlled Intervention Trial — NCT00554294.
• Registry record last updated 2009-07-16; status last verified 2009-05.
• Results first posted 2009-07-16.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

21. Record Summary

The registry describes a completed randomized, parallel-group school intervention with no masking. Enrollment was 2950, and the study period ran from 2006-04 to primary completion in 2007-06. The brief description identifies 32 elementary schools in two German cities and a one-school-year intervention focused on promoting tap-water consumption.

Intervention schools received water dispensers, drinking bottles and lessons; control schools continued the usual curriculum without the environmental intervention. The primary outcome was the number of participants meeting the International Obesity Task Force overweight criteria after one school year. The posted counts were 385 of 1641 in the intervention group and 683 of 1309 in the control group.

The sponsor’s logistic-regression analysis reported an odds ratio of 0.69, with a 95% confidence interval from 0.48 to 0.98 and p < 0.05. The unadjusted recalculation reported a risk difference of -0.287, with a 95% confidence interval from -0.321 to -0.253 and p < 0.001. Important design and analysis details, including the allocation unit, missing-data procedures, analysis population, and model specification, are not reported in the registry record.