Effectiveness of Arginine as a Treatment for Sickle Cell Anemia: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Effectiveness of Arginine as a Treatment for Sickle Cell Anemia (NCT00513617), a randomized phase 2 study. The primary outcome is Gardos Channel Activity.

2. Clinical Question

The trial asks whether arginine supplementation is associated with a change in Gardos channel activity compared with placebo in people with sickle cell anemia. The primary measure was taken 12 weeks after randomization. The registry record describes low-dose and high-dose arginine arms as well as a placebo arm, allowing each dose to be considered against placebo.

This is a question about a continuous physiological measure, not a count of pain episodes or another symptom outcome. The registry summary describes the broader study as evaluating blood cell function and disease symptoms, but the primary endpoint reported for this analysis is Gardos channel activity. The clinical question therefore needs to be interpreted in terms of this measure and its timing; it should not be expanded to conclusions about other outcomes that are not included in the primary result.

3. Trial Design

The registry describes a completed phase 2 randomized trial with a parallel-group model and quadruple masking. Randomization indicates that assignment to study groups was intended to occur by chance, while quadruple masking indicates that the registry records masking at that level. Further details about which roles were masked, how masking was maintained, or the allocation procedure are not reported in the registry record.

The listed arms were low-dose arginine, high-dose arginine, and placebo. The low-dose description is 0.05 g/kg/day arginine, the high-dose description is 0.10 g/kg/day arginine, and placebo is described as no arginine. The posted primary outcome summaries include 12 participants in the low-dose arm, 11 in the high-dose arm, and 14 in the placebo arm. These are the arm counts for the reported primary endpoint summaries; the registry record does not explain how they relate to the overall enrollment of 128.

The registry lists the trial start as 2004-06 and primary completion as 2007-09. It identifies UCSF Benioff Children's Hospital Oakland as the sponsor and lists anemia, sickle cell as the condition. The registry record marks the study completed. It does not report the original design assumptions, the planned allocation ratio, or a sample-size justification, so those aspects of the design cannot be reconstructed from the posted record.

The primary outcome was measured 12 weeks after randomization. The primary population is described as all randomized and dosed subjects, labeled intent to treat, with no imputation used. The record also lists four secondary outcomes: soluble vascular cell adhesion molecule, 8-iso-PGF2a, endothelin-1, and fetal hemoglobin. It records seven outcome measures and six analyses, but does not provide enough information to infer the full analysis plan from those counts alone.

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.

Before assignment: All subjects were to be without hydroxyurea, transfusion, and arginine for 90 days prior to enrollment. Prior to randomization, blood was drawn for baseline efficacy and safety measurements.

Overall Study

Analysis population for the primary outcome, as defined in the registry: All subjects that were randomized and dosed - Intent to Treat (ITT) No imputation used.

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 Gardos Channel Activity, measured over 12 weeks after randomization.

The primary endpoint is Gardos channel activity, a continuous measure reported in mmol/10^13 cells x min. The registry describes the Gardos channel as a calcium (Ca2+)-activated K+ channel. The arm summaries are means with standard deviations, so the central result describes average activity within each group, while the standard deviation describes variation among the observations contributing to each group summary.

The trial uses this measure as its primary physiological endpoint at 12 weeks after randomization. In general, a continuous physiological measure can capture variation in a biological process across participants and can be compared between randomized groups using differences in their means. Here, the endpoint addresses the study’s stated interest in blood cell function; it is not itself a direct measure of pain episodes, symptoms, or longer-term clinical outcomes. The registry record does not report a minimum clinically important difference or explain how a change of a given size in this measure should be interpreted clinically.

When reading the results, the direction of the mean difference alone does not establish whether a change is beneficial or harmful. That requires a justified interpretation of the measure’s scale and clinical context, neither of which is specified beyond the endpoint definition and unit. The reported time point also limits what can be said: the primary result concerns activity measured 12 weeks after randomization, rather than effects at other times.

7. Results Posted in the Registry

• Low Dose: mean -0.03 versus 0.11 on control; difference -0.14 (95% CI -0.34 to 0.06), standardized effect d = -0.54; z-test p = 0.161.
• High Dose: mean 0.00 versus 0.11 on control; difference -0.10 (95% CI -0.34 to 0.13), standardized effect d = -0.35; z-test p = 0.383.

8. Statistical Interpretation

9. Statistical Methods Explained

The result is a comparison of means for a continuous endpoint. The posted arm summaries give a mean and standard deviation for each group, while the recalculations compare each arginine-dose group with placebo. A mean difference expresses the separation between the group averages on the endpoint’s measurement scale. A negative value means the estimated mean in the arginine group is lower than the placebo mean under the subtraction order used for the comparison. The sign alone does not say whether that direction is clinically favorable.

What does the standard deviation describe?

The standard deviation describes how spread out individual observations are around an arm’s mean. It is not the same as the uncertainty in the mean or the uncertainty in the difference between two means. In general, uncertainty in a mean depends on both the spread of observations and the number of observations contributing to the estimate. Uncertainty in a difference also reflects information from both groups. The posted arm summaries include standard deviations, but the endpoint summary counts are small, so estimates can remain imprecise even when a point estimate is available.

The recalculated comparisons are explicitly unadjusted calculations from posted summary statistics. They are not a substitute for the sponsor’s analysis. The recalculated p-values are 0.161 for low dose versus placebo and 0.383 for high dose versus placebo. The sponsor reports mixed models, with p-values of 0.080 and 0.133 for the corresponding comparisons. Because the methods and reported estimate definitions differ, the p-values should be read alongside their method labels rather than treated as duplicate outputs from one identical procedure.

What does a confidence interval contribute?

A confidence interval gives a range of estimates compatible with the data under a statistical procedure and its assumptions. In general, a narrower interval indicates greater precision, while a wider interval indicates less precision. For these recalculated comparisons, both intervals include zero and extend in both directions from it. This means the posted interval ranges do not isolate a difference in one direction. The registry record does not state the confidence level for these intervals, so it is not appropriate to assign a level that is not recorded.

A confidence interval is not a guarantee that the true effect lies within the displayed limits, and it is not a range of likely outcomes for individual participants. It is also not a measure of clinical importance. Clinical interpretation would require contextual information about the endpoint scale and what size of change matters; the record does not report a threshold for that purpose. The sponsor analyses do not include confidence limits in the reported results.

How should the p-values be read?

In general, a p-value describes how unusual data at least as incompatible with a null hypothesis as the observed data would be, assuming the null model and analysis assumptions hold. It is not the probability that the null hypothesis is true, nor the probability that a result occurred by chance. A p-value also does not state the size of an effect. The unadjusted recalculated values and the sponsor’s mixed-model values are therefore best read as features of their respective analyses, together with the estimated differences and available uncertainty intervals.

The registry record does not report the sponsor’s full model specification, covariates, variance assumptions, or other details needed to independently reproduce the mixed-model results. It also does not report whether a multiplicity procedure was used. Those missing details limit how precisely the sponsor analysis can be evaluated from the registry record alone.

What is meant by an unadjusted comparison?

For the recalculations, unadjusted means that the comparisons are based on the posted summary statistics rather than on a reported covariate-adjusted or mixed-model analysis. The recalculated estimates use the posted group means and pooled variability summarized in the comparison calculations. This approach can be useful for a straightforward description of the observed group separation, but it does not reproduce a model that uses participant-level data or any additional structure. The record does not report participant-level data, so the recalculation cannot assess model choices beyond the summary-statistic calculation.

In general, adjustment can change an estimate or its uncertainty when a model accounts for other variables or features of the data. That general possibility does not explain the differences between these particular results: the registry record does not provide enough information to identify why the estimates and p-values differ. The recalculated and sponsor results should therefore remain clearly labeled rather than blended into one result.

What does the analysis population mean?

The registry identifies the primary population as all subjects randomized and dosed, calls it intent to treat, and states that no imputation was used. This wording describes the population and missing-data approach as recorded; it does not establish that every randomized participant contributed an endpoint observation. The outcome summary counts are 12 for low dose, 11 for high dose, and 14 for placebo. Further detail about participant flow, reasons for unavailable measurements, or the handling of missing values is not reported in the registry record.

In general, imputation is a method for replacing or estimating missing values for analysis. Saying that no imputation was used does not, on its own, explain whether observations were missing or how the analysis handled them. The record does not report enough detail to determine whether the available-case summaries, the sponsor model, and the stated population label refer to exactly the same analyzed observations.

Finally, the trial is randomized, parallel, and quadruple masked according to the registry. These design labels describe important features of how the comparison was organized, but do not supply missing details about implementation. The allocation method, concealment details, mask-specific roles, analysis plan, and original sample-size rationale are not reported in the registry record. Statistical interpretation should stay within the results and methods that are actually recorded.

10. Confidence Intervals and Precision

Precision concerns how tightly the data locate an estimated difference. The recalculated mean differences are -0.142 for low dose versus placebo and -0.103 for high dose versus placebo. Their posted interval limits are -0.340 to 0.057 and -0.335 to 0.129, respectively. Both ranges include zero, so neither interval confines the estimate to a difference in one direction.

The ranges are broad enough to include estimates below and above zero. This indicates uncertainty in both the magnitude and direction of the unadjusted differences. The intervals should not be read as establishing equivalence or as proving that the groups are the same: a range that includes zero can also include nonzero differences. The registry record does not give a clinical threshold for judging whether any value within these ranges would matter in practice.

The interval level is not stated for the recalculated limits. The sponsor’s mixed-model results report point estimates and p-values but no confidence limits. Therefore, the posted information offers a direct interval-based view of precision for the recalculations, but not for the sponsor estimates. In general, confidence intervals communicate estimation uncertainty more directly than a significance label, while still depending on the method and assumptions used.

These estimates are based on the arm counts attached to the endpoint summaries: 12 for low dose, 11 for high dose, and 14 for placebo. The registry enrollment is 128, and the relationship between enrollment and the endpoint counts is not reported in the registry record. The limited endpoint counts and the broad interval ranges are relevant when judging how precisely the mean differences are estimated; they do not establish why those counts differ from enrollment.

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 participants per arm for 80% power and participants per arm for 90% power. They are not total event counts: the record’s planning assumptions specify per-arm participants for non-time-to-event outcomes, and Gardos channel activity is a continuous outcome. In plain terms, the 80% column gives the planned number per arm under the stated assumptions to have an 80% chance of detecting the chosen effect if that effect and the assumptions hold. The 90% column describes the corresponding planning target for a 90% chance.

The planning illustration uses a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect: 100%, 75%, or 50%. For each comparison, the figures are planning scenarios based on that comparison’s observed effect, not the original trial’s design. The original study’s sample-size assumptions and justification are not reported in the registry record.

Smaller effects require more participants to detect under the same planning assumptions. That is why planning for 75% or 50% of the observed effect calls for larger per-arm sample sizes than planning for the full observed effect. These figures are illustrations for future planning, not evidence about how well this trial was powered after it was completed. They are not observed power.

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, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -0.142; pooled standard deviation σ = 0.261.
2. n = 2 × (1.96 + 0.842)² × 0.261² / -0.142² = 53.3, rounded up to 54 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
delta <- -0.1418                      # observed difference in means
sigma <- 0.261255767878912                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^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 (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(513617)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 54 (80% power), 72 (90% power)
 75% of observed effect: n per arm = 95 (80% power), 127 (90% power)
 50% of observed effect: n per arm = 214 (80% power), 286 (90% power)
power.t.test: 54.3 
Simulated power at n = 54 per arm: 0.803 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-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 sponsor reports mixed-model analyses of mean differences in final values. For low dose versus placebo, the reported mean difference is -0.125 with p = 0.080. For high dose versus placebo, it is -0.098 with p = 0.133. Confidence limits are not reported for these sponsor analyses.

The recalculated comparisons are simple unadjusted calculations from the posted summary statistics. They give mean differences of -0.142 for low dose and -0.103 for high dose, with p-values of 0.161 and 0.383, respectively. The recalculated estimates and p-values are not identical to the sponsor’s reported results. The registry record does not provide enough model detail to explain the differences, so the two sets should be presented as results from different analyses rather than treated as interchangeable.

The sponsor analysis is the trial’s reported analysis; the recalculation is a supplementary summary-statistic comparison. The record does not report the full mixed-model specification or whether the model included additional terms. It also does not report an adjustment for multiple comparisons. Those omissions constrain interpretation but do not justify attributing the result differences to a particular model feature.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists seven outcome measures, four secondary outcomes, and six analyses. The four named secondary outcomes are soluble vascular cell adhesion molecule, 8-iso-PGF2a, endothelin-1, and fetal hemoglobin. The primary outcome is Gardos channel activity. These counts and names indicate that the record includes more than one outcome or analysis, but do not by themselves establish how many statistical hypotheses were tested or how they were organized.

Whether the sponsor adjusted for multiple comparisons is not reported in the registry record. No multiplicity procedure, testing hierarchy, or division between confirmatory and exploratory analyses is described in the available record. Without that information, the reported p-values should not be assumed to account for the number of outcomes or comparisons. In general, testing multiple hypotheses can increase the chance of at least one apparently unusual result if each test is considered separately; adjustment methods are designed to address that issue in specified ways. This general point does not establish that any particular adjustment was required or used in this trial.

The primary population is stated as all subjects randomized and dosed, described as intent to treat, with no imputation used. The primary endpoint summaries list 12 low-dose, 11 high-dose, and 14 placebo participants. The registry record does not explain the relationship between these endpoint counts and enrollment of 128, identify reasons for missing measurements, or give a detailed account of participant flow. It does not report additional missing-data methods beyond stating that no imputation was used.

In general, an analysis-population label and a missing-data method answer different questions. A population label describes who is intended to contribute to an analysis; an imputation statement describes whether missing values are filled in using a statistical procedure. The registry wording should be retained as reported, without inferring that all randomized participants had an observed primary endpoint. Details about sensitivity analyses, alternative analysis populations, and the treatment of missing observations are not reported in the registry record.

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

The results describe mean Gardos channel activity in the low-dose, high-dose, and placebo groups at 12 weeks after randomization. In the recalculated comparisons, each arginine group has a lower estimated mean than placebo, but both confidence intervals include zero and the unadjusted p-values are 0.161 and 0.383. The sponsor’s mixed-model analyses also report negative mean differences, with p-values of 0.080 and 0.133. These are statistical descriptions of the posted endpoint results, not conclusions about whether arginine improves health.

The endpoint is a physiological measure of activity in a calcium-activated K+ channel. The record does not provide a threshold for a clinically meaningful change or explain how the observed differences relate to symptoms. Consequently, the results do not establish the clinical importance of the estimates. They also do not establish that there is no difference: the recalculated intervals include both zero and values in either direction, and the sponsor analyses do not report confidence limits.

The registry summary describes a study intended to evaluate blood cell function and disease symptoms, but the primary result addressed is specifically Gardos channel activity. The secondary outcomes listed in the record are separate measures; findings for those outcomes are not part of these primary comparisons. No conclusion about pain episodes, other symptoms, or the secondary outcomes should be inferred from the primary endpoint alone.

The trial’s randomized, parallel, quadruple-masked design is relevant context for interpreting group comparisons, but the registry record does not give detailed information about implementation, the full statistical plan, missing data, or multiplicity. The overall enrollment is 128, while the reported endpoint summaries contain the smaller arm-specific counts noted above; the relationship is not explained. These limitations affect how far the result can be generalized beyond the reported comparisons.

In short, the posted results estimate differences in a continuous physiological endpoint and show substantial uncertainty around the unadjusted comparisons. They neither establish clinical benefit nor rule out potentially meaningful differences. Clinical interpretation would need contextual evidence about the measure and its relevance, together with fuller reporting of the analysis and endpoint data.

16. Why This Trial Matters Statistically

This trial is statistically informative because it places a continuous physiological endpoint at the center of a randomized comparison with two active dose groups and a placebo group. That design creates a clear set of dose-versus-placebo questions, while the result illustrates why an estimated mean difference should be read together with its uncertainty rather than as a stand-alone number.

The two recalculated comparisons point in the same direction, but their intervals extend across zero and the endpoint counts are limited. This makes precision a central part of interpretation. The sponsor’s mixed-model results are also reported, and their estimates and p-values differ from the summary-statistic recalculations. Seeing both emphasizes the importance of keeping analysis methods distinct and asking what information is needed to reproduce a reported model.

The trial also highlights reporting questions that matter in clinical research: how endpoint counts relate to enrollment, what missing-data procedures were used, whether multiple comparisons were addressed, and what clinical change in a physiological measure would be meaningful. The registry record leaves these matters unreported, so the statistical account can describe the observed comparisons but cannot fill in the missing design or analysis details.

Finally, the planning illustrations show how the required sample size changes when the target effect is smaller, under stated planning assumptions. They should be read as prospective planning scenarios, not as a retrospective verdict on the completed trial. Together, these features make the study a useful example of distinguishing point estimates, uncertainty, sponsor analyses, and planning calculations.

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.

The registry record leaves important details unavailable for evaluating the primary result. It does not report the original sample-size justification, planned allocation ratio, allocation procedure, or a detailed statistical analysis plan. The study is identified as randomized, parallel, and quadruple masked, but the record does not specify how randomization and masking were implemented or which roles were masked.

The relationship between enrollment of 128 and the primary endpoint summary counts of 12, 11, and 14 is not reported in the registry record. Participant flow, reasons for unavailable endpoint measurements, and the handling of missing observations beyond the statement that no imputation was used are also not reported. The primary population is labeled intent to treat but described as randomized and dosed subjects; the record does not explain the wording or its operational application.

The sponsor reports mixed-model results but does not provide confidence limits or enough model detail to reproduce the analyses from the registry record. The recalculated intervals do not have a stated confidence level in the record. The record also does not report whether adjustment for multiple comparisons was used, or provide a clinical threshold for interpreting changes in Gardos channel activity.

The primary endpoint is measured at 12 weeks after randomization. The record does not provide results at other time points in the primary result summary, so the findings should not be extended to longer-term outcomes. These limitations do not invalidate the posted statistics, but they restrict what can be concluded about precision, missing data, analysis choices, and clinical relevance.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Effectiveness of Arginine as a Treatment for Sickle Cell Anemia — NCT00513617.
• Registry record last updated 2017-03-29; status last verified 2009-06.
• Results first posted 2009-08-04.
• 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

This completed phase 2 trial is recorded as randomized, quadruple masked, and parallel. The registry lists enrollment of 128 and identifies UCSF Benioff Children's Hospital Oakland as sponsor. The primary endpoint is mean Gardos channel activity, measured 12 weeks after randomization and reported in mmol/10^13 cells x min.

The three arm descriptions are low-dose arginine at 0.05 g/kg/day, high-dose arginine at 0.10 g/kg/day, and placebo described as no arginine. The primary outcome population is described as all randomized and dosed subjects, labeled intent to treat, with no imputation used. The posted endpoint summaries include 12 participants for low dose, 11 for high dose, and 14 for placebo.

The sponsor reports mixed-model comparisons against placebo. Summary-statistic recalculations provide unadjusted comparisons and intervals that include zero. The record lists four secondary outcomes and does not report the original sample-size justification, full model details, multiplicity handling, or the relationship between enrollment and the endpoint counts.