Cognitive Training With and Without tDCS to Improve Cognition in HIV: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Cognitive Training With and Without tDCS to Improve Cognition in HIV (NCT02647645), a randomized na study. The primary outcome is Working Memory: Participants' Rate of Improvement.

2. Clinical Question

The registry frames the study as a pilot investigation of cognitive training, either alone or paired with tDCS, for cognitive function in people with HIV-related mild neurocognitive disorder. The comparison therefore concerns the addition of active rather than sham stimulation to cognitive training. It is not a comparison of cognitive training against no training.

Working memory was measured using the Digits Backward trial of the Digit Span subtest of the Wechsler Adult Intelligence Scale Fourth Edition. Participants repeated digits in reverse order, and raw scores represented the largest number of digits a participant could remember and repeat. Higher scores were considered better. The primary endpoint is described as participants’ rate of improvement, with a 3-week timeframe.

The registry also lists the Center for Epidemiological Studies Depression Scale and the Patient’s Own Assessment of Function as secondary outcomes. Whether the posted primary comparison establishes a meaningful or durable difference cannot be determined from the p-value alone. The registry record does not report a clinical threshold for an important change in working-memory scores.

3. Trial Design

This was a randomized, parallel-group study with single masking. The registry lists an enrollment of 21 and marks the study as terminated. The recorded start was 2015-09, and primary completion was 2017-05. The registry does not report why the study was terminated or how termination affected recruitment, follow-up, or analysis.

The two groups both received cognitive training. One group also received active tDCS, while the comparator group received sham tDCS. This arrangement is intended to distinguish the effect associated with adding active stimulation from the effect of cognitive training and the experience of sham stimulation. The registry names tDCS as a device and cognitive training as a procedure.

The primary endpoint analysis is described as an overall estimated marginal means analysis using repeated-measures analysis of covariance. Treatment group was included as a fixed factor, with age, gender, race, education, helper T cell count, and log viral load as covariates. The registry says the cognitive measures were administered by an evaluator blind to treatment assignment. It does not specify the identity of the person or group masked under the single-masking designation, or describe how masking was maintained.

The arm-level primary results include 9 participants in each group. The primary outcome population is identified as all participants who completed treatment. The record does not explain how this population relates to the enrolled population, or provide a participant flow accounting for enrollment, assignment, completion, and analysis. Randomization methods, allocation concealment, treatment-session details, and the reason for termination are not reported in the registry record.

The registry lists 3 outcome measures and 3 analyses, alongside the primary working-memory endpoint and 2 named secondary outcomes. The record does not establish whether the analyses correspond one-to-one with the outcomes or whether other analyses were planned. The study is described as developing pilot data, so its results should be read as estimates from a small study rather than as a definitive evaluation.

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

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants who completed treatment.

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 Working Memory: Participants' Rate of Improvement, measured over 3 weeks.

The primary endpoint is participants’ rate of improvement on a verbal working-memory task over 3 weeks. Working memory was assessed with the Digits Backward trial of the Digit Span subtest of the Wechsler Adult Intelligence Scale Fourth Edition. Participants repeated a sequence of digits in reverse order; the raw score was the largest number of digits they could remember and repeat. The possible score range was zero to 10, and higher scores were considered better. The registry notes that normal persons typically remember from 5 to 7 digits, but this contextual description does not define a threshold for meaningful change in this study.

Working memory is one aspect of cognitive function, and a task-based score gives a standardized way to record performance at assessment. Repeated measurement can be used to estimate change over the stated timeframe. The registry describes the reported values as overall estimated marginal means from a repeated-measures analysis of covariance, rather than simply identifying them as raw endpoint scores.

The analysis population is listed as all participants who completed treatment. That matters because the endpoint does not necessarily summarize every participant enrolled or randomized. The registry does not report the assessment schedule, the exact definition of the rate-of-improvement measure, or a threshold for a change that would be considered clinically important. Those omissions limit interpretation of the values as evidence about the size or practical importance of change.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis is described as repeated-measures analysis of covariance, with treatment group as a fixed factor and age, gender, race, education, helper T cell count, and log viral load as covariates. The posted arm values are estimated marginal means from this model. The comparison is reported as an ANCOVA superiority analysis with p = 0.02. The registry does not report the full model specification, the test statistic, degrees of freedom, or a confidence interval for the treatment contrast.

What does covariate adjustment do?

In general, covariate adjustment incorporates selected participant characteristics into a model when estimating a group comparison. An adjusted mean is a model-based summary that accounts for the covariates included in the analysis; it is not necessarily the same as the unadjusted average of individual scores. Adjustment can improve precision when covariates are informative, but interpretation depends on the model being appropriate and the covariates being measured and handled as intended. The registry names the covariates but does not report their distributions, coding, missingness, or the rationale for their selection.

What does repeated-measures analysis address?

In general, repeated-measures methods analyze observations collected from the same participants over time. Measurements from one person are usually related rather than independent. A model can represent that dependence while estimating patterns across time and treatment groups. The registry identifies the method as repeated-measures analysis of covariance but does not report the assessment schedule, covariance structure, or the precise model used for time and treatment-by-time effects. The phrase rate of improvement is used for the endpoint, but the registry does not give the formula or model term that defines that rate.

How should the means and standard errors be read?

The reported means are 6.61 for cognitive training and 7.45 for cognitive training with tDCS, each with a standard error of 0.62. A standard error describes the uncertainty of an estimated mean under the analysis; it is not the spread of individual participant scores. A confidence interval gives a range of values compatible with the data and model at a stated confidence level. No confidence interval for the between-group contrast is reported, so the precision of that contrast cannot be read directly from the posted means and standard errors.

In general, uncertainty for a difference between adjusted means depends on the model and the covariance between the estimates. It cannot be recovered reliably by treating the two standard errors as if they fully described the contrast. The registry record does not provide the model-based contrast estimate or its standard error, and a recalculated two-arm continuous comparison is not estimable from the posted summary information.

What does the p-value convey?

In general, a p-value is calculated under a null hypothesis and model assumptions. It describes the compatibility of the observed data with that null hypothesis; smaller values can indicate greater incompatibility, but do not communicate effect size or clinical importance. The reported p = 0.02 belongs to the sponsor’s ANCOVA superiority analysis. The registry does not state the significance threshold used in the trial’s analysis plan, nor does it report whether the p-value was adjusted for multiple outcomes or analyses.

The label superiority identifies the type of reported comparison; it does not supply an effect estimate or confidence interval. A p-value alone cannot show how large the adjusted contrast was, how precise it was, or whether a change would be important to participants. Those questions require estimates and context not reported in the registry record.

Why can over-dispersion matter?

Over-dispersion is a general concern when observed variability is greater than a statistical model assumes. If a model understates variability, uncertainty estimates can be too small and tests can appear more decisive than warranted. This issue is most often discussed for count outcomes, whereas this primary endpoint is continuous. The registry does not report a dispersion assessment or other model diagnostics, so no conclusion about such diagnostics can be drawn for this analysis.

More generally, model checks help assess whether assumptions about the outcome, residual variation, and dependence among repeated measurements are reasonable. The registry does not report diagnostics, sensitivity analyses, or alternative model results. Their absence from the record does not establish that they were or were not performed.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Cognitive Training", "Cognitive Training With tDCS")
n   <- c(9, 9)
m   <- c(6.61, 7.45)
sp  <- c(0.62, 0.62)
sd  <- sp * sqrt(n)      # standard errors converted to SDs
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Cognitive Training  9  6.610  1.860  5.180  8.040
CARM  Cognitive Training With tDCS  9  7.450  1.860  6.020  8.880

11. Analyses Reported by the Sponsor

The registry reports the sponsor’s comparison as an ANCOVA superiority analysis with p = 0.02. The model is described as a repeated-measures analysis of covariance with treatment group as a fixed factor and age, gender, race, education, helper T cell count, and log viral load as covariates. Its posted adjusted means are 6.61 for cognitive training and 7.45 for cognitive training with tDCS, each with a standard error of 0.62.

The registry does not report the adjusted between-group contrast, its confidence interval, or a model-based standard error for the contrast. A recalculated unadjusted comparison is not estimable from the posted summary information, so there is no independent numerical contrast to set against the sponsor’s analysis. The group means and standard errors do not reproduce the repeated-measures, covariate-adjusted ANCOVA result.

The reported p-value should therefore be read as the sponsor’s stated model-based comparison, not as a recalculated result. The record does not report enough information to verify the model calculation independently or to assess the precision of the treatment contrast. It also does not provide a confidence interval or clinical-importance criterion that would help interpret the magnitude of the result.

12. Multiplicity, Missing Data and Analysis Populations

The registry names the primary working-memory outcome and lists the Center for Epidemiological Studies Depression Scale and the Patient’s Own Assessment of Function as secondary outcomes. It records 3 outcome measures and 3 analyses. Whether the reported ANCOVA p-value was adjusted for testing multiple outcomes, time points, or analyses is not reported in the registry record. The record also does not identify a prespecified multiplicity procedure or explain how the primary and secondary outcomes were prioritized in the analysis.

Multiplicity matters in general because conducting several statistical tests can increase the chance that at least one result appears unusual by chance. Adjustment methods can control that issue in different ways, but the appropriate interpretation depends on the planned testing family and analysis strategy. Without details of the prespecified analysis plan, it is not possible to determine whether multiplicity was addressed or how the p-value should be interpreted in relation to other outcomes.

The endpoint population is described as all participants who completed treatment. The primary arm summaries list 9 participants in each group, while enrollment is listed as 21. The registry does not provide a participant flow account, explain how many participants were randomized, or state why the analysis population differs from the overall enrollment count. It does not report how participants without completed treatment or missing measurements were handled.

Missing-data methods, imputation, sensitivity analyses, and the assumptions used for incomplete repeated measurements are not reported in the registry record. The record also does not say whether all randomized participants were included in the primary analysis or whether the analysis was restricted to treatment completers as described. These details are important because analysis-population choice and missingness assumptions can affect a model-based estimate. Their absence prevents a full assessment of how robust the reported comparison is to incomplete data.

13. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The safety figures list 0 serious adverse events among 11 participants at risk in the cognitive-training group and 0 among 10 participants at risk in the cognitive-training-with-tDCS group. Other adverse events are also listed as 0 in both groups, and deaths are listed as 0 in both groups. No top serious adverse-event terms are listed.

These are event counts within the recorded groups and risk populations. They should not be read as proof that adverse events cannot occur or that the interventions have equivalent safety. Counts from a small study provide limited information about uncommon events, and the registry does not report the safety observation period, event definitions, severity grading, or details of safety monitoring. The safety denominators also differ from the primary endpoint arm counts, and the registry does not explain the relationship between these populations.

The figures describe what was recorded in this trial; they do not establish a general safety profile. The reason for the terminated status is not reported, so it should not be inferred from the adverse-event entries.

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

The posted findings show higher estimated marginal working-memory means in the cognitive-training-with-tDCS arm than in the cognitive-training arm: 7.45 versus 6.61. Each mean has a standard error of 0.62. The registry reports an ANCOVA superiority comparison with p = 0.02. These results describe this study’s analysis of a verbal working-memory task over 3 weeks among participants who completed treatment.

They do not, by themselves, establish the magnitude or precision of the adjusted group difference. The registry does not report the contrast estimate or its confidence interval, and the posted summary values do not allow a reliable recalculation of the model-based comparison. The p-value is not an effect-size measure and does not show whether a change would be important to participants. The registry does not identify a threshold for clinical importance.

The comparison tests cognitive training with active tDCS against cognitive training with sham tDCS. It does not directly compare cognitive training with no intervention, nor does it isolate every possible contribution of treatment experience or expectations. The record does not report enough details about masking to assess how participants or other study personnel may have known or not known group assignment. It does state that the cognitive measures were administered by an evaluator blind to treatment assignment.

The study is described as developing pilot data and lists 21 participants enrolled, while the primary results include 9 participants per arm. The analysis population is treatment completers, and a full participant flow is not reported. The trial is marked terminated, with no reason given in the registry record. These features limit how broadly the result can be interpreted. No longer-term primary outcome result or evidence of persistence is reported in the record.

Accordingly, the statistical result is best understood as a model-based finding from a small, specific study, not as a general conclusion about cognitive training or tDCS. The record leaves important questions about precision, missing data, multiplicity, and clinical importance unanswered.

15. Why This Trial Matters Statistically

This trial raises several statistical issues that are common in small clinical studies. The design compares two groups that both received cognitive training, with active tDCS in one group and sham tDCS in the other. That makes the treatment contrast specific: it concerns adding active stimulation in the context of cognitive training, rather than comparing the full intervention package with no treatment.

The primary outcome is a continuous task score measured repeatedly, and the registry describes a covariate-adjusted repeated-measures analysis. Such models can use measurements over time and adjust for selected participant characteristics, but interpretation requires knowing what measurements were collected, how the model represented change, and how incomplete observations were handled. The registry names the method and covariates but does not report several details needed to evaluate those choices.

The result also illustrates the difference between a p-value and an effect estimate. The record reports p = 0.02, but does not provide the adjusted contrast or its confidence interval. The arm means indicate the observed direction, while their standard errors describe uncertainty around each mean rather than the comparative contrast. Without the model-based effect estimate and uncertainty, readers cannot assess the precision or practical size of the reported difference.

Finally, the primary analysis population is treatment completers, and the record does not explain participant flow from enrollment to analysis. A complete account of who was randomized, who completed treatment, and how missing outcomes were handled would help readers evaluate potential selection and missing-data concerns. These are especially important questions when interpreting pilot findings and a terminated study.

16. 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 provides limited information for independently evaluating the primary comparison. It reports adjusted arm means and standard errors, along with an ANCOVA p-value, but not the adjusted between-group contrast or its confidence interval. A recalculated continuous two-arm comparison is not estimable from the posted summary information. This means the comparative effect’s size and precision cannot be described from the available statistics.

The primary endpoint is called a rate of improvement, but the registry does not give the calculation used to define that rate or the schedule of repeated assessments. The model is identified as repeated-measures analysis of covariance, but its full specification, covariance structure, diagnostics, and sensitivity analyses are not reported. Covariate distributions and missingness are also not described.

The registry identifies the analysis population as participants who completed treatment. It lists 9 participants per arm for the primary outcome and an overall enrollment of 21, but does not provide a participant flow diagram or explain these counts. Missing-data handling, reasons for non-completion, and whether any analyses included all randomized participants are not reported in the registry record.

The record lists secondary outcomes and multiple analyses but does not state whether multiplicity was addressed. It does not report the significance threshold in the trial’s analysis plan or a clinical-importance threshold for working-memory change. The reason for the terminated status is also not reported. These omissions restrict interpretation of the result’s robustness, generalizability, and practical importance.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Cognitive Training With and Without tDCS to Improve Cognition in HIV — NCT02647645.
• Registry record last updated 2021-06-22; status last verified 2021-06.
• Results first posted 2018-10-15.
• 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.

20. Record Summary

NCT02647645 is a randomized, parallel-group, single-masked study listed as terminated. It enrolled 21 participants and compared cognitive training with sham tDCS against cognitive training with active tDCS. The primary outcome was participants’ rate of improvement on a verbal working-memory task over 3 weeks, analyzed among participants who completed treatment.

The posted adjusted means were 6.61 for cognitive training and 7.45 for cognitive training with tDCS, each with a standard error of 0.62. The registry reports an ANCOVA superiority analysis with p = 0.02, but does not report a confidence interval or adjusted contrast estimate. The registry also lists depression and function as secondary outcomes and records 0 serious adverse events in either group. Participant flow, missing-data handling, multiplicity procedures, and the reason for termination are not reported in the registry record.