The Effect of Alternative Keyboards on Discomfort and Typing Kinematics: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for The Effect of Alternative Keyboards on Discomfort and Typing Kinematics (NCT00801983), a randomized na study. The primary outcome is Musculoskeletal Discomfort.

2. Clinical Question

The discomfort outcome was assessed using the Work Discomfort Survey. Participants rated discomfort and activity limitations on an 11-point numerical scale, with 0 defined as no discomfort or limitations and 10 as unbearable discomfort or major limitations. Because the recorded responses were severely skewed toward 0, the investigators dichotomized the outcome as discomfort yes or no. The primary endpoint therefore concerns the share classified as having discomfort, rather than changes in the original rating scale.

This framing matters to interpretation. A binary result can answer whether participants were classified as reporting discomfort, but it does not retain the full range of severity ratings. The registry’s reported outcome is not a direct measurement of a particular posture, a diagnosis of an upper-extremity disorder, or a measure of typing performance. The relationship between the keyboard condition and those other outcomes is not established by the posted primary endpoint.

3. Trial Design

The registry describes a randomized, unmasked, crossover trial with 85 participants enrolled. The study was marked completed. The record lists a start date of 2009-01 and a primary completion date of 2011-04. The phase is listed as not applicable. The sponsor was the University of Pittsburgh.

Participants were assigned to one of two keyboard orders. In one sequence, participants used a typical keyboard first and an alternative keyboard second; in the other, they used the alternative keyboard first and the typical keyboard second. The study description states that each keyboard was used for 6 months before switching to the other. It describes weekly reporting of musculoskeletal pain and kinematics assessments just before and just after each 6-month intervention at participants’ worksites. Details about completion of those assessments are not reported in the registry record.

The trial description says 75 computer users would be randomly assigned, whereas the registry lists 85 enrolled. The number enrolled is not the same as the number contributing to each posted outcome. The arm-level outcome summaries list 77 participants for each keyboard condition, and the recalculation uses 77 in each condition. The record does not explain this difference in participant counts, how many people completed both periods, or whether the same set of participants contributed data under each keyboard condition.

A crossover design allows participants to experience both conditions, so comparisons may be influenced by within-participant dependence, time period, sequence, or carryover. Those are general considerations for crossover studies, not documented features of the reported analysis. The record identifies the sponsor’s method as GEE but does not describe its specific model terms, correlation structure, or handling of period and sequence. The masking status is listed as none; details about outcome-assessor masking are not reported in the registry record.

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.

Period 1

Reasons for not completing this period:

Period 2

Reasons for not completing this period:

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.

Population: GEE analyses allows a subjects inclusion even if they did not complete the full trial. Although we had a total of 15 subjects not complete the full trial we had sufficient data on 77 subjects to complete the GEE statistical analyses. With the crossover design we compared 77 subjects using the standard keyboard to 77 subjects using the alternative

6. Primary Endpoint

The registry lists the primary outcome as Musculoskeletal Discomfort, measured over 6 months and 12 months.

The primary endpoint is musculoskeletal discomfort, reported as the percentage of subjects with musculoskeletal discomfort at 6 months and 12 months. The Discomfort Survey assessed symptoms and activity limitations in the neck and shoulder, back, and bilateral lower arms, including elbows, forearms, wrists, and hands. Participants also reported work schedule and medication used for pain control, but those items are not the stated primary endpoint.

Participants rated discomfort and limitations on an 11-point numerical rating scale. The record defines 0 as no discomfort or no limitations and 10 as unbearable discomfort or major limitations. The investigators report that responses were severely skewed toward no discomfort, so they dichotomized the outcome into discomfort yes or no. The posted endpoint therefore represents the classification of whether discomfort was reported, rather than the original rating’s severity or change in rating.

Dichotomization can make a skewed outcome easier to summarize as a proportion, but it discards distinctions among participants with different nonzero ratings. It also treats any response classified as discomfort as part of the same category. The threshold used to turn the rating into yes or no is not reported in the registry record, beyond the statement that the final outcome was discomfort yes or no. The endpoint does not by itself quantify posture, typing kinematics, or a clinical diagnosis of musculoskeletal disorder.

The registry lists the endpoint timeframe as 6 months and 12 months. How these time points were operationalized in the posted binary arm summaries is not reported in the registry record. That limits interpretation of whether the displayed counts represent a particular period, a combined summary, or another analysis choice.

7. Results Posted in the Registry

• Alternative Keyboard: 19.5% (15/77) versus 19.5% (15/77) on control; risk difference 0.0 percentage points (95% CI -12.5 to 12.5), risk ratio 1.00 (0.53–1.90); unadjusted two-proportion z-test p = 1.000.

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports a binary outcome: discomfort classified as yes or no. For a binary outcome, a proportion describes the share of participants with the event, and a risk difference compares the proportions between conditions by subtracting one from the other. A risk difference of 0.0 means the estimated proportions are the same in the calculation. The recalculated risk difference is 0.0, based on the posted summary statistics. This simple calculation does not adjust for the crossover structure.

Why does the crossover structure matter?

In a parallel-group trial, each participant is usually assigned to one condition. In a crossover trial, participants receive more than one condition in sequence. When the same participant contributes outcome data under both conditions, the two measurements may be correlated because they come from the same person. A statistical comparison that treats all observations as independent can fail to reflect that dependence. General methods for crossover data can account for repeated measurements and may consider period or sequence effects, but the exact model must be reported to know what was done.

The sponsor’s analysis is identified as GEE, or generalized estimating equations. In general, GEE is a method for estimating average effects when observations may be correlated, such as repeated measurements from the same participant. The registry does not state the specific working correlation structure, model terms, covariates, or variance estimator. It also does not state whether the analysis included period or sequence. Those details cannot be inferred from the method label alone.

What does a p-value convey?

A p-value describes how compatible the observed data are with a specified null hypothesis under the assumptions of the statistical model. It is not the probability that the null hypothesis is true, and it does not measure the clinical importance of a result. The sponsor reports p > .05 for its GEE analysis. The recalculated unadjusted p-value is 1.0. These values arise from different reported approaches; the sponsor’s own analysis takes precedence because the discrepancy flag is true. The registry does not provide enough information to reconstruct the sponsor’s p-value or explain the difference.

A p-value above .05 does not establish equivalence or prove that there is no effect. An equivalence conclusion requires a prespecified margin and an analysis designed to assess whether the uncertainty is contained within that margin. Such a margin and an equivalence analysis are not reported in the registry record. The p-value also does not show how precisely the effect has been estimated; the confidence interval is needed for that assessment.

What does dichotomizing the rating change?

The original survey used an 11-point scale, but the analysis converted responses to a binary discomfort outcome because of severe skew toward 0. In general, dichotomizing an outcome can simplify interpretation as a proportion with an event. It also loses information about the magnitude of nonzero ratings and can make results depend on the threshold used to define the categories. The threshold for the yes/no classification is not reported in the registry record. The posted binary summaries therefore cannot show whether the keyboard conditions differed in the full distribution or severity of discomfort ratings.

The record identifies the outcome unit as percentage of subjects with musculoskeletal discomfort and gives a value of 20.0 in each arm. It also lists 15 events in each condition for the recalculation. The relationship between the arm-level values and the event counts is not explained in detail, and the analysis timing within the crossover is not fully described. The recalculation is accordingly a descriptive, unadjusted comparison of the posted summaries, not a substitute for a participant-level crossover analysis.

How should planning calculations be understood?

In general, sample-size calculations estimate the number of participants needed to detect a chosen effect under specified assumptions about significance level, allocation, and statistical method. The planning assumptions described for the illustrations use a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect defined as a stated share of the observed effect. Such calculations are planning illustrations, not evidence about what the completed trial could detect after the fact. They are not observed power, and they do not establish the trial’s original design rationale. The registry does not report the trial’s own sample-size justification.

The stated illustration approach depends on the observed effect. When that estimate is uncertain, planning values derived from it can also be unstable. The registry does not report enough numerical detail about the planning illustrations to interpret particular participant targets. More broadly, any planning estimate is distinct from the observed enrollment of 85 and from the outcome summaries listing 77 participants in each keyboard condition.

Finally, no protocol-level detail is reported for the sponsor’s statistical model, handling of missing observations, or analysis population. Those omissions prevent a full assessment of whether the analysis matched the randomized crossover design and how sensitive the result might be to assumptions about incomplete follow-up.

10. Confidence Intervals and Precision

A confidence interval describes a range of effect values reasonably compatible with the data and the assumptions used to calculate it. It conveys uncertainty around an estimate; it is not a guarantee that the true effect lies within the interval, nor does it indicate that every value in the range is equally plausible. Its interpretation depends on the method used and on whether that method fits the study design.

The recalculated risk difference is 0.0, with a 95% confidence interval from -0.125 to 0.125. The interval includes 0.0, consistent with no difference in the point estimate, but it also includes differences in either direction. The interval therefore does not establish that the two keyboard conditions are equivalent. The registry does not report an interval for the sponsor’s GEE estimate, so the precision of the sponsor’s estimate cannot be assessed from the reported GEE results.

This interval is a simple unadjusted recalculation from the posted summary statistics. It does not necessarily account for the fact that participants could experience both keyboard conditions. If the observations are correlated within participants, an interval that treats them as independent may not represent the uncertainty from a crossover analysis. The exact sponsor model and its interval are not reported in the registry record, so the unadjusted interval should be treated as a separate descriptive calculation rather than as the sponsor’s primary inferential result.

The width of an interval is affected by the amount of information in the data and the variability of the outcome, as well as by the analysis method. The registry reports 77 participants per arm summary and 15 events in each condition for the recalculation. It does not report enough detail about outcome completeness, participant overlap between conditions, or the sponsor model to evaluate how those factors affect precision. The sponsor’s p-value above .05 cannot fill that gap: a p-value does not provide an effect range or describe the magnitude of uncertainty on its own.

11. Analyses Reported by the Sponsor

The sponsor reports a GEE analysis comparing the typical and alternative keyboard conditions, with p > .05. No sponsor effect estimate, confidence interval, or exact p-value is reported. The GEE result is therefore limited: it indicates that the sponsor did not report a statistically significant difference under that analysis, but it does not show the estimated direction or size of the difference or how precisely it was estimated.

A separate unadjusted recalculation from the posted summary statistics gives a risk difference of 0.0, a confidence interval from -0.125 to 0.125, and p = 1.0. The arm summaries list a value of 20.0 for both conditions, and the recalculation lists 15 events in each condition. These calculations do not reproduce the sponsor’s GEE model and should not be treated as the sponsor’s primary analysis.

The discrepancy flag is true, so the sponsor’s own analysis takes precedence. The two reported p-value statements both fall above .05, but the registry does not explain the discrepancy or provide enough detail to determine why the numerical results differ. It would be inappropriate to attribute the difference to a particular model choice, participant set, or data handling decision without that information. The missing sponsor estimate and interval also mean that the GEE result cannot be used to assess the size or precision of the treatment contrast.

12. Multiplicity, Missing Data and Analysis Populations

The registry reports one primary outcome measure and one analysis, and no secondary outcomes are posted. The study description nevertheless lists aims concerning discomfort, posture during keyboard use, and associations between posture and discomfort. The posted outcome record does not give results for those other aims. Whether additional outcomes were planned, measured, analyzed, or left unreported is not reported in the registry record.

Multiplicity arises when several outcomes, time points, subgroups, or analyses are tested. In general, performing many tests can increase the chance of obtaining a statistically significant result by chance, so trial protocols may specify how multiple comparisons will be handled. The registry does not report a multiplicity strategy, a prespecified hierarchy of outcomes, or whether any adjustment was applied. Because only one outcome analysis is posted, the available record does not permit an assessment of the broader set of analyses that may have been planned or conducted.

The outcome timeframe is listed as 6 months and 12 months, but it is not clear how the binary arm summaries relate to each time point or whether separate tests were conducted. The sponsor result is described as a GEE analysis without an effect estimate or model details. It is not reported whether time was modeled explicitly or whether the two time points were analyzed together. These omissions matter because repeated observations across conditions and time points can create dependence and potentially multiple opportunities for comparison.

Missing-data methods are not reported in the registry record. The record does not state how incomplete surveys, missed visits, or participants who did not complete both crossover periods were handled. The analysis population is also not reported: it is unclear whether the posted summaries include all randomized participants, only those with available outcome data, or another group. The registry lists 85 enrolled and 77 participants in each arm-level outcome summary, but does not explain the difference or the extent of participant overlap between conditions.

Without this information, it is not possible to assess whether missingness could have affected the estimated result or whether the analysis preserved the randomized comparison. The registry also does not report sensitivity analyses for missing outcomes or alternate outcome definitions. These are limitations of the available record, not evidence that any particular handling choice was used.

13. Safety Results

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

Other (non-serious) adverse events

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

The serious adverse event summaries list 85 participants at risk in each keyboard condition and 0 affected participants in each condition. The summaries also list 85 at risk for other adverse events and 0 affected participants in each condition. No serious adverse event terms are listed.

These figures describe what is recorded in the registry. In a crossover design, participants may contribute exposure to both keyboard conditions, so the two condition-level at-risk counts should not be read as necessarily representing separate groups of people. The record does not explain how adverse events were assigned to a condition or how the exposure periods were handled in these summaries.

Deaths are listed as null, so death counts are not reported in the registry record. The zero affected counts do not establish that no events occurred outside the registry’s reporting definitions or observation periods. The registry does not provide further detail on collection procedures, event severity definitions, follow-up completeness, or attribution. These summaries should therefore be read as registry-reported counts, not as a comprehensive account of all possible harms.

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

The posted results show equal arm-level values of 20.0 for the typical and alternative keyboard conditions. The unadjusted recalculation reports 15 events in each condition among 77 participants per arm summary, a risk difference of 0.0, and a confidence interval from -0.125 to 0.125. The sponsor’s reported GEE analysis gives p > .05, without an effect estimate or confidence interval. Because the discrepancy flag is true, the sponsor’s analysis takes precedence over the simple recalculation for the trial’s reported inference.

These results do not establish that the keyboard conditions are equivalent. A nonsignificant result is not proof that an effect is absent, and the unadjusted confidence interval includes differences in either direction. The registry does not provide an equivalence margin or an equivalence analysis. Nor does it report the sponsor’s estimated effect and its uncertainty, which would be needed to assess the size and precision of the primary GEE result.

The endpoint is a dichotomized discomfort measure, not the full range of ratings. The trial description also includes aims about posture and kinematics, but no secondary outcomes are posted. The binary discomfort summaries cannot establish whether typing posture changed, whether a particular posture was associated with less discomfort, or whether participants’ full symptom severity distributions differed between keyboard conditions.

Interpretation is further limited by the crossover design and missing analysis details. The registry does not specify how repeated observations, time period, sequence, or missing outcomes were handled in the sponsor’s model. Participant counts in the arm summaries are lower than the enrollment count, and the record does not explain the difference. These uncertainties do not invalidate the posted results, but they constrain what can be concluded from them. The result is best read as a reported absence of statistical significance in the sponsor’s GEE analysis, alongside an unadjusted recalculation with a null point estimate and appreciable uncertainty.

15. Why This Trial Matters Statistically

This trial illustrates why a registry result needs to be interpreted in the context of its design and outcome definition. The study used a randomized crossover arrangement, in which participants were assigned to keyboard orders and used both a typical and an alternative keyboard. That structure differs from a comparison in which separate groups use only one keyboard. Within-participant dependence can matter to both estimates and uncertainty, so an unadjusted comparison of arm summaries and a model-based crossover analysis do not necessarily answer the question in the same way.

The outcome also illustrates a common trade-off in clinical measurement. The investigators report that the rating data were severely skewed toward no discomfort and were dichotomized into yes or no. This makes the posted result interpretable as a proportion with discomfort, but it removes information about the intensity of reported discomfort. A reader cannot recover the original rating distribution from the binary summaries.

The difference between the sponsor’s GEE result and the unadjusted recalculation is a useful reminder that a reported p-value cannot be reconstructed from summary values without the details of the original model. The registry identifies GEE but does not report its estimate, interval, model terms, or treatment of repeated observations. Since the discrepancy flag is true, the sponsor’s analysis takes precedence, while the unadjusted recalculation remains a limited descriptive comparison.

Finally, the record shows why enrollment, outcome denominators, and analysis populations should not be treated as interchangeable. It lists 85 enrolled, while the outcome summaries list 77 participants per condition, and it does not explain the relationship. Careful statistical interpretation depends not only on the headline result but also on knowing which participants and time points contributed to it.

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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

Several details needed to evaluate this particular crossover trial are not reported in the registry record. The record does not explain why 85 participants were enrolled when the study description says 75 computer users would be assigned, or why the outcome summaries list 77 participants per condition. It does not state how many participants completed both keyboard periods, whether the same participants contributed to both condition summaries, or how missing observations were handled.

The timing and construction of the posted binary outcome are also unclear. The endpoint timeframe lists 6 months and 12 months, but the record does not explain how the arm-level values and event counts correspond to those time points. It states that discomfort ratings were dichotomized because of severe skew toward 0, but does not report the threshold used to define yes versus no. The original rating distributions are not posted, so the effect of dichotomization on the result cannot be assessed.

The analysis description is sparse. GEE is named, but the registry does not report the model terms, correlation structure, handling of period or sequence, missing-data approach, or analysis population. The sponsor’s estimate and confidence interval are absent. The registry also does not explain the discrepancy with the recalculated unadjusted result. As a result, the GEE analysis cannot be independently reconstructed from the record.

The record lists a planned enrollment of 75 in the study description and actual enrollment of 85, but a sample-size justification for the completed trial is not reported in the registry record. No multiplicity plan is reported, despite several aims in the study description, and no secondary outcomes are posted. Details about participant characteristics, eligibility criteria, adherence to each keyboard condition, and the completeness of the weekly reports are not reported. These limitations restrict interpretation of both the result’s precision and its applicability beyond the described computer-user population.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: The Effect of Alternative Keyboards on Discomfort and Typing Kinematics — NCT00801983.
• Registry record last updated 2014-08-22; status last verified 2014-08.
• Results first posted 2014-08-22.
• 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

The registry record describes a completed, randomized, unmasked crossover trial of typical and alternative keyboards, sponsored by the University of Pittsburgh. It lists 85 enrolled participants, with a start date of 2009-01 and primary completion date of 2011-04. The study description says participants were to use each keyboard for 6 months, in one of two randomized orders.

The primary outcome was musculoskeletal discomfort, assessed with a survey and dichotomized as yes or no because ratings were reported to be severely skewed toward no discomfort. The outcome timeframe is 6 months and 12 months. The arm summaries list a value of 20.0 for each keyboard condition and 77 participants in each summary. The unadjusted recalculation gives a risk difference of 0.0 and a confidence interval from -0.125 to 0.125. The sponsor reports a GEE analysis with p > .05 but no estimate or confidence interval; because the discrepancy flag is true, the sponsor’s own analysis takes precedence.

No secondary outcomes are posted. Serious and other adverse event summaries list 0 affected participants in each condition; deaths are not reported in the registry record. Details about missing data, analysis populations, and the sponsor model are not reported.