Using these concepts, we compared 1) the estimated efficacy of weekly and daily dose frequencies, 2) the predicted efficacy of weekly and daily supplementation regimens, and 3) the estimated trial effectiveness of weekly and daily supplementation regimens.
The comparison of efficacy of dose frequencies tested whether there was a difference in hematologic response per tablet ingested between a weekly dose frequency and a daily one. In the comparison of efficacy of supplementation regimens, the difference in hematologic response over a 12-wk period was predicted as if compliance had been 100%. The comparison of trial effectiveness of supplementation regimens showed the actual difference in hematologic effect over a 12-wk period, including the effect of differential compliance.
The questionnaires were coded and data was entered by using the SPSSWIN statistical package (version 7.5.1; SPSS Inc, Chicago). Data were verified by checking for consistency and range. For data analysis, the SPSSWIN (version 9.0) and STATA 7 (Stata Corporation, College Station, TX) statistical packages were used. The statistical methods included Student’s t test, analysis of variance (ANOVA), Bonferroni test for multiple comparisons, correlation analysis, three-factor repeated-measures ANOVA, multivariate regression analysis, and hierarchical ANOVA. Statistical significance was set at P < 0.05. Variables were tested for normal distribution. For nonnormal distributions, medians and percentiles were presented, and comparisons were made by using Mann-Whitney U test statistics. Lowess (20) smoothed plots were used to visualize dose-response relations.
The risk of confounding was fundamentally different between the analyses of efficacy and trial effectiveness. The efficacy analysis depended on differential compliance to produce a range of tablet intakes. The number of tablets taken during the 2 regimens was therefore not randomized but depended on individual behavior and was potentially associated with confounding factors. Careful analyses of potentially confounding factors were therefore essential in the efficacy analyses. In the trial effectiveness analyses, however, the interventions (weekly and daily regimens) were randomized, as were any potentially confounding factors. Control for confounding in this type of analysis is theoretically not necessary but was done nevertheless to increase plausibility (19). All analyses were tested for the confounding effects of maternal age, fundal height, initial hemoglobin concentration, parity, and SES. For a given analysis, potential confounders were identified as those variables with a P value < 0.20 for any linear or nonlinear association between hemoglobin outcomes and tablet ingestion (for efficacy analyses) or for any linear or nonlinear association of hemoglobin outcomes across the daily or weekly regimens (for effectiveness analyses). Potential confounding was accounted for by introducing potentially confounding factors into the crucial analyses as main effects and as pertinent interactions. Confounding by the measured variables was regarded as an unimportant contribution to the main effects if the effect of their linear or nonlinear introductions changed the variables describing the dose response or the interaction between daily and weekly regimens by
Using these concepts, we compared 1) the estimated efficacy of weekly and daily dose frequencies, 2) the predicted efficacy of weekly and daily supplementation regimens, and 3) the estimated trial effectiveness of weekly and daily supplementation regimens.
The comparison of efficacy of dose frequencies tested whether there was a difference in hematologic response per tablet ingested between a weekly dose frequency and a daily one. In the comparison of efficacy of supplementation regimens, the difference in hematologic response over a 12-wk period was predicted as if compliance had been 100%. The comparison of trial effectiveness of supplementation regimens showed the actual difference in hematologic effect over a 12-wk period, including the effect of differential compliance.
The questionnaires were coded and data was entered by using the SPSSWIN statistical package (version 7.5.1; SPSS Inc, Chicago). Data were verified by checking for consistency and range. For data analysis, the SPSSWIN (version 9.0) and STATA 7 (Stata Corporation, College Station, TX) statistical packages were used. The statistical methods included Student’s t test, analysis of variance (ANOVA), Bonferroni test for multiple comparisons, correlation analysis, three-factor repeated-measures ANOVA, multivariate regression analysis, and hierarchical ANOVA. Statistical significance was set at P < 0.05. Variables were tested for normal distribution. For nonnormal distributions, medians and percentiles were presented, and comparisons were made by using Mann-Whitney U test statistics. Lowess (20) smoothed plots were used to visualize dose-response relations.
The risk of confounding was fundamentally different between the analyses of efficacy and trial effectiveness. The efficacy analysis depended on differential compliance to produce a range of tablet intakes. The number of tablets taken during the 2 regimens was therefore not randomized but depended on individual behavior and was potentially associated with confounding factors. Careful analyses of potentially confounding factors were therefore essential in the efficacy analyses. In the trial effectiveness analyses, however, the interventions (weekly and daily regimens) were randomized, as were any potentially confounding factors. Control for confounding in this type of analysis is theoretically not necessary but was done nevertheless to increase plausibility (19). All analyses were tested for the confounding effects of maternal age, fundal height, initial hemoglobin concentration, parity, and SES. For a given analysis, potential confounders were identified as those variables with a P value < 0.20 for any linear or nonlinear association between hemoglobin outcomes and tablet ingestion (for efficacy analyses) or for any linear or nonlinear association of hemoglobin outcomes across the daily or weekly regimens (for effectiveness analyses). Potential confounding was accounted for by introducing potentially confounding factors into the crucial analyses as main effects and as pertinent interactions. Confounding by the measured variables was regarded as an unimportant contribution to the main effects if the effect of their linear or nonlinear introductions changed the variables describing the dose response or the interaction between daily and weekly regimens by < 10%.
การแปล กรุณารอสักครู่..
