Choice of the primary analysis in longitudinal clinical trials

Mallinckrodt, Craig; Watkin, John G.; Molenberghs, Geert; Carroll, Raymond J.

doi:10.1002/pst.124

Cited by 114 publications

(91 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is known to behave acceptably in the case of MAR drop-out, whereas the LOCF method patently does not. This conclusion is in agreement with that reached by other studies [15,16]. The ICH E9 Guidelines [17] say little about missing values apart from 'A trial may be regarded as valid, nonetheless, provided the methods of dealing with missing values are sensible.'…”

Section: Discussionsupporting

confidence: 89%

Handling drop‐out in longitudinal clinical trials: a comparison of the LOCF and MMRM approaches

Lane

2007

Pharmaceutical Statistics

229

185

View full text Add to dashboard Cite

This study compares two methods for handling missing data in longitudinal trials: one using the lastobservation-carried-forward (LOCF) method and one based on a multivariate or mixed model for repeated measurements (MMRM). Using data sets simulated to match six actual trials, I imposed several drop-out mechanisms, and compared the methods in terms of bias in the treatment difference and power of the treatment comparison. With equal drop-out in Active and Placebo arms, LOCF generally underestimated the treatment effect; but with unequal drop-out, bias could be much larger and in either direction. In contrast, bias with the MMRM method was much smaller; and whereas MMRM rarely caused a difference in power of greater than 20%, LOCF caused a difference in power of greater than 20% in nearly half the simulations. Use of the LOCF method is therefore likely to misrepresent the results of a trial seriously, and so is not a good choice for primary analysis. In contrast, the MMRM method is unlikely to result in serious misinterpretation, unless the drop-out mechanism is missing not at random (MNAR) and there is substantially unequal drop-out. Moreover, MMRM is clearly more reliable and better grounded statistically. Neither method is capable of dealing on its own with trials involving MNAR drop-out mechanisms, for which sensitivity analysis is needed using more complex methods.

show abstract

Section: Discussionsupporting

confidence: 89%

Handling drop‐out in longitudinal clinical trials: a comparison of the LOCF and MMRM approaches

Lane

2007

Pharmaceutical Statistics

229

185

View full text Add to dashboard Cite

show abstract

“…Changes in C-peptide AUC (DAUC) were calculated from baseline to month 12 and, for data from the DIA-AID 1 study, also from baseline to month 24 and from month 12 to month 24, using the mixedeffects model for repeated measurements (13,14). Relative treatment effect was defined as the ratio between the DAUC values of the DiaPep277-treated group and the placebo group and is expressed as a percentage.…”

Section: Data Analysesmentioning

confidence: 99%

Evaluation of Long-Term Treatment Effect in a Type 1 Diabetes Intervention Trial: Differences After Stimulation With Glucagon or a Mixed Meal

Pozzilli

Peled²,

Elias³

et al. 2014

Diabetes Care

View full text Add to dashboard Cite

OBJECTIVEEndogenous insulin secretion, measured by C-peptide area under the curve (AUC), can be tested using both the glucagon stimulation test (GST) and the mixed-meal tolerance test (MMTT). This study compares these two stimulation methods using long-term data from patients newly diagnosed with type 1 diabetes or with latent autoimmune diabetes. RESEARCH DESIGN AND METHODSA recently completed phase 3 intervention study with DiaPep277 demonstrated improved glycemic control and a significant treatment effect of glucagon-stimulated C-peptide secretion. Unexpectedly, MMTT failed to detect differences between the treated and control groups. Data from 343 patients in two balanced-randomized, double-blind, placebo-controlled, parallel-group trials of DiaPep277 were used to compare and correlate between GST-and MMTT-derived C-peptide AUC. Pearson's correlations were calculated for absolute C-peptide AUC at baseline and 12 and 24 months and for long-term changes in AUC (ΔAUC). RESULTSThe absolute AUC values obtained at any single time point by the two tests were well correlated in both data sets (r = 0.74-0.9). However, the correlations between the ΔAUC were much weaker (r = 0.39-0.58). GST-stimulated C-peptide secretion was stable over the fasting glucose range permitted for the test (4-11.1 mmol/L), but MMTT-stimulated C-peptide secretion decreased over the same range, implying differences in sensitivity to glucose. CONCLUSIONSMeasurement of long-term changes in stimulated C-peptide, reflecting endogenous insulin secretion, during the course of intervention trials may be affected by the method of stimulation, possibly reflecting different sensitivities to the physiological status of the tested subject.

show abstract

“…38,39 With the mixed-effect approach, information from the observed data is not explicitly imputed, and no additional data manipulation or analysis is required to accommodate the missing data. Two sensitivity analyses were performed with the mixed-effect model repeated-measure approach (MMRM) on different data sets (but without replacing missing data).…”

Section: Discussionmentioning

confidence: 99%