Rovner BW, Casten RJ, Hegel MT, Leiby B.
Preventing Cognitive Decline in Black Individuals With Mild Cognitive Impairment: A Randomized Clinical Trial.
JAMA Neurol. 2018 Dec 1;75(12):1487-1493.
PubMed.
For a randomized controlled trial, one expects that a variety of important risk factors or determinants of cognitive decline or its measurement (APOE, education, existing vascular and neurodegenerative pathology) would be distributed equally, if randomization worked and if they did not also influence dropout. The missing completely at random (MCAR) assumption is an unusually strong one for an observational study (Missing at Random: MAR is more commonly used) and it is not totally clear that the imputation technique applied to support the MCAR assumption actually answers concerns.
It appears that despite having another full neuropsychology battery (Uniform Data Set), the study chose to report a dichotomous six-word deficit on Hopkins Verbal Learning Test, and slope of Trails B to indicate the outcome; it would have been nice to see the others, at least for context and domain effects.
Single-blind studies always seem to raise some questions regarding how they might influence outcome in the “treatment” group due to demand characteristics. It is possible there could be some effect here, however, since the study period runs through 24 months and one might expect that such an effect would have washed out by that time.
All in all, this is an interesting study. I would be more convinced if they could verify that some of the other risk factors mentioned above were actually equally distributed by randomization and that dropout was not also leading to “reverse causation” inference.
So for me it raises more questions than reassurances.
There is a general consensus amongst researchers in aging that engagement in its various forms, exercise, cognitive, and social, are major factors in preserving health, including cognitive health and well-being in the elderly. However this concept has been difficult to prove in controlled trails and certainly for all racial and ethnic groups. In this elegant study of African-American volunteers using a sophisticated intervention “behavioral activation” that includes most components of engagement, Dr. Rovner and his colleagues demonstrated that this intervention significantly prevented cognitive decline in these participants.
The authors, however, are suitably cautious in their conclusions. The decline prevention was detected in cognitive testing and the clinical significance of the finding is still uncertain. A rather important point that they emphasized was that the effects of the intervention were time limited, lasting up to 2two years, suggesting that engagement process needs to be a lifetime process.
Despite these caveats this research study provides an important confirmation for the engagement hypothesis, especially in a population that is particularly vulnerable to dementia.
We are grateful for Dr. Hendrie’s kind words and reasonable caution. We agree that engagement in cognitive, physical, and/or social activities needs to be a lifetime process. We inculcated that message by having participants develop activity goals that they wanted to achieve, often emerging from lifelong or newly found interests. This conferred personal and cultural relevance and hopefully long-term clinical benefit. We are especially indebted to Dr. Hendrie and his research, which has provided the groundwork and rationale for this clinical trial.
We appreciate Dr. Kukull’s thoughtful comments. He identifies the importance of randomization to establish comparability of treatment groups in clinical trials. We have no evidence that the treatment groups in this RCT differed with respect to any important risk factors. Table 1 demonstrates this. In the modified intent to treat sample, there is also no evidence of any imbalances between treatment groups, despite more participants dropping out after baseline in the Behavioral Activation group. As to the concern about reverse causation, participants who dropped from the study had comparable cognitive performance at baseline to those who remained in the study. This fact mitigates against concerns about reverse causation.
The multiple imputation analysis assumes missing data are Missing at Random (MAR) and includes all randomized participants. The results do not drastically differ from results of analysis under the Missing Completely at Random assumption. Further, analyses of the continuous HVLT score, which assume MAR, all show significant benefit of Behavioral Activation vs. Supportive Therapy.
The paper includes a supplementary table that compares performance on multiple cognitive tests (representing various cognitive domains) by treatment group. We reported results that differed significantly (all comparisons favored Behavioral Activation over Supportive Therapy).
Clinical trials of behavioral interventions are unavoidably single-masked, as participants know the nature of the treatment they receive. In this clinical trial, both study treatments were credible interventions. This fact increases our confidence in the validity of the results.
Comments
University of Washington, National Alzheimer Coordinating Center
For a randomized controlled trial, one expects that a variety of important risk factors or determinants of cognitive decline or its measurement (APOE, education, existing vascular and neurodegenerative pathology) would be distributed equally, if randomization worked and if they did not also influence dropout. The missing completely at random (MCAR) assumption is an unusually strong one for an observational study (Missing at Random: MAR is more commonly used) and it is not totally clear that the imputation technique applied to support the MCAR assumption actually answers concerns.
It appears that despite having another full neuropsychology battery (Uniform Data Set), the study chose to report a dichotomous six-word deficit on Hopkins Verbal Learning Test, and slope of Trails B to indicate the outcome; it would have been nice to see the others, at least for context and domain effects.
Single-blind studies always seem to raise some questions regarding how they might influence outcome in the “treatment” group due to demand characteristics. It is possible there could be some effect here, however, since the study period runs through 24 months and one might expect that such an effect would have washed out by that time.
All in all, this is an interesting study. I would be more convinced if they could verify that some of the other risk factors mentioned above were actually equally distributed by randomization and that dropout was not also leading to “reverse causation” inference.
So for me it raises more questions than reassurances.
View all comments by Walter KukullThere is a general consensus amongst researchers in aging that engagement in its various forms, exercise, cognitive, and social, are major factors in preserving health, including cognitive health and well-being in the elderly. However this concept has been difficult to prove in controlled trails and certainly for all racial and ethnic groups. In this elegant study of African-American volunteers using a sophisticated intervention “behavioral activation” that includes most components of engagement, Dr. Rovner and his colleagues demonstrated that this intervention significantly prevented cognitive decline in these participants.
The authors, however, are suitably cautious in their conclusions. The decline prevention was detected in cognitive testing and the clinical significance of the finding is still uncertain. A rather important point that they emphasized was that the effects of the intervention were time limited, lasting up to 2two years, suggesting that engagement process needs to be a lifetime process.
Despite these caveats this research study provides an important confirmation for the engagement hypothesis, especially in a population that is particularly vulnerable to dementia.
View all comments by Hugh HendrieThomas Jefferson University
We are grateful for Dr. Hendrie’s kind words and reasonable caution. We agree that engagement in cognitive, physical, and/or social activities needs to be a lifetime process. We inculcated that message by having participants develop activity goals that they wanted to achieve, often emerging from lifelong or newly found interests. This conferred personal and cultural relevance and hopefully long-term clinical benefit. We are especially indebted to Dr. Hendrie and his research, which has provided the groundwork and rationale for this clinical trial.
View all comments by Barry RovnerThomas Jefferson University
We appreciate Dr. Kukull’s thoughtful comments. He identifies the importance of randomization to establish comparability of treatment groups in clinical trials. We have no evidence that the treatment groups in this RCT differed with respect to any important risk factors. Table 1 demonstrates this. In the modified intent to treat sample, there is also no evidence of any imbalances between treatment groups, despite more participants dropping out after baseline in the Behavioral Activation group. As to the concern about reverse causation, participants who dropped from the study had comparable cognitive performance at baseline to those who remained in the study. This fact mitigates against concerns about reverse causation.
The multiple imputation analysis assumes missing data are Missing at Random (MAR) and includes all randomized participants. The results do not drastically differ from results of analysis under the Missing Completely at Random assumption. Further, analyses of the continuous HVLT score, which assume MAR, all show significant benefit of Behavioral Activation vs. Supportive Therapy.
The paper includes a supplementary table that compares performance on multiple cognitive tests (representing various cognitive domains) by treatment group. We reported results that differed significantly (all comparisons favored Behavioral Activation over Supportive Therapy).
Clinical trials of behavioral interventions are unavoidably single-masked, as participants know the nature of the treatment they receive. In this clinical trial, both study treatments were credible interventions. This fact increases our confidence in the validity of the results.
View all comments by Barry RovnerMake a Comment
To make a comment you must login or register.