Author's reply

There is a lack of research into the effect of pharmacological interventions for dementia in people with Down syndrome and Alzheimer's disease. In this this paper we used routinely collected clinical data to explore the effect of cholinesterase inhibitors and/or memantine on survival and function in this group. These therapies are recommended by the National Institute for Health and Clinical Excellence for dementia treatment and the guideline includes people with Down syndrome.¹ Although subject to limitations given the observational rather than randomised design (discussed in more detail in the paper) our results support the use of antidementia drugs for people with Down syndrome who develop Alzheimer's disease.

We welcome Professor Buckley's interest in our work, and she is right to highlight the complexities of medication decision-making. We would expect that individual treatment decisions consider the best available evidence, personal circumstances and comorbidities, and incorporate the views and preferences of people with intellectual disability and their carers. Indeed, it is the aim of our analysis to expand the evidence base to enable informed decision-making. We will address Professor Buckley's concerns in turn.

Professor Buckley questioned our use of Kaplan–Meier estimates, a standard approach to survival analysis, and the figures based on these. This type of survival analysis enables use of data from all individuals to be included in the analysis, by censoring participants either at the date of death or at the date of their last clinic assessment so that information contained in survival times is taken into account accurately. It would be inaccurate to report average ages of death as these cannot be compared meaningfully between both groups because not all of those in the cohort died and those in the treated group were less likely to die; this would not be captured by reporting the age of death of only those known to have died.

Factors other than medication prescription that might influence survival were accounted for using Cox regression (Table 2). Variables added to the final analysis were site, age at dementia diagnosis, gender and degree of intellectual disability. Professor Buckley points out that we did not account for stage of dementia as a potential confounder. We agree that this would have been desirable, but there is no standard method of recording this that is in regular use in clinical services. She suggested that the DLD (a screening tool for dementia in individuals with intellectual disabilities) could be used for this purpose, but the scores obtained from the DLD reflect both degree of intellectual disability and dementia-related decline, with higher scores indicating lower levels of functioning, whether because of intellectual disability or dementia or a combination of both conditions.² For that reason the authors of the tool have recommended it be used sequentially to identify decline over time from an individual baseline. It cannot be used as a cross-sectional staging tool because a high score could indicate a long-standing level of intellectual ability rather than dementia and a low score might not exclude dementia in those with mild intellectual impairment. Unfortunately clinician ratings of mild, moderate or severe dementia were incompletely recorded in clinical notes (Table 1) and these data were not recorded beyond baseline, thus could not be included in our analyses.

However, in order to examine change in cognitive and functional ability over time from a baseline, our analysis of DLD scores was conducted using coefficients (i.e. based upon the mean difference between the scores of those on medication and those not on medication) that did take account of baseline DLD scores (Table 3). As requested by Professor Buckley, we now report raw DLD data at baseline, and follow-up visits for all individuals for whom this is available. Mean baseline DLD cognitive score in the untreated group was 30.54 (95% CI 26.49–34.60) and in the treated group 25.35 (95% CI 23.29–27.41); at first follow-up assessment DLD cognitive scores were 27.80 (95% CI 23.24–32.35) in the untreated group and 22.34 (95% CI 20.16–24.52) in the treated group; at second follow-up 31.62 (95% CI 26.17–27.08) (untreated) and 23.90 (95% CI 21.85–25.88) (treated); and at third follow-up 34.86 (95% CI 27.49–42.23) (untreated) and 26.20 (95% CI 23.90–28.51) (treated). These unadjusted data highlight the difference between the group means in cognitive score at baseline and other time points and appear to demonstrate a generally slower rate of cognitive decline in people prescribed medication, with DLD cognitive scores of those not treated worsening by approximately 14% (increase in scores of 4.32 on average, from a baseline of 30.54) by the third follow-up visit, compared with a worsening in DLD cognitive score of only 3% on average (increase in scores of 0.85 on average) in those prescribed medication. We have included the third time point here, which we did not include in the analysis in the paper, although as indicated by the width of the confidence interval, the number of observations at this time point is small, particularly in the untreated group. The numbers included at each time point are slightly different from those reported in the paper because of missing data in certain individuals precluding adjustment by baseline value. While we agree that research abstracts are limited by word counts, we believe our reporting is balanced and fair and call for more research in this field, including clinical trials of medication where the limitations of observational designs could be overcome.