Author's response to reviews Title: The Older People, Omega-3, and Cognitive Health (EPOCH) trial design and methodology: A randomised, double-blind, controlled trial investigating the effect of long-chain omega-3 fatty acids on cognitive ageing and wellbeing in cognitively healthy older adults. Authors: Vanessa Danthiir (vanessa.danthiir@csiro.au) Nicholas R Burns (nicholas.burns@adelaide.edu.au) Ted Nettelbeck (ted.nettelbeck@adelaide.edu.au) Carlene Wilson (carlene.wilson@flinders.edu.au) Gary Wittert (gary.wittert@adelaide.edu.au) Version: 2 Date: 14 September 2011 Author's response to reviews: see over
14 th September, 2011 Dear Editors, Please find below our response to the reviewer s comments on our manuscript, The Older People, Omega-3, and Cognitive Health (EPOCH) trial design and methodology: A randomised, double-blind, controlled trial investigating the effect of long-chain omega-3 fatty acids on cognitive ageing and wellbeing in cognitively healthy older adults. We appreciate and thank the reviewers for their constructive comments. We have addressed each of these as detailed below. As requested, we have also now included a list of abbreviations used in the manuscript, along with their meanings, on pp.27-18. Sincerely, Vanessa Danthiir Responses to reviewers Reviewer: Natalie Sinn 1. It would be helpful to clarify up front that n-3 blood analysis was a measure of erythrocyte (red blood cell) PUFA status which is very important. We have now clearly stated that n-3 PUFAs were measured in erythrocyte membranes (p12). 2. Page 10: Participants had to be willing to not commence their own n-3 supplementation throughout the study and were excluded if they were taking n-3 supplements but no mention is made of fish consumption i.e. were they asked to limit their normal consumption of fatty fish to, for example, less than once a week or fortnight? As the authors note, the daily supplement dosage is equivalent to the n-3 PUFAs in 150g of salmon which has a similar impact on cell membrane composition. We did not ask participants to limit their consumption of fatty fish. We did ask participants to maintain their usual diet over the course of the study, and have now included a statement to this effect [page 11]. We collected detailed data relating to dietary n-3 PUFA intake at each of the four time points, allowing us to control for any effect of changes, or differences in dietary n-3 PUFA intake on outcomes [this has now been noted on page 15].
3. Page 10: The cut-off score of 24/30 on the MMSE for possible dementia is more commonly used as a cut-off for mild cognitive impairment, whereas 18/30 tends to be a cut-off for possible dementia. Therefore perhaps the methodology needs to be modified to state that people with dementia were excluded (i.e. that elderly people ranging from mild cognitive impairment to normal cognitive function were included in the study). The traditional MMSE cut-off score for identifying cognitive impairment and possible dementia (Lezak, 1984) was used, consistent with our exclusion criteria which specify as exclusions dementia [and] received a score less than less than 22/27 on a telephone-administered version [37] of the Minimental State Examination (MMSE) (equivalent to < 24/30 on the standard MMSE, the traditional cut-off for possible dementia). Although we have used a cut-off score that is used to screen out demented older adults, we agree that the suggested cut-off scores vary between publications and populations, with some publications suggesting that for highly educated adults, ie. University educated, the cut-off should be higher. However, less than 24 is often stated as the traditional cut-off for dementia using the MMSE, and for this reason that was the cut-off score utilised. We have clarified (p. 7): The primary aim of the current trial was to comprehensively assess over 18-months the efficacy of a DHA-rich fish-oil supplement on slowing cognitive decline in non-demented older adults. 4. Page 10: In the exclusion criteria the MMSE cut-off is stated as <24/30 whereas under Screening and information sessions it is given as <22/27 both cited as the telephone administered versions. This inconsistency needs to be clarified. Also, what was the purpose of administering the complete MMSE following written consent was it for further screening or an outcome measure? I think this needs to be clarified here We have now clarified the cut-off point on MMSE used to exclude participants; in both instances we say that participants scoring less that 22/27 on the telephone administered MMSE were excluded, and note that this score is equivalent to the traditional cut-off of less than 24/30 on the standard MMSE. p10: less than 22/27 on a telephone-administered version [37] of the Mini-mental State Examination (MMSE) (equivalent to < 24/30 on the standard MMSE, the traditional cut-off for possible dementia). We have clarified the purpose of administering the complete MMSE, on page 11: administration of the complete MMSE was undertaken, to corroborate the telephone screening and to serve as an outcome measure. 5. Page 12: would be useful to clarify that each placebo capsule contained 990mg of olive oil (if that is the case).
oil. We have clarified (p.13) that each placebo contained 990mg of olive 6. When asking participants which experimental group they thought they were in, to assess blinding adequacy, were they asked why they thought they were in a specific group? I.e. it should be considered that some beneficial effects of fish oil include reduced joint pain which might lead people to suspect that they are in the treatment group even if blinding is effective. Participants were not asked why they thought they were in a specific group, although in hindsight this would have been interesting information to collect. 7. Does the statistical analysis take multiple assessments into consideration? I note here that factor analysis was conducted to potentially reduce the number of assessments; will that also be used to reduce the number of variables in the final analysis? If so, this should be clarified. Modelling of latent cognitive variables will be employed in the final analyses to reduce the number of tests that will be conducted, following the confirmatory factor analyses of the baseline data, and providing the cognitive factors prove to be measuring the same construct across time. The assessment of intervention effects will be conducted on these cognitive factors (as opposed to individual tests). We have clarified this now on pp22-23: The cognitive domains will be modelled as latent variables, providing the cognitive factors prove to be measuring the same construct across time, following the initial planned modelling of the tasks and cognitive factors on baseline data using confirmatory factor analysis (CFA). The main effect of the intervention on each cognitive domain will initially be assessed separately... 8. Is it appropriate to give actual recruitment numbers and missing data in the methodology paper or would they be more appropriately stated in the paper reporting study outcomes? On consideration, we agree with the reviewer that details on actual recruitment numbers and missing data would be most appropriately stated in the paper reporting study outcomes, hence we have deleted mention of these. However, the one exception is the numbers for baseline since confirmatory factor analysis was conducted on this data and reported in the current MS; the N for the baseline analyses is included on p24. The following sentences have been deleted: p8. Recruitment commenced in June 2007 and data collection finished Nov 2009; 391 participants completed baseline assessment and 354 completed the final assessment. p.26 The trial has been completed with excellent retention rates and minimal missing data.
Reviewer Alan Dangour 1. Was a systematic review conducted to inform the section on crosssectional and prospective studies? If not, then statements such as "To date, only one published study has reported a negative relationship between cognition and biomarkers of n-3 LC PUFAs [22] or dietary intake levels" are not appropriate. How can you be sure? Please modify these background sections (which appear not to be based on systematic reviews) accordingly. A comprehensive but not systematic review was conducted to inform the background sections on cross-sectional and prospective studies, and we have therefore modified the section on cross-sectional and prospective studies accordingly, as suggested by the reviewer. The modifications are in bold in the instances cited below: p4. only a handful of the aforementioned studies that also examined relationships between cognitive outcomes and dietary intake levels of n-3 LC PUFAs have found significant positive relationships p5. The sentence cited by the reviewer has been changed to ; in contrast, Laurin et al [22] reported a negative relationship between cognition and biomarkers of n-3 LC PUFAs. 2. In the section on participant recruitment you state in one place that you exclude individuals with an MMSE score of <24 out of 30, while a few lines later you say you exclude those with an MMSE score of <22 out of 27. I understand that this is because in the latter case this is a telephone version of the test, but this is not particularly clear in the paper. This has now been clarified; please see our response to the same point by the other reviewer, at point 2 above. 3. You must provide more information on the composition of the "fish oil" added to the placebo capsule, provide a rationale for its addition and justify the inclusion of the "intervention" in the placebo arm. On p.13 we have now provided detail of the composition of the fish oil added to the placebo capsule [ equivalent to 1.8mg EPA and 1.2mg DHA ] On pages 11 and 13 we have added more detail [in bold below] to justify the addition of the small amount of fish-oil to the placebo capsules: p11. and 1% fish-oil was added to the placebo oil to help maintain blinding in the event of, for example, accidental piercing of the capsules or an aftertaste. p13. The minimal amount of fish oil was added to help mask any difference between the placebo and active capsules and was not processed under nitrogen, rendering it less potent than that used in the active formulation.
4. You have collected a huge amount of data at multiple time points and it's really important to be clear about what you plan to do with it. For example, you have collected a measures of "carotenoids, folate and B12" at baseline and final appointment. What are you going to do with these? Is there reason to believe that they will change because of the intervention? Are you planning to include them as co-variates? Are you just interested in cross-sectional associations of levels and outcome variables? At present you include the statment "Analyses may include these markers as covariates or mediating variables." This is not sufficient. More clarity is needed for all the variables you have collected. We have now included the following statement on page 7: The trial was set within the broader context of examining nutritional, health, and lifestyle factors associated with better cognitive and well-being outcomes in the elderly, both cross-sectionally and longitudinally. Detail relating to the variables collected, in particular the purpose of including each, has been included see pp.14-16. On page 14 we have included the following: At baseline and final assessment, blood was taken for measurement of: carotenoids, folate, and Vitamin B12, all having been associated with cognitive functioning in the elderly, to examine cross-sectional associations between these measures and outcomes, and if necessary in the intervention analyses, to control for any potential differences between groups at baseline or any changes in these measures that differ between groups (despite randomisation) 5. I would like to see a list of specific secondary outcome measures rather than broad categories such as "lipid metabolism" or "mood". Please specify each secondary outcome measure that will be report. I would also urge you to think carefully about your secondary outcomes - how many of these do you really believe might be influenced by the intervention? Do they all need to be listed as secondary outcomes? Also, surely change in n-3 LCP status is a measure of compliance rather than a secondary outcome measure? We have revised our secondary outcomes on p16. so that specifics, as opposed to broad categories, are given, as follows. We have removed broadly defined wellbeing and functional capacity, encompassing, and replaced mood with positive and negative affect, and added life satisfaction. We have replaced markers of lipid metabolism, inflammation, and oxidative stress with triglycerides, total cholesterol, HDL, LDL, homocysteine, CRP, MDA. For each measure that is listed there is reason to believe that the intervention might affect it. We have removed n-3 LCP status from the list of secondary outcomes, although we will also be investigating whether changes in n-3 LCP status are related to changes in the cognitive outcomes.
6. The number of tests conducted on each participant is large. It is important to state how long the appointments took and the order in which the tests were administered. Also, some statement on participant burden/fatigue and how you over-came it would be useful. We have now expanded upon the previous details of the duration of the testing session, on p17, to now read: The testing session lasted approximately four hours and consisted of four sections of around 50 min duration, with a break of approximately 10 min between each section. During this break participants were provided with standard refreshments, including decaffeinated tea/coffee. The speed tasks were interspersed between the accuracy based tasks and to also help minimise fatigue, paper and pencil tasks were interspersed with computerised tasks, which were administered in an adjacent room. The order of tasks was fixed across participants and testing sessions. We have now noted in the appendix the order of presentation of the tasks. The sequence number is listed after the name of each task. 7. I think it would be useful to provide more information on the rate of decline in cognitive function expected in the placebo group. As you are aware, we in the OPAL study found no change in cognitive function over 24 months of intervention. There are of course differences between trials, (you are conducting a trial in a population group with a wider age category (65-90), and are measuring cognitive function 4 times and conducting growth curve modelling), but your trial is shorter - and trial duration is likely to be critical. So some statement to support your assumption that there will be a measurable decline in cognitive function in the placebo group over 18 months seems appropriate. Normative data on rates of decline with ageing for cognitive tasks does not currently exist; hence we cannot state an expected rate of cognitive decline for the placebo group. However, we have amended (in bold) our statement on p8 regarding cognitive decline to give an indication of potential magnitude and support our statement that decline of the cognitive abilities vulnerable to ageing effects begins by the early 30s so although participants were screened for cognitive impairment, all were expected to experience cognitive decline to some extent. For instance, meta-analyses of cross-sectional data suggest that processing speed declines, on average, by approximately 20% at age 40 and by 40-60% at age 80. Further, under the section on the cognitive test battery, on p16, we have already noted that the focus of the battery was on constructs that are sensitive to age-related changes, and that Most of the cognitive domains were assessed using tasks in which speed of response was the main outcome measure. The psychometric properties of speed tasks make them more sensitive to subtle cognitive changes.
8. I propose that the statistical analysis section is checked by an expert. Two of the authors (VD and NB) are experienced in latent variable modelling and together have numerous publications utilising these types of analyses. Further, the statistical analysis section on the intervention analyses and power was incorporated in the successful proposal for funding for the trial to the Australian National Health and Medical Research Council, which was reviewed by a grant review panel (consisting of on average, ten members), and two expert external assessors. We believe one of these expert reviewers was knowledgeable and experienced in latent growth curve modelling. Further, the main reference for the analytical framework to be applied for the latent growth curve modelling of the intervention and for the power calculation has been provided as a citation on pp.22-23 (Muthén & Curran, 1997). Lezak, M. D. (1984). Neuropsychological assessment in behavioral toxicology--developing techniques and interpretative issues. Scand J Work Environ Health, 10 Suppl 1, 25-29. Muthén, B. O., & Curran, P. J. (1997). General longitudinal modeling of individual differences in experimental designs: A latent variable framework for analysis and power estimation. Psychological Methods, 2, 371-402. Newkirk, L. A., Kim, J. M., Thompson, J. M., Tinklenberg, J. R., Yesavage, J. A., & Taylor, J. L. (2004). Validation of a 26-point telephone version of the Mini-Mental State Examination. J Geriatr Psychiatry Neurol, 17(2), 81-87.