Adherence to intervention
As part of a sensitivity analysis, investigation CACE was undertaken using instrumental variable two-stage least squares regression. In this analysis, outcomes among those who ‘complied’ with the intervention were compared with outcomes among a group of ‘would-be compliers’ in the control arm. As described, treatment compliance was defined as (self-report of) taking 80–120% of study medication. As is evident in Table 17, the CACE analysis showed no differences with the conclusions drawn from the primary analysis.
| Analysis | n | Adjusted difference in means | 95% CI | p-value |
|---|---|---|---|---|
| Primary | 171 | –2.79 | –6.46 to 0.89 | 0.136 |
| CACE | 166 | –3.23 | –7.14 to 0.69 | 0.106 |
Missing data
A thorough investigation of the data showed that approximately 19% of the data or data sets were missing or incomplete, the majority of which related to the data collected from the various assessment tools used to collect some of the secondary outcomes. Reassuringly, there was no evidence that the extent of missingness differed by trial arm (p = 0.70), with 18% missingness in the placebo arm compared with 20% in the intervention arm.
A further exploration of the potential association between baseline characteristics and data missingness was also undertaken. There was no evidence of an association between baseline brain volume (p = 0.76), MMSE (p = 0.12), duration of diagnosis of AD (p = 0.13) and whether or not participants were taking dementia drugs (0.54) with respect to data missingness of the primary outcome. Similarly, there was no evidence of an association between the level of primary outcome data missingness and gender (p = 0.14), whether participants were hypertensive or normotensive (p = 0.63), or their number of years of prior education (p = 0.11). By contrast, we did find evidence of association between participant age and data missingness (p = 0.01) in that older participants were more likely to have missing data.
Multiple imputation using chained equations was used to impute missing data (Table 18). Variables included in the imputation model included all variables in the primary analysis model, baseline variables predictive of missingness and variables associated with the primary outcome. The findings from the analysis on imputed data are consistent with those from the primary analysis model: there was no evidence of a difference in brain volume at 12-month follow-up between the treatment arms.
| Analysis | n | Adjusted difference in means | 95% CI | p-value |
|---|---|---|---|---|
| Primary | 171 | –2.79 | –6.46 to 0.89 | 0.136 |
| MICE | 211 | –1.26 | –5.77 to 3.25 | 0.580 |
Post hoc exploratory analyses
In addition to the primary and secondary outcome analyses that have been described in the protocol, and given the overall lack of evidence in support of the role of losartan as a potential intervention for AD, we considered the potential value of three additional exploratory analyses. These were intended to make use of data gathered to inform the design of future studies of these types of BP-lowering medications in this participant group and in relation to the selection of other alternative outcome measures.
Did pre-existing hypertension affect primary and secondary outcomes?
We first sought to explore further what impact, if any, a prior diagnosis of hypertension may have on the primary outcome and secondary outcomes, given the long-standing recognition of association between mid-life hypertension and increased risk of AD.20,58 It was therefore considered that participants with pre-existing hypertension, for which they would be receiving treatment at the time of entry to the study, may plausibly have different associations with various outcomes not only at baseline, but also in how they responded over time to the intervention.
The full data are presented in Tables 19 and 20. Table 19 presents the data and regression analyses related to the presence or absence of hypertension on the various imaging data. Table 20 similarly presents the data and regression analyses related to the presence or absence of hypertension on the various cognitive, behavioural and functional outcomes. The regression analyses of these data according to the presence or absence of hypertension at baseline and 12-month follow-up, and according to intervention or placebo, showed that there was no association with any of the MRI-based data on brain volume or hippocampal volume. Similarly, but with the exception of DEMQOL-proxy, there were no significant differences for any of the cognitive (ADAS-Cog and MMSE), behavioural (NPI) or functional (BADLS, DEMQOL) outcome measures. The association with DEMQOL-proxy (interaction coefficient 6.31, 95% CI 0.42 to 12.20; p = 0.04) was unexpected and, given that DEMQOL-proxy is a companion-reported outcome, we cannot discern a reasonable explanation for this association and it most likely represents a type 1 false-positive error.
| Atrophy measure | Placebo, mean (SD) [IQR] | Intervention, mean (SD) [IQR] | Regression analysis | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | 12 months | Baseline | 12 months | Interaction coefficient | 95% CI | p-value | |||||
| Norm | Hyper | Norm | Hyper | Norm | Hyper | Norm | Hyper | ||||
| TBV (ml) | 1048 (109) | 1023 (115) | 1029 (112) | 1009 (114) | 1023 (106) | 1022 (106) | 1044 (97) | 998 (101) | –2.62 | –10.43 to 5.19 | 0.509 |
| LVV (ml) | 44 [34–62] | 50 [38–67] | 49 [38–66] | 54 [42–70] | 48 [38–59] | 50 [31–76] | 51 [40–67] | 55 [38–86] | 1.00 + | 0.97 to 1.04 | 0.795 |
| THV (ml) | 5.1 (1.0) | 5.0 (0.9) | 5.1 (1.0) | 4.7 (0.9) | 5.2 (0.9) | 5.2 (0.9) | 5.0 (0.9) | 5.1 (0.9) | –0.02 | –0.08 to 0.04 | 0.482 |
| LHV (ml) | 2.5 (0.5) | 2.4 (0.5) | 2.5 (0.5) | 2.3 (0.4) | 2.5 (0.5) | 2.5 (0.5) | 2.5 (0.5) | 2.5 (0.5) | –0.02 | –0.05 to 0.02 | 0.397 |
| RHV (ml) | 2.6 (0.5) | 2.5 (0.5) | 2.6 (0.5) | 2.4 (0.5) | 2.6 (0.5) | 2.7 (0.5) | 2.5 (0.5) | 2.6 (0.5) | –0.01 | –0.04 to 0.03 | 0.748 |
| TBV-BSI (ml) | – | – | 19.9 (10.8) | 18.3 (9.8) | – | – | 20.1 (10.9) | 19.8 (10.7) | 1.81 | –4.43 to 8.05 | 0.567 |
| LVV-BSI (ml) | – | – | 5 [3–7] | 5 [3–7] | – | – | 5 [2–8] | 5 [3–7] | 0.24 | –1.87 to 2.36 | 0.819 |
| LHV-BSI (ml) | – | – | 0.11 (0.07) | 0.12 (0.06) | – | – | 0.10 (0.08) | 0.12 (0.08) | 0.01 | –0.03 to 0.06 | 0.604 |
| RHV-BSI (ml) | – | – | 0.11 (0.06) | 0.12 (0.06) | – | – | 0.10 (0.07) | 0.11 (0.08) | < 0.0001 | –0.05 to 0.05 | 0.998 |
| Scheltens score, n (%) | |||||||||||
| 0 | 13 (23) | 7 (14) | 7 (13) | 2 (4) | 13 (22) | 6 (13) | 9 (18) | 5 (13) | 1.31 | 0.28 to 6.25 | 0.732 |
| 1 | 19 (34) | 23 (46) | 15 (29) | 13 (28) | 23 (40) | 20 (43) | 21 (41) | 15 (38) | |||
| 2 | 17 (30) | 15 (30) | 19 (37) | 24 (52) | 8 (14) | 12 (26) | 9 (18) | 12 (30) | |||
| 3 | 7 (13) | 5 (10) | 11 (21) | 7 (15) | 14 (24) | 9 (19) | 12 (24) | 7 (18) | |||
| 4 | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 1 (3) | |||
| WMH | 6917 [2791–13,778] | 12,551 [3640–21,071] | 8176 [3029–16,097] | 13,929 [5152–24,833] | 8802 [2805–22,589] | 12,387 [1981–25,837] | 11,036 [3023–23,763] | 13,903 [2126–28,034] | 0.96 + | 0.85 to 1.09 | 0.545 |
| Secondary outcome measure | Placebo, mean (SD) [IQR] | Intervention, mean (SD) [IQR] | Regression analysis | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | 12 months | Baseline | 12 months | Interaction coefficent | 95% CI | p-value | |||||
| Norm | Hyper | Norm | Hyper | Norm | Hyper | Norm | Hyper | ||||
| ADAS-Cog | 20 (7) | 18 (6) | 25 (12) | 23 (11) | 19 (8) | 21 (8) | 22 (12) | 24 (13) | –1.98 | –6.58 to 2.61 | 0.395 |
| MMSE | 22 (3) | 23 (4) | 19 (6) | 20 (5) | 22 (4) | 22 (4) | 19 (6) | 19 (6) | 0.58 | –1.65 to 2.81 | 0.609 |
| NPI | 8 [3–14] | 4 [2–15] | 8 [4–15] | 7 [3–21] | 8 [4–18] | 8 [3–18] | 9 [3–19] | 7 [2–16] | 0.86 + | 0.51 to 1.45 | 0.567 |
| NPI-proxy | 5 [2–10] | 5 [1–9] | 6 [2–9] | 6 [2–14] | 5 [2–10] | 6 [1–9] | 6 [2–10] | 4 [2–7] | 0.78 + | 0.47 to 1.27 | 0.307 |
| BADLS | 5 [3–9] | 5 [2–9] | 7 [3–15] | 7 [4–14] | 4 [1–11] | 9 [3–18] | 9 [3–16] | 12 [3–19] | 0.71 + | 0.48 to 1.04 | 0.075 |
| DEMQOL | 94 [85–103] | 96 [87–101] | 93 [86–100] | 94 [85–102] | 96 [89–102] | 93 [86–102] | 95 [87–105] | 99 [88–107] | 1.12 + | 0.96 to 1.30 | 0.146 |
| DEMQOL-proxy | 90 [82–98] | 95 [84–100] | 92 [82–98] | 95 [85–101] | 93 [86–100] | 90 [75–99] | 93 [83–99] | 94 [83–100] | 6.31 | 0.42 to 12.20 | 0.036 |
Exploring the characteristics of non-randomised participants of the open-label phase
In addition to providing all participants with some reassurance and the opportunity to try the intervention, we used an open-label phase to aid the selection of participants who were most likely to complete the main study. The effectiveness of this is supported by favourable study retention rates in the randomised phase. With this in mind, we wanted to better understand the characteristics of participants who entered the open-label phase but did not go on to the randomised phase, in which, as previously described, there was an excellent balance between the intervention and placebo arms in baseline participant demographics and outcome measures.
Table 21 details the reasons why open-label participants were not randomised. A number of important observations are evident.
| Recorded characteristic | Discontinuation | MRI failure | Total | |
|---|---|---|---|---|
| ≤ 7 days | 8–14 days | |||
| Number of participants | 27 | 15 | 8 | 50 |
| Hypertensive (i.e. taking BP medications) | 5 | 2 | 0 | 7 |
| Normotensive | 22 | 13 | 8 | 43 |
| Discontinuation in relation to intervention | ||||
| Considered unrelated | 3 | 0 | 8 | 11 |
| Considered related or possibly related | 24 | 15 | – | 39 |
| Recorded reasons for discontinuation | ||||
| BP | 17 | 9 | – | 26 |
| Bloods | 2 | 2 | – | 4 |
| Compliance | 1 | 0 | – | 1 |
| AEs | 5 | 3 | – | 8 |
| Other reasons | 2 | 1 | – | 3 |
| MRI | – | – | 8 | 8 |
| Protocol reason | 23 | 11 | 8 | 42 |
| Non-protocol reason | 4 | 4 | – | 8 |
First, over half of participants who withdrew from this phase of the study were in the first 7 days of treatment (at the lower dose of the active medication).
Second, and notable with respect to the relevance of hypertension in this group of participants, was that 86% of the discontinued participants were normotensive; two-thirds of those who withdrew within the first 7 days of treatment and all who withdrew for reasons related to MRI were normotensive.
Third, only 22% of the participants were thought to have discontinued for reasons that were unrelated to losartan. By contrast, the majority of the participants (78%) withdrew for reasons that were considered to be related to the intervention. Among these, there was less certainty that losartan was responsible for discontinuation of six participants, three of whom reported dizziness, confusion and fatigue; one participant reported nausea, one participant had a fall and one participant reported lethargy that continued during and after wash out.
Fourth, of the 50 participants who entered the open-label phase who were not randomised, just over 50% discontinued owing to BP changes that were deemed out of range or clinically significant, and another 20% (n = 8) were not randomised because of either unsuitable or incomplete baseline MRIs; a further 20% discontinued because of reported AEs. Of the eight participants who reported AEs, six are the same six participants whose reasons for discontinuation were less certain to be related to the open-label losartan. Of the remaining two participants, one developed a rash on their lower back and arm that coincided with a drop in their BP and the other developed bradycardia around the same time as their BP dropped. Two participants reported that they wished to discontinue because they viewed the trial to be too problematic (n = 1) or too time-consuming (n = 1), and the remaining participant reported that they felt that the drug was reducing their cognitive function.
Fifth, and finally, 82% of discontinuations were for protocol reasons, and only eight discontinuations were deemed non-protocol related.
Exploration of MRI-based measures of hippocampal atrophy as an alternative outcome measure
Although we used whole-brain atrophy as the primary outcome measure in this study, we were fortunate to be able to extract data on hippocampal atrophy from the relevant scan sequences. Given the potential utility of hippocampal atrophy as an outcome measure for dementia progression in its own right, and that it is measurable with scanning intervals of > 6 months, this study provided a welcome opportunity to explore it in the study population and also explore the relationship between other atrophy-related measures. Indeed, the original development of the visual Scheltens rating scale84 we used in this study as a minimisation variable to expedite randomisation was created because MRI-based methods for measuring hippocampal atrophy were not available at that time.
Table 22 presents a summary of additional measures of hippocampal and ventricular atrophy. We examined total hippocampal volume (THV), as well as left and right hippocampal volumes (LHV and RHV, respectively), at both baseline and 12-month follow-up. We also present change in volume using the BSI method92 applied to the individual hippocampi (LHV-BSI and RHV-BSI). We also explored data regarding lateral ventricle volume (LVV), which we examined at each time point, as well as change over time according to intervention arm (LVV-BSI). We also present these MRI-derived data alongside an exploratory examination of the subjective assessor-derived Scheltens ratings that were captured at baseline to inform randomisation and where these were also captured at 12-month follow-up.
| Placebo | Intervention | Regression analysis, adjusted mean difference (95% CI)a | p-value | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | 12 months | Baseline | 12 months | |||||||
| n | Mean (SD)/median [IQR] | n | Mean (SD)/median [IQR] | n | Mean (SD)/median [IQR] | n | Mean (SD)/median [IQR] | |||
| LVV (ml) | 106 | 47 [35–64] | 86 | 50 [40–67] | 105 | 48 [35–69] | 83 | 51 [40–78] | 0.99b (0.98 to 1.01) | 0.443 |
| LVV-BSI (ml) | – | – | 86 | 5 [3–7] | – | – | 83 | 5 [3–8] | –0.45 (–1.45 to 0.56) | 0.380 |
| LHV (ml) | 106 | 2.5 (0.5) | 87 | 2.4 (0.5) | 105 | 2.5 (0.5) | 84 | 2.5 (0.5) | 0.01 (–0.01 to 0.03) | 0.337 |
| LHV-BSI (ml) | – | – | 87 | 0.11 (0.06) | – | – | 84 | 0.11 (0.08) | –0.01 (–0.03 to 0.01) | 0.385 |
| RHV (ml) | 106 | 2.6 (0.5) | 87 | 2.5 (0.5) | 105 | 2.6 (0.5) | 84 | 2.6 (0.5) | 0.01 (–0.003 to 0.03) | 0.112 |
| RHV-BSI (ml) | – | – | 87 | 0.12 (0.06) | – | – | 84 | 0.11 (0.08) | –0.01 (–0.03 to 0.01) | 0.231 |
| THV (ml) | 106 | 5.0 (1.0) | 87 | 4.9 (0.9) | 105 | 5.2 (0.9) | 84 | 5.0 (0.9) | 0.02 (–0.01 to 0.05) | 0.144 |
| Scheltens score | ||||||||||
| 0 | 20 | 19% | 9 | 9% | 19 | 18% | 14 | 15% | 0.31c (0.14 to 0.68) | 0.004 |
| 1 | 42 | 40% | 28 | 29% | 43 | 41% | 36 | 40% | ||
| 2 | 32 | 30% | 43 | 44% | 20 | 19% | 21 | 23% | ||
| 3 | 12 | 11% | 18 | 18% | 23 | 22% | 19 | 21% | ||
| 4 | 0 | 0% | 0 | 0% | 0 | 0% | 1 | 1% | ||
From Table 22 it is evident that there were no differences according to trial arm for LVV, LHV or RHV at baseline and 12-month follow-up (p = 0.529, p = 0.360 and p = 0.099, respectively). Similarly, the mean change in each of these variables between 12-month follow-up and baseline was also non-significant (LVV-BSI, p = 0.641; LHV-BSI, p = 0.388; and RHV-BSI, p = 0.233). Similarly, there were no significant differences in THV (p = 0.143). These data, therefore, are entirely consistent with the absence of any significant differences recorded for the primary outcome of whole-brain atrophy.
Scheltens scores84 collected during the study are also presented in Table 22. There was a suggestion of a difference between arms whereby the odds of being in a higher (more severe) Scheltens score category at 12 months are lower for participants in the intervention arm than for those in the control arm (p = 0.004). Although this observation, based on the Scheltens rating scale, could be viewed as a possible treatment benefit, it is very important that this subjective rating scale is not overinterpreted when compared with the objective MRI-based measures that we found no effects from. Furthermore, the Scheltens scale rates atrophy of the MTA, which includes the hippocampus, and Scheltens scores not directly comparable with the MRI-based rating of the hippocampus.
Finally, given that both MRI-based measures and Scheltens ratings were focused on assessing levels of atrophy in participants, we thought that it may be helpful to present the correlations between the variables across the entire study population as there were no treatment-based differences. Table 23 summarises the levels of correlation between the MRI-based measures of atrophy and the observer-rated levels of atrophy by Scheltens scale. There was evidence of a strong correlation between THV and both LHV and RHV, whereas there was a weak negative correlation between LVVs and THVs, LHVs and RHVs. Of particular note was the observation that there was negligible correlation between subjective observer-based Scheltens scores and any of the MRI-based measures.
| Atrophy measure | LVV | THV | LHV | RHV | TBV-BSI | LVV-BSI | LHV-BSI | RHV-BSI | S_S |
|---|---|---|---|---|---|---|---|---|---|
| LVV (ml) | 1 | ||||||||
| THV (ml) | –0.40 | 1 | |||||||
| LHV (ml) | –0.36 | 0.87 | 1 | ||||||
| RHV (ml) | –0.31 | 0.86 | 0.50 | 1 | |||||
| TBV-BSI (ml) | 0.79 | –0.47 | –0.45 | –0.37 | 1 | ||||
| LVV-BSI (ml) | 0.98 | –0.40 | –0.35 | –0.32 | 0.80 | 1 | |||
| LHV-BSI (ml) | 0.42 | –0.65 | –0.71 | –0.42 | 0.51 | 0.42 | 1 | ||
| RHV-BSI (ml) | 0.46 | –0.64 | –0.45 | –0.67 | 0.46 | 0.46 | 0.62 | 1 | |
| S_S | 0.11 | 0.05 | 0.02 | 0.07 | 0.12 | 0.12 | –0.06 | –0.03 | 1 |