The RADAR trial was designed to be the first RCT to formally test the angiotensin hypothesis in AD.20 Central to this hypothesis is the role of AngII, a potent vasoconstrictor known for decades to be implicated in hypertension, acting through the AngII type 1 receptor.110 AngII levels are in excess in AD,35 as a direct result of excess levels of ACE131,32 or reductions in the activity of ACE235 that have been found in post-mortem brain tissue and compared with brain tissue from non-demented individuals. The relevance of these findings to the pathogenesis of AD is that AngII is a multifunctional peptide with known functions in the synthesis of the inflammatory mediator TNFα,36,37 reduction of acetylcholine release,38,39 promotion of glutamatergic effects,40 and promotion of oxidative stress and mitochondrial dysfunction,111 all of which are major sequelae of AD pathology. Furthermore, AngII has been shown in rodents to promote the production of Aβ and alter the levels of phosphorylation of tau protein, analogous to AD-related changes that give rise to neurofibrillary tangles.41,42 This has been supported by our own findings of reduced ACE2 activity in post-mortem AD brain tissue that coincided with elevated levels of AngII, was highly correlated with Aβ and tau pathology and was inversely related to ACE1 activity.35,43
These largely preclinical and ex vivo data have gone some way towards providing a possible mechanistic explanation for findings, spanning several decades, suggesting that hypertension in mid-life7,8 and later life9 and stroke10 all increase risk of dementia. In other words, the association between hypertension and dementia could be related to excess RAS activity in mid-life that contributes to hypertension, whereas the occurrence of dementia later in life derives from what would have been the coincident chronic exposure of the brain to elevated levels of AngII. Alternatively, it may relate to the fact that, although a number of hypertension treatments lower BP effectively, they do not all necessarily lower levels of AngII. Moreover, the more recent finding that cerebrovascular pathology and dysfunction are likely to be the earliest pathological event in the development of AD18 also reinforces this possibility. Similarly, several studies both prior to and since the commencement of the RADAR trial have reported that AngII-targeting drugs [AngII type 1 receptor antagonists (AT1RAs) and ACE1-Is] are associated with a lower incidence of AD than other types of antihypertensive drugs and AT1RAs tend to be more beneficial than ACE1-Is in slowing the rates of progression, hospitalisation and mortality,11–14 although not all studies have found such effects.19 Similarly, these RAS-acting drugs have also been found to be associated with lower rates of conversion of MCI to AD, as well as less AD-related tau neuropathology in both post-mortem brain tissue and CSF from living AD participants recruited to the ADNI.15–17
In summary, AngII signalling is a plausible mechanism that can connect the traditional pathological hallmarks of AD with mechanisms involved in BP regulation and cerebrovascular dysfunction, which have been widely reported but poorly understood for decades. As a result, there is now a strong rationale for the use of AngII-targeting drugs as an intervention in AD. The RADAR trial therefore tested the use of the AT1RA losartan, the prototype drug in the AT1RA class, for which there are considerable data on its use as an effective antihypertensive drug for a wide range of ages. Losartan has also been reported to cross the BBB,44 a characteristic that is important in potential treatments for AD. It has also been reported to improve CBF,22 a surrogate marker of cognitive performance in humans.45–47 Losartan has also been reported to reduce neuronal damage following ischaemia in rat stroke models,23 to inhibit the production of Aβ and AD-related modifications of tau protein in aged rodents41,42 and, in low doses (i.e. having no effect on BP), to reduce AD-related pathology and cognitive decline in transgenic mouse models of AD.48 It has also been suggested that, given its role in reducing hypertension, losartan may provide a benefit by reducing ischaemia-mediated WMH.24
General comments on RADAR design and study population
The RADAR trial recruited to 93% of its intended target. However, the use of an open-label phase, preferential participant retention rates and the ability to secure a high proportion of primary outcomes for all completed participants meant that the power of the study to detect our target difference in brain atrophy at 12 months (assumed therapeutic benefit) of at least 3.8 ml/year with a two-sided α = 0.05 was reduced only slightly, by 2% (from 84% to 82%).
Our decision to investigate a 12-month period of treatment and follow-up was based on previous studies conducted by ADNI in which MRI has revealed measurable brain atrophy in this period among AD patients, equivalent to 15.2 ml/year (SD 8.6 ml/year). We had proposed that a relative difference in atrophy between groups of 25%, equivalent to an absolute difference in an annual atrophy rate of 3.8 ml in TBV, would be a clinically meaningful outcome that was measurable. However, it is now worth acknowledging that, since the design and commencement of the RADAR trial in 2012, follow-up periods of 18–24 months have been recommended for investigations where disease modification is being explored and this, therefore, serves as a potential limitation of this study.
Examination of the baseline characteristics of the RADAR study shows very clearly that our randomisation process, which included minimisation variables of age and Scheltens score to attempt good balance in trial arms related to baseline atrophy, was very successful. There were no significant differences in the baseline characteristics of study population in terms of any of the traditional demographic variables or in the baseline primary and secondary outcome measures when examined cross-sectionally.
It was also evident that the level of data missingness was similar in both arms following review at the end of the study. On closer examination it was clear that there were no predictors of the factors that may contribute, such as baseline brain volume, MMSE, duration of diagnosis of AD or whether or not participants were taking dementia drugs (0.539). There were also no differences in data missingness based on the gender or number of years of education of participants, and whether or not they had hypertension did not have any effect. The only variable that was found to be associated with data missingness was participant age (p = 0.01), in that data were more likely to be missing in older participants, arguably not unexpectedly, and the majority of these missing data were related to the assessment-based secondary outcome measures. There may also have been a slight contribution to data missingness from MRI-based outcome measures, because the duration of scans may have been too short, when shortened at the participant’s request, to secure the primary outcome that was taken in the initial MRI sequences. Furthermore, more recruitment sites than anticipated were unable to support the collection of the secondary MRI-based outcome measures related to WMH volume and CBF. However, unless these sites recruited participants who on average were slightly older than those recruited by sites that could support the retrieval of these outcomes, this could have made a small contribution to the levels of data missingness associated with older participants. Our recorded levels of 88% data completeness are comparable to those of another study, NILVAD,112 that recently reported no significant therapeutic benefit of the calcium receptor antagonist nilvadipine in patients with AD. This study also included a single-centre substudy that examined MRI outcome measures similar to those used in the RADAR trial; that study reported 89% data completeness, compared with 88% in the RADAR trial, and, as was found in the RADAR trial, the distribution of missing data was random.113
It was reassuring to see that there were no significant differences between the intervention and placebo arms of the study in terms of either SAEs or AEs, highlighting the safety of the RADAR trial. This is further supported by comparable retention rates throughout the main study in both treatment arms. These favourable rates are believed to be largely the result of the carefully designed open-label phase; following stakeholder feedback and suggestions, this was undertaken before the randomisation phase. This showed that, in the absence of such an open-label phase, approximately 20% of participants may have entered the RADAR trial and undergone baseline MRI scans only to withdraw within weeks of commencing medication. The open-label phase, therefore, contributed significantly to improving the overall welfare of participants and is likely to have reduced unnecessary participant and study partner burden for a proportion of participants, as well as helping those participants who proceeded to the main study to feel less anxious and more reassured about the possible implications of taking part.
Finally, on a related point regarding safety, only one unblinding request was made during the entire trial, for a participant attending hospital with a chest infection. Few other questions were raised about the need for or possibility of emergency unblinding during the study. This is most likely because the intervention, losartan, is a repurposed drug that is already used, as with a number of related drugs, in clinical practice to treat hypertension and some other cardiac conditions. Thus, in the case of any event necessitating emergency unblinding, it would be reasonable to assume that the patient had been taking losartan and to use existing knowledge about the drug to manage treatment accordingly. Thus, an obvious advantage of using repurposed drugs rather than experimental interventions in a trial is that important safety data can be collected in a relevant patient population, while simultaneously determining the drug’s therapeutic efficacy.
Primary outcome
The primary outcome of the RADAR trial was the measurement, using vMRI, of change in brain atrophy between baseline and the 12-month follow-up after treatment with losartan. The RADAR trial60 was originally designed to provide 84% power, requiring a sample size of 228 participants. The final power was 82% because the number of participants recruited was slightly smaller than originally hoped, but the effects of which were mitigated somewhat by more favourable retention rates, possibly because of the effectiveness of the open-label phase in ensuring the identification of participants who were most likely to tolerate the intervention. The primary outcome, brain atrophy after 12 months’ treatment post randomisation, was chosen on the assumption that an absolute difference in TBV between groups of at least 3.8 ml/year would constitute a clinically meaningful difference. TBV measured using vMRI (T1-MPRAGE) is now widely validated as a surrogate marker of cognitive decline and AD pathology.75,76,85–88
The analysis showed that there was some progression of brain atrophy in participants with AD in both treatment arms in the RADAR trial. However, there were no significant differences between treatment arms when TBV was analysed using either the 12-month volume cross-sectionally adjusted for baseline or the more sensitive BSI method, showing that losartan afforded no therapeutic benefit following 12 months of treatment. Given the variety of mechanisms reported to be associated with AngII, a number of which are regularly described as significant in the pathogenesis of AD, one may have expected that interference with a target that is a significant contributor to all of these mechanisms may have had a more appreciable effect. These results are therefore surprising and somewhat disappointing and provide yet another example of promising preclinical research and supportive population-based studies failing to translate into RCT evidence of effectiveness.114 Although loss of original study power was negligble, it remains possible that our powering of the study at 82–84% was on the low side; however, we are reasonably confident that this is unlikely, given the consistently similarly null findings from the secondary analyses.
The RADAR trial is a seminal study from the perspective of formally testing the AT1RA losartan in a formally diagnosed group of people with AD as well as the first use, in a non-commercial study, of vMRI to objectively determine the primary outcome measure. To the best of our knowledge, no other studies with which to compare these results have been conducted to date. Yet these data will provide a sobering benchmark for other small-scale trials that are ongoing. A similarly designed and sized Phase II trial in hypertensive AD patients, SARTAN-AD, which is being undertaken in Canada, commenced after the RADAR trial. It was designed to compare perindopril with telmisartan (https://clinicaltrials.gov/ct2/show/NCT02085265). However, unlike the RADAR trial, this study has failed to recruit even a fraction of the number of participants required because of high rates of prior exposure to drugs related to the intervention, a reliance on the recruitment of only participants with hypertension, and also issues with identifying sufficient numbers of recruitment centres to aid recruitment (Professor Sandra Black, SARTAN-AD chief investigator, University of Toronto, personal communication, 2019). The SARTAN-AD trial is the only study that is close enough to the RADAR trial to allow the results to be compared, particularly given the intention to use vMRI to measure the primary outcome. However, beyond the ongoing concerns of whether or not the study will be successfully completed, its original design will have asked a different research question from the RADAR trial. It was designed to compare the effects of ACE1-Is (perindopril) and AT1RAs (telmisartan) on brain atrophy; however, the restriction of recruitment to people with pre-existing hypertension (unlike the RADAR trial) will limit the generalisability of the SARTAN-AD study. Furthermore, the participants recruited to the SARTAN-AD study include those with prior exposure to related medications and the design of the trial is such that participants could be randomised onto a related drug (ACE1-I or AT1RA). This means that participants in the SARTAN-AD study could have had a prior exposure to the intervention, whereas the RADAR trial was more conservative and recruited only individuals with no previous exposure to the intervention.
The continuation and success of the SARTAN-AD study has important consequences, not only as it will be an additional study to help test the angiotensin hypothesis in AD, but also because it will serve as an important basis of comparison for the RADAR trial. Furthermore, among the other ongoing studies that are exploring the role of other RAS-acting drugs in relation to cognitive impairment, and will be discussed in further detail later, none currently includes investigations of people with AD, thus reinforcing the uniqueness of the RADAR trial.
Only one other study lends itself to some consideration as a comparator with the RADAR trial findings. The RCT NILVAD112 tested the efficacy of the BP-lowering drug nilvadipine in mild-to-moderate AD. Because nilvadipine is a calcium channel blocker and not an AT1RA, the comparison of the studies is imperfect because the drugs work by different mechanisms; however, as both drugs work to lower BP, comparisons in that regard may be worthy of consideration. NILVAD found no any evidence that nilvadipine confers therapeutic value in mild-to-moderate AD. However, NILVAD also included a small substudy, undertaken at a single centre, that included MRI measures of brain (and hippocampal) volume, WMH volume and CBF using similar but not identical approaches to those used in the RADAR trial. It is notable that these data were generated after a follow-up of just 6 months post randomisation and from a group of 44 participants allocated equally to the intervention and placebo arms.113 Despite the participant group being approximately 20% of that presented in the RADAR trial, the data from NILVAD showed the absence of any therapeutic benefit from BP-lowering with respect to whole-brain volume. In agreement with the RADAR trial findings, this substudy found no evidence to support any benefit from BP-lowering treatment with nilvadipine in relation to either brain volume or the volume of WMH.113 By contrast, the authors did report that nilvadipine increased hippocampal blood flow, although they did not identify differences in global CBF or CBF in the posterior cingulate cortex, an area they chose to investigate regional differences alongside the hippocampus.113
Secondary outcomes
A number of secondary outcomes were investigated in the RADAR trial.
We first investigated rates of AD progression, as assessed by changes in various cognitive and behavioural assessments, including the ADAS-Cog,96 the NPI97 and the MMSE.98 We also explored participant function and quality of life using the BADLS100 as well as examining change in quality of life (DEMQOL, DEMQOL-proxy99). These data were collected at the same time points as the primary outcomes, with an additional collection between these points to explore the short-term benefits.
A cursory examination of all of the assessment scores at each of the three time points (baseline, and 6 and 12 months) showed a worsening of all scores among participants in both the intervention and the placebo arms. However, a comparison of change in scores from baseline to follow-up, similar to the exposure period to the intervention for the primary outcome, showed that there were no significant differences in the secondary outcomes between the intervention arm and the placebo arm that would suggest a therapeutic benefit.
The study also sought to investigate the change in the volume of WMH, which was measured in our MRI protocol using T2-FLAIR, given that it may predict 1-year cognitive decline,24 as well as to explore CBF, using ASL MRI methods in these participants. For ease of discussion and because of their relatedness these will be discussed together. These imaging outcomes and the specific measurement of BP as an additional outcome measure were important, as BP is a key outcome in assessing the efficacy of hypertension treatments and thus may have provided insight into the associations between BP and the secondary MRI measures of cerebrovascular damage and dysfunction.
The levels of data collection for these outcome measures were lower than anticipated across all of the recruitment centres. Indeed, it was known from the outset of the trial that data on WMH and CBF, even if collected at the same level as the primary outcome, would be underpowered to provide any definitive findings and, therefore, these variables were proposed as secondary outcomes.
Data on WMH volume could be collected from only approximately half of the RADAR participants (i.e. from 15 of the 24 sites), reinforcing that these analyses were predominantly exploratory and required appropriate cautious interpretation. The number of participants for whom data were collected was approximately the same in the placebo (n = 51) and intervention arms (n = 54). Comparison of WMH volume at both baseline and 12-month follow-up showed no statistical differences between treatment arms, and regression analysis to measure the change in WMH volume showed that there were no statistically significant differences that may indicate that losartan had any beneficial effect in lowering WMH volume (p = 0.697).
The intended similarly exploratory analysis of CBF was slightly more limited by the number of recruitment sites that could support it. However, the analysis of CBF was also subject to additional limitations posed by the complexities of retrieving data from nine different centres that each used different MRI facilities, different hardware and, as was identified, sometimes adopted different interpretations of the MRI protocols. Although data were collected from a total of 94 participants with similar distribution in the trial arms, it became apparent that there was significant variation in the numbers of scans that could be retrieved that were suitable for analysis once we had applied our QA approaches, both to get indicative results on the potential effect of the intervention and to pilot how data from several centres may be harmonised for analysis.
With these two objectives in mind, we selected data from sites 19 and 10, which between them accounted for almost half of participants from whom data were collected. Our QA of the data from these two centres involved exploration of two approaches. We used either the application of stringent criteria to the data, which were deemed to be of sufficient quality to allow accurate measurement of CBF from the T2-FLAIR scan, or a relatively newer option that uses a spatial CoV approach, which used data to calculate the ratio of the SD of CBF to the mean of CBF in scans. The dual approach was used because the latter is being proposed as a novel robust, but arguably more forgiving, approach when working with data from multicentre studies to allow estimates (proxy) measurements of CBF.106,107 As is evident from the data generated from pilot studies in the two largest sites in the RADAR trial, the number of scans that were able to be analysed was slightly higher using the CoV approach than when using the more restrictive focus on purely CBF scans; in this study, there were too few CBF scans to allow any meaningful analysis. Moreover, owing to differences in the parameters with which the ASL sequences were run at these two sites, it was not possible to readily combine these data to provide a larger data set that would enable a comparison.
Because of the exploratory nature of CBF as an outcome measure and the complexities of this set of data, exploration of the influence of losartan on CBF following the treatment period was restricted to the CoV data from sites 19 and 10. Unfortunately, all efforts to retrieve data in a usable format from the remaining sites have been unsuccessful at the time of writing. From the CoV data available from sites 19 and 10 there were no obvious differences between treatment arms at either baseline or 12-month follow-up; nor was there any evidence of a difference in mean change in CBF analysed using spatial CoV analysis.
The exploratory data suggest that losartan has no effect on either WMH or CBF in patients with AD following 12 months of treatment, although, given the complexities of the latter data, noted previously, some of these conclusions need to be interpreted cautiously because the study was not designed around the secondary analyses. As mentioned, the recently completed NILVAD trial112 provides some limited basis of comparison for the RADAR trial findings. The NILVAD trial found no evidence to suggest that lowering BP with nilvadipine has any beneficial effect on brain volume; however, it also found no evidence of a nilvadipine-associated change in the volume of WMH.113 By contrast, the authors reported that nilvadipine had a beneficial effect on increasing hippocampal blood flow, but not on differences in global CBF or CBF in the posterior cingulate cortex, an area that was preselected for investigation of regional blood flow differences, alongside the hippocampus.113 Again, the fact that this substudy included the measurement of these variables after only 6 months of treatment, compared with 12 months of treatment in the RADAR trial, may have been a limiting factor. Alternatively, it is possible that observations of increased hippocampal blood flow were a true and early but transient phenomenon after participants were randomised to the intervention. However, the fact that there were no similar changes in global or posterior cingulate cortex CBF, and no differences were found in either total brain or hippocampal volume, raises some questions about the impact of the observed increase in hippocampal blood flow on the brain overall. Similarly, the limited numbers of participants for whom data were available and the short follow-up period meant that further analysis of the relationship with cognitive decline, which may have been slightly informative, was not possible.
On balance, it is important to note that the findings of the NILVAD trial are interesting but, given the number of participants, need to be considered with the appropriate levels of caution. The findings of the RADAR trial largely agree with those of the NILVAD substudy and the various MRI outcome measures examined, with the exception of the observed differences in hippocampal blood flow, something that our study did not set out to investigate.
- It was important to pay due consideration to change in BP over time in the RADAR trial given the normal prescribed purpose of losartan. Interestingly, by virtue of this function, the measurement of BP provided the opportunity to have an important positive control that the intervention was biologically active (i.e. observable reductions in BP in the intervention arm would be proof that the medication was active). The other reason why examination of BP was valuable is that previous research has suggested that variations in BP might be associated with the volume of WMH and CBF, which in turn may affect cognitive performance.45–47 Yet there have also been some inconsistent findings.49,50
The findings clearly showed that there was a significant reduction in BP in participants who were randomised to losartan. This equated to a modest drop of approximately 5 mmHg in systolic BP and of 3 mmHg in diastolic BP, although these values are lower than in the NILVAD imaging substudy, which found a drop of in BP of closer to 10 mmHg.113 Although the RADAR trial has longer follow-up data (12 months vs. 6 months) and a larger study population (193 vs. 44) than the NILVAD trial, in contrast to the main NILVAD trial112 and imaging substudy,113 we found no evidence of an association between reduced BP and change in the primary or secondary outcome measures.
The results of the RADAR and NILVAD trials, being intervention studies in patients with mild-to-moderate AD, can be compared with the findings of the SPRINT-MIND study65 of the main SPRINT (NCT01206062) intervention trial.66 SPRINT compared the effectiveness of treating systolic BP in a population of hypertensive individuals aged > 50 years (without diabetes or a history of stroke or AD) with the intention of lowering systolic BP to below 120 mmHg, compared with the normal normotensive threshold of 140 mmHg. The sample size in the SPRINT study was an order of magnitude greater (n = 9361 participants randomised) than that of the NILVAD trial (n = 511 randomised) and the RADAR trial (n = 211 randomised). The SPRINT-MIND substudy was similarly larger (n = 2636) and investigated the impact of the two BP-lowering approaches (evenly distributed in the study population) with respect to cognitive outcomes, namely the rate of new cases of probable dementia (the primary cognitive outcome) and the rate of development of MCI, which was investigated along with, or as a composite outcome measure with, the rate of probable dementia.
First, SPRINT-MIND investigated of small-vessel disease in the brain using MRI65 and its association with cognitive outcomes. No such association was observed, although it must be acknowledged that the early stoppage of the main SPRINT trial, because of the earlier than expected observed therapeutic benefit from more aggressive BP-lowering, meant that the planned follow-ups in SPRINT-MIND were curtailed, resulting in fewer data than was originally planned.65 Yet the lack of observed findings is consistent with the limited findings of both the RADAR trial and the NILVAD trial. The observation from the main SPRINT trial that lowering BP to below 120 mmHg reduced rates of MCI and their composite secondary cognitive outcome measure is not consistent with the findings of the RADAR trial. Yet there are a few notable differences between the studies. First, participants recruited to SPRINT (and SPRINT-MIND), although of comparable lower age to participants in the RADAR trial (50 years in SPRINT and 55 years in the RADAR trial), were hypertensive with no apparent cognitive impairment whereas those in the RADAR trial had mild-to-moderate AD. A similar difference exists between the NILVAD trial and SPRINT. Second, the outcome measures in SPRINT were measures of disease prevention, whereas those in the RADAR trial and the NILVAD trial were measures of reduction in disease progression. Third, the benefits observed in SPRINT were accompanied by a reduction in systolic BP to below the target of 120 mmHg, whereas neither the RADAR trial nor the NILVAD trial achieved such a reduction after 12 months’ treatment [RADAR: mean 133 (SD 21) mmHg; NILVAD: mean 132 (SD 16) mmHg]. This poses the possibility that the reductions in systolic BP achieved in both the RADAR trial and the NILVAD trial may be insufficient to achieve any benefit that may have translated into the chosen primary and secondary outcomes. However, it is also possible that the advanced stage of disease among participants in both the RADAR trial and the NILVAD trial could have negated any potential benefit that may be gained in persons without dementia with no obvious neurodegenerative disease.
In HYVET-COG (Hypertension in the Very Elderly Trial cognitive function assessment), which targeted very elderly people with severe hypertension (systolic BP 160–200 mmHg), the use of indapamide with an optional add-on of the ACE1-I perindopril was compared with placebo.115 The target BP level in HYVET-COG was a systolic BP of 150 mmHg, higher than both the RADAR trial and the NILVAD trial achieved, but the trial failed to demonstrate any reduction in the incidence of dementia. Like SPRINT-MIND, the full scope of HYVET-COG was limited by the early termination of the main HYVET study because of the observed therapeutic benefit.115 Yet it may be reasonable to think that the negative findings in HYVET-COG, as in the RADAR trial and the NILVAD trial, may be attributable to insufficient lowering of BP. In the case of the RADAR trial, this would imply that the insufficient BP lowering, rather than the involvement of dysfunctional biochemical systems (i.e. cRAS), could explain the failure to detect any therapeutic benefit of losartan, and the current results cannot exclude this possibility.
Finally, the results of the RADAR trial differ from those of Tedesco et al., and others, who reported that losartan might confer a beneficial effect on cognitive function.49–51,74 There are again some notable differences to consider. The Tedesco et al.50,51 study was smaller (n = 69), involved people with hypertension but not dementia, and used a smaller maximum dose of losartan (50 mg) but for 26 months. We cannot exclude the possibility that the RADAR trial treatment period was too short to demonstrate a similar cognitive benefit in AD participants. The NILVAD study involved treatment for 78 weeks (compared with 52 weeks in the RADAR trial) and, thus, if BP-lowering is an important factor, the longer exposure periods may have benefited participants in both the RADAR trial and the NILVAD trial.
Given that the RADAR trial is one of a few studies to use a combination of MRI measurements that are reported to serve as a proxy for cognitive decline in a population with mild-to-moderate AD, it was also important to explore the potential associations between the primary and secondary outcomes. This examination may help elucidate which outcome measures may be more suitable in future studies. However, as the data from the RADAR trial were null for the primary and secondary outcome measures, it is not possible to determine which outcome measures would be the most informative for a similar study. It was reassuring to see pertinent correlations between MRI-based measures of hippocampal volume and other relevant measures. However, what may have been somewhat surprising is the lack of any strong correlation between the observer-rated measures of hippocampal atrophy (Scheltens scores) and the MRI-based measures of hippocampal atrophy. Any potential issues this may have caused would have been mitigated in our analysis because we adjusted for Scheltens scores as one of our minimisations.
Levels of drug compliance and tolerability were also an important consideration for the RADAR trial. As already described, there were high levels of compliance (88%) across all participants, with compliance only slightly lower (86%) in the intervention arm than in the placebo arm (90%), which sensitivity analyses confirmed was comparable across study arms. Similarly, the numbers of SAEs and AEs were similar in both treatment arms, showing that the intervention was safe to use.
Post hoc exploratory analyses
We considered a small number of additional post hoc exploratory analyses intended to explore questions that the current data may be able to answer to inform future studies.
The first question that could be explored, given that people with and without hypertension were recruited to the RADAR trial, was whether pre-existing hypertension had an effect on the primary and secondary outcomes. This would potentially shed some further light on whether or not any differences may be apparent in the characteristics of people with AD (in a randomised study) in terms of baseline characteristics or how they changed after 12 months of treatment with the BP-lowering agent losartan. The comparability of all the baseline characteristics, as well as the baseline primary and secondary outcomes, suggested that prior hypertension had no bearing on any of the outcome measures in the RADAR trial. In the recent NILVAD substudy that explored similar MRI-based outcomes to those used in the RADAR trial, about 30% of participants enrolled in the study had prior hypertension. However, the authors of the NILVAD study did not comment on the potential effects of prior existence of hypertension and their outcome measures. This is most likely because the arms in the substudy were already relatively small (n = 22 in each arm) and, therefore, meaningful comparisons of the effects with the various MRI-based outcome measures and assessments would have been significantly underpowered. However, given the considerable evidence reported to date that midlife hypertension is thought to increase risk of developing AD,20,58 which the recent findings of the SPRINT-MIND study also support,66 the absence of obvious differences between the treatment and intervention arms is curious. We did not have access to information regarding duration of prior hypertension at the time of writing; the investigators of the SPRINT-MIND study reported on some ongoing analyses of the medication types that participants were taking while trying to achieve the different target BP levels. It was suggested that RAS-acting medications (e.g. ACE1-Is and AT1RAs) appeared to have a greater benefit than other medications, with some indication that AT1RAs (e.g. losartan) offer the most favourable outcome. As yet there has been no publication to support this statement; however, if found to be true and the evidence significant, it would not only provide renewed evidence in support of the legitimacy of investigating losartan in the RADAR trial but also point to the fact that durations of follow-up longer than the 12 months used in the RADAR trial (median follow-up of participants in SPRINT-MIND was 3.14 years) are likely to be needed.66
Second, the favourable rates of retention of participants, leading to high rates of completion of the primary outcomes and of data completion, were thought to be largely due to the inclusion of an open-label phase as a preceding step to the main randomisation phase. This prompted an examination of the characteristics of those participants who were not randomised, with a view to informing potential revisions to a future study protocol.
There were a number of notable observations from the open-label study. Over half of the participants withdrew within 1 week of entering the study while on the low titration dose for the intervention. Among the 50 participants who withdrew, a surprisingly high proportion (86%) were normotensive. Furthermore, a large number of those people discontinued at the commencement of the study were normotensive participants within the first 7 days (i.e. at the lower dose of the active medication). Yet there was no evidence of unevenness in the randomised phase, showing that our randomisation process worked effectively. Another observation was that almost 20% of the enrolled participants did not proceed to the randomised phase because of an unsuccessful, or lack of, completion of a baseline MRI.
Finally, an exploration of MRI-based measures of hippocampal atrophy was also undertaken as a means to pilot this alternative outcome measure for future trials. As was observed for the primary outcome, there was no evidence to suggest that losartan had a beneficial effect on THV or the volume of individual hippocampi. It was also considered relevant to explore the possible correlation between the MRI-based measures of hippocampal atrophy and rater-based ones (Scheltens score). It was somewhat unexpected to note the poor levels of correlation between these two measures, given that both are intended to measure the same pathological phenomenon.
Conclusions (implications for health care, recommendations for research)
The RADAR trial is the first RCT, to our knowledge, to investigate whether or not the AT1RA losartan has any therapeutic benefit in mild-to-moderate AD and to formally test the angiotensin hypothesis in AD, a hypothesis that is currently gathering increasing levels of support. Furthermore, the RADAR trial is a long overdue study of a BP-lowering intervention in a clinically defined population that numerous observational studies have suggested may benefit from a BP-lowering drug because of decades-old observations of a link between mid-life hypertension and an increased risk of AD. The growing number of preclinical studies that have demonstrated how various means of reducing the synthesis or signalling of AngII reduce rates of cognitive decline and neuropathological hallmarks of AD have continued to support this. Furthermore, there have continued to be some supportive findings, albeit inconsistently, from more recent observational studies. However, the RADAR trial, which used a more objective MRI-based measure as the primary outcomes rather than, as in many studies, assessment-based measures, found no evidence that losartan may provide a therapeutic option for the treatment of AD in the future. This robustly designed study, which had adequate randomisation between treatment arms and favourable participant retention and primary outcome completion rates, showed that losartan did not alter the assumed rate of clinically relevant brain atrophy and unambiguously excluded a reduction in rate of at least 3.8 ml/year. Similarly, there was no significant evidence to support a therapeutic benefit offered by losartan against any of the secondary outcome measures, although this needs to be stated with some caution as the study was not designed around the analysis of these clinical secondary outcomes.
Our study does not exclude the possibility that the duration of treatment, the severity of disease in the participants or the extent to which the drug penetrates the brain are factors that may have resulted in a failure to detect an effect on our primary and secondary outcomes. As described, evidence from the recent SPRINT-MIND study66 showed some reduction in dementia incidence after > 2 years of BP-lowering treatment. However, SPRINT-MIND also examined the therapeutic benefit of aggressive BP-lowering in general, rather than specifically investigating inhibitors of AngII function, or any other individual class of BP-lowering medication. Further analyses of this would be very welcome. Thus, it is not possible to exclude the possibility that more aggressive BP reduction, and for longer, might have had some therapeutic benefit; however, the uniformity of the null findings across all of the primary and secondary outcomes may suggest that AngII inhibition by losartan has no further future as an intervention to be tested.
The findings of the RADAR trial should not serve as a reason for premature termination of investigation of the role of AngII in AD. The evidence in favour of conducting this study was already strong at the time that the RADAR trial commenced, and supportive data have continued to be published in the intervening years. This means that further studies are still required to properly interrogate this hypothesis, as has happened in recent decades to test the therapeutic benefit of stopping or preventing Aβ synthesis, despite several large negative studies.114 It must be stated that, since the RADAR trial concluded, there has been some renewed interest in the Aβ-lowering intervention aducanumab following subsequent analysis of additional data that became available, after the cut-off point used for a futility analysis that triggered the discontinuation of these trials in early 2019. These data, on a subset of participants, suggested that some participants receiving aducanumab at higher doses and for longer demonstrated reductions in brain Aβ levels, as measured by imaging, and encouraged the commercial owners of aducanumab to seek regulatory marketing approval. However, this development is not without some controversy and scepticism as, despite observable reductions in brain Aβ levels, these reductions were not found to be reflected in separate clinical outcomes also measured in the study, casting some further doubt on the clinical efficacy of the drug.116,117 Therefore, some of the considerations that relate to the failure of other drugs testing other mechanisms in AD trials may also apply to the RADAR trial, not least the recruitment of participants with a level of disease that may be insurmountable by any intervention by the time clinical symptoms manifest. Indeed, the design of the RADAR trial and its undertaking coincided with growing debate and now acknowledgement that intervention trials may be more likely to be successful in patient populations in the earliest detectable stages of AD, such as those with MCI; longer follow-up periods of at least 18–24 months should be used to try to measure any form of disease modification.4,5 The recent developments around aducanumab, and in particular the apparent disconnect between evidence of Aβ clearance following longer periods of exposure to the intervention, and in a dose-dependent manner, but in the absence of supportive data around clinical efficacy show once more that there may yet further revisions to perceptions of what may be deemed appropriate follow-up periods in future trials.
Although the findings of the RADAR trial are disappointing with regard to the angiotensin hypothesis of AD, and to patients and their families everywhere hoping for a new treatment, it is possible that the findings of other ongoing related trials will prove to be important in informing whether or not an AngII-inhibiting intervention may have a future role in the treatment and/or lowering the risk of developing AD. Indeed, the recent findings around aducanumab show the importance of having several studies properly test a hypothesis. Aducanumab is still just one in a long succession of compounds that have been tested or are being tested for therapeutic benefit of targeting Aβ. Yet, it is arguably only the first, following a long line of several other failed attempts over several decades,114 to show some promise in a relatively safe manner. With this in mind, it is notable and fortunate that most of the ongoing studies are investigating populations of people who, unlike those in the RADAR trial, have no cognitive impairment but either are at risk of developing dementia [i.e. the HEART study (NCT02471833) and the rrAD study (NCT02913664)] or have MCI [i.e. the CEDAR study (NCT02646982) and the CALIBREX study (NCT01984164)] and thus will provide information on populations of individuals at an earlier stage in the natural history. Moreover, all of these studies use different sartan drugs [rrAD uses losartan, whereas other studies use telmisartan (HEART) or candesartan (CEDAR and CALIBREX)]. Thus, not only will these studies provide information on different outcome measures, but they will do so in various less impaired populations, using a number of other interventions that had similar candidacy to losartan, which was investigated in the RADAR trial. It is therefore conceivable that the contribution of the RADAR trial to the development of any future AngII-inhibiting intervention may be not only in the likely exclusion of losartan as a candidate intervention, but also in providing a robust study design with which to test any other sartan that shows promise in the future.