Sparks DL, Sabbagh MN, Connor DJ, Lopez J, Launer LJ, Browne P, Wasser D, Johnson-Traver S, Lochhead J, Ziolwolski C.
Atorvastatin for the treatment of mild to moderate Alzheimer disease: preliminary results.
Arch Neurol. 2005 May;62(5):753-7.
Please login to recommend the paper.
To make a comment you must login or register.
In response to the paper by Sparks et al.: Recently, several studies reported an absence of noticeable effects on cognition in treated AD patients. All of these studies had as a common denominator the use of low or moderate statin dosages, and for most of these studies treatment extended 3 months or less. Results were disappointing; apart from occasional indications of altered APP processing, no indications of altered cognitive performance were observed (1,2,3).
However, a study by Friedhof and Buxbaum with healthy volunteers already indicated that altering APP processing may require higher levels of statins in humans (4).
This was confirmed and extended by a pilot study (prospective, double blind, placebo-controlled) designed to evaluate whether cerebral Aβ levels respond to statin treatment (5). Following 6 months of high-level simvastatin treatment (80 mg), a significant drop in CSF Aβ was found in the statin-treated AD group. Potentially more important, the decline in MMSE performance was significantly reduced as compared to the placebo-treated group. However, no difference in ADAS-cog was observed and the data indicated that patients in the moderate AD group profited less than those with mild AD from this experimental therapy. With the limited number of patients (20 statin-treated + 17 controls completed the trial), several questions where left open. Would moderate AD patients profit from longer treatment? Would other statins show similar effects? Most importantly, would other trials with a comparable design be able to repeat these findings?
Although the number of treated patients still remains low (26 statin-treated patients + 22 controls completed the trial), Larry Sparks et al. now present very informative answers to several of these questions:
Taken together, one can only congratulate Larry Sparks et al. for this breakthrough. Nevertheless, one has to keep in mind that all of the above-mentioned studies are pilot trials designed to give direction, not to provide final answers. Of course, many open questions remain. The top three on my list are: Can we go below 80 mg when treatment is done for longer times? Would it help to treat AD patients earlier, as our pilot study indicated? Finally, if patients are treated as soon as the diagnosis becomes possible, would the remaining regenerative potential of the brain prevent further disease progression?
Fortunately, all of these questions can be addressed relative safely. Several studies are ongoing which, unlike the previous studies, are powered to answer these questions. It is the results of such studies that may allow us to eventually come to final conclusions regarding the use of statins in prevention and therapy of AD.
The analysis of a subset of patients for whom CSF samples were available before and after production of anti-Aβ antibodies suggests that successful immunization with Aβ may retard further neurodegeneration. Although the number of patients studied is very small, the veracity of these findings is supported by recent animal modeling studies from the laboratories of Frank La Ferla and Karen Ashe.
This week marks publication of a provocative study by Sparks et al. (Sparks et al., 2005). This study suggests that treatment with atorvastatin reduces the progression of Alzheimer disease (AD) in subjects with mild to moderate forms of the disease. The results show benefits that are statistically significant in multiple categories, including ADAS-COG, GDS, and activities of daily living. In many ways, the results observed by Sparks et al. reproduce results observed in a study reported by Simons et al. three years ago, where they treated patients with mild to moderate Alzheimer disease with simvastatin and observed significant reductions in β amyloid levels and significant decrease in the rate of cognitive loss (Simons et al., 2002). These two small studies both provide evidence that statins can prevent the decline in cognitive function in subjects with mild to moderate Alzheimer disease.
The positive results observed by Sparks and Simons contrast sharply with the negative results reported by the PROSPER study and the Heart Study Group (Shepherd et al., 2002; Group 2002). For simplicity, I will refer to the studies by Sparks et al. and Simons et al. as "Alzheimer studies," and I will refer to the studies for PROSPER and Heart Study Group as "cardiovascular," although I am aware that the cardiovascular studies had a significant neurologic component. The cardiovascular studies were originally designed to examine the effects of statins on the incidence of cardiovascular disease, and the investigators added on a cognitive component to examine whether statins might also influence the incidence of Alzheimer disease. The seemingly contradictory results among the cardiovascular studies and the Alzheimer studies raise important questions that beg resolution. While we do not currently know the reason for the contradictory results, these studies differ in important respects, which could provide important clues for future and current studies testing the efficacy of statins in AD.
The subject populations differed greatly between the cardiovascular studies (PROSPER and the Heart Study Group) and the Alzheimer studies (Sparks et al. and Simons et al.). The cardiovascular studies were truly massive studies, one with ~9,000 subjects initially (2,891 treated with pravastatin) and the other with ~20,000 subjects initially (~10,000 treated with simvastatin). In comparison, the two Alzheimer studies were small. The study by Sparks examined 56 subjects (25 treated subjects completing the study), while the study by Simons examined 37 subjects (17 treated) (Sparks et al., 2005; Simons et al., 2002). However, the actual number of Alzheimer cases studied in each group was surprisingly similar. Only 31 subjects (0.3 percent) of the subjects in the Heart PROSPER Study were characterized as having Alzheimer disease, and the PROSPER study did not specifically list people with the diagnosis of Alzheimer disease (Shepherd et al., 2002; Group 2002).
The differing numbers of patients among the two types of studies reflect an important conceptual difference between the cardiovascular and Alzheimer studies. The study by Sparks et al. sought to measure progression of AD. Because of the focus on progression of symptoms, Sparks and colleagues designed their study to carefully quantify cognitive function among the study participants. The Simons study also quantified progression of cognitive loss in Alzheimer subjects, although the level of detail was far less than that documented by Sparks and colleagues. By contrast, the cardiovascular studies did not strongly address the issue of Alzheimer disease. The PROSPER study looked at cognitive decline among the entire cohort of subjects, which is a general measure that does not specifically address the pathophysiology of Alzheimer disease. It is possible that any effect relevant to Alzheimer disease was lost among the signal of other factors contributing to cognitive decline. The Heart Study Group examined the incidence of Alzheimer disease rather than progression of the disease. Because of this objective, the cardiovascular studies simply assessed whether there was dementia, which is a binary "yes/no" measure. I believe that if the results of Sparks et al. are borne out by future studies, the distinction between looking at progression of cognitive loss in Alzheimer cases, rather than incidence of Alzheimer disease or cognitive loss in the general population, will be key factors.
A second important difference might lie in dosing; I have heard other investigators mention this as a significant consideration. I do not know how important this issue is, but it is well worth considering. The cardiovascular studies utilized doses of statins that were at the moderate end of the recommended doses (up to 40 mg QD of simvastatin or pravastatin). In contrast, Sparks and colleagues used dosing that is on the moderate to high end (40/80 mg) of recommended dosing. This difference could have important consequences. Studies by Sparks et al., Friedhoff et al., Simons et al., and Hogland et al. have all examined the effects of statins on Aβ levels (Sparks et al., 2005; Friedhoff et al., 2001; Hoglund et al., 2004; Simons et al., 2002). Sparks used a relatively high dose of atorvastatin (40/80 mg QD) and has reported at meetings a dose-dependent reduction in Aβ. Friedhoff used a high dose of slow release lovastatin, and observed a reduction in Aβ levels associated with statin use (Friedhoff et al., 2001). In contrast, Hoglund et al. used doses of statins on the lower end of the recommended range (20 mg atorvastatin QD or 40 mg simvastatin QD) and failed to observe a reduction of Aβ levels with statin use (Simons et al., 2002). The confusing aspect of this issue is that Simons and colleagues also used 40 mg QD of simvastatin and observed a reduction in Aβ. Although not entirely consistent, these results raise the possibility that higher levels of statins might exert effects relevant to AD not observed with lower doses of statins. If true, use of lower doses of statins might also contribute to the negative outcome of the cardiovascular studies with respect to prevention of Alzheimer disease. However, this argument is quite tenuous.
It is possible that the small sample sizes of the Sparks and Simons studies contributed to false positive results. However, it is also possible that the results from both the cardiovascular studies and the Alzheimer studies are real. How could this be? Perhaps statins reduce the progression of AD, but do not prevent the incidence of AD and also do not prevent other forms of dementia. Our current knowledge of the pathophysiology of AD does not provide a clear mechanism for distinguishing between processes that might occur earlier in the disease process, and be associated with incidence of AD, from those that occur later in the disease and be associated with progression of existing disease. However, these studies might be identifying just such a distinction.
Just a quick note on dosing in the statin studies. The Simons study used 80 mg simvastatin; 40 mg were used for the first month, then patients were put to 80 mg. One reason for doing this was that at the time the study was initiated, the use of 80 mg simvastatin was rather new and we anticipated that it would be safer to start with a lower dose. By now, it appears that this was an overly cautious procedure.
This puts the dosing of the Alzheimer sudies, which found a beneficial cognitive response in a distinct group, using at least twice the statin amount than other studies which did not observe a beneficial effect. Ben Wolozin very importantly raises the point of lower doses in respect to "beneficial side effects" and to prevention.
In this recent paper, Sparks and colleagues have reported encouraging preliminary data showing a beneficial effect of statin treatment on cognitive decline due to probable AD. Suggestions of a link between cholesterol metabolism and AD have come from many scientific arenas over the years, but have yet to be fully elucidated. Results from epidemiological studies have shown an association between hypercholesterolemia and AD, but the data have been mixed (Jarvik et al., 1995; Kalmijn et al., 1997; Kuo et al., 1998; Notkola et al., 1998; Romas et al., 1999). The initial retrospective studies showing reduced AD/dementia risk with statin use were very provocative (Jick et al., 2000; Wolozin et al., 2000); however, results from more recent prospective studies of statin use have been mixed (Group, 2002; Shepherd et al., 2002; Sparks et al., 2005). Clearly, many variables can contribute to the outcome of such studies, including clinical characteristics of the patient population, specific statin, dosage and length of treatment, clinical and biological outcome measures, and so on. Results from animal studies demonstrating alterations in AD-related pathology in response to high-fat diets or cholesterol-lowering drugs are compelling (Howland et al., 1998; Refolo et al., 2000; Fassbender et al., 2001; Refolo et al., 2001), and the in vitro data showing cholesterol effects on Aβ generation provide a plausible mechanistic explanation (Simons et al., 1998; Frears et al., 1999; Fassbender et al., 2001).
The preliminary findings described in the current paper are certainly promising in terms of identifying a possible clinical therapy, but the mechanism of action remains to be defined. The pleiotrophic effects of statins, including inflammatory and vascular effects, could conceivably impact AD pathogenesis. The authors themselves acknowledge the possibility of a non-cholesterol-lowering mechanism for the observed effect on cognition. In terms of the possible Aβ connection, it would be interesting to know if CSF and/or plasma levels of Aβ in their cohort were altered with statin use, comparing levels at study entry with levels at the time of study completion. This would provide some insight into effects on Aβ metabolism as a possible mechanism of action.
Despite the abundant literature suggesting a connection between cholesterol metabolism and AD-related processes, I am still not convinced that the effects of hypercholesterolemia or statin use on cognitive or biological (e.g., plaque load, Aβ levels) outcomes involve cholesterol per se. We recently reported a lack of effect of endogenous plasma cholesterol levels on brain Aβ levels and pathology in the PDAPP mouse model of AD (Fagan et al., 2004). We wanted to test the role of plasma cholesterol levels on Aβ pathology in this model without resorting to the use of non-physiologic high-fat diets (known to cause pathology in multiple systems) or pharmacologic manipulations with drugs with possible pleiotrophic effects (e.g., statins). To do this, we took a genetic approach and bred PDAPP mice with mice lacking ApoA1. ApoA1-null mice have severely reduced plasma cholesterol levels (by ~75 percent) due to the virtual absence of HDL, the primary lipoprotein in mice. Despite a marked reduction in plasma (and brain, ~40 percent) cholesterol levels in PDAPP/ApoA1-/- mice, Aβ-related parameters were not changed. Furthermore, while plasma ApoE levels actually increased in PDAPP/ApoA1-/- mice compared to littermate controls, brain ApoE levels remained unchanged. We hypothesized that it is perhaps the level of brain ApoE, and not the level of brain or plasma cholesterol per se that influences Aβ metabolism, and by extension, perhaps dementia risk. In view of the published literature, it is conceivable that effects of high-fat diets and statin treatment previously attributed to cholesterol may actually be due to altered levels of brain ApoE. High-fat diets not only increase the level of cholesterol, but also ApoE, in the brain (Sparks et al., 1995; Howland et al., 1998; Wu et al., 2003), and statins decrease them both (Naidu et al., 2002; Petanceska et al., 2003). Thus, it has not been possible to distinguish putative effects of cholesterol from those of ApoE in the many studies published to date. Does this mechanistic nuance have any bearing on whether statins will have therapeutic value in AD? No, probably not. Consideration of this alternative hypothesis does, however, open up the possibility of additional targets (e.g., ApoE) that warrant exploration, and cautions against an automatic presumption of a cholesterol mechanism in the hypercholesterolemia/statin/AD connection.
Recent epidemiological studies (Jick et al., 2000; Wolozin et al., 2000; Heart Protection Study Group, 2002; Shepherd et al., 2002; Zandi et al., 2005) have led to contradictory conclusions regarding the efficacy of statin treatment for AD. As Dr. Wolozin points out in his commentary, it may be that statins reduce AD progression, rather than decrease disease incidence. In support of this, the double-blind, placebo-controlled randomized pilot trial by Sparks et al. offers some intriguing findings, suggesting that statins may be of some benefit in reducing dementia progression in both mild and moderate AD patients (Sparks et al., 2005). A significant benefit of atorvastatin treatment for 12 months was observed for GDS score, and trends toward significant differences for ADAS-cog, CGIC, and NPI were seen between the atorvastatin and placebo-groups, although significance was not obtained for MMSE or ADCS-ADL scores.
It is widely believed that the potential beneficial effects of statin treatment as an AD therapeutic are related to the cholesterol-lowering properties of statins. However, whether benefit is mediated through changes in serum or brain cholesterol, or whether the effects are direct (e.g., lower brain cholesterol causing reduced Aβ generation) or indirect (e.g., lower serum cholesterol causing improved cerebrovascular function, thus reducing AD progression) remains unknown. As Drs. Gandy and Petanceska point out in their commentary, lipophilic statins such as simvastatin penetrate the blood-brain barrier (BBB), whereas hydrophilic statins such as atorvastatin do not. Interestingly, Simons et al. reported a 26-week randomized, placebo-controlled, double-blind trial using simvastatin that showed a significant benefit in the MMSE scores of the simvastatin treated group as compared to placebo (Simons et al., 2002). As anticipated, statin treatment in both Sparks and Simons trials significantly lowered plasma cholesterol levels, and in the Simons study, cerebral cholesterol metabolism was also affected following treatment. It is interesting to note that both trials showed statin-related cognitive benefits, although one used a BBB-penetrant statin while the other did not. This would seem to support the notion that the beneficial effect was associated with lower serum, rather than brain, cholesterol levels. As discussed by Drs. Wolozin and Hartmann, use of high-dose simvastatin was also associated with reduced Aβ levels (Simons et al., 2002). However, whether or not atorvastatin treatment affected either cerebral cholesterol or Aβ levels (or both) was not determined in the Sparks study, and the mechanisms underlying the potential benefits of statins on measures of cognition remain a mystery.
Given the relatively short time frame of both the Sparks and Simons trials, it should be considered that any putative beneficial effects that the statins exert may be entirely independent of changes in amyloid load and could instead be a secondary effect due to an improvement in cardiovascular or cerebrovascular function. Indeed, recent data has indicated that atorvastatin treatment promotes both angiogenesis and neuronal plasticity (Chen et al., 2005). Obviously, further trials involving the use of statins of different lipophilicity, over a range of doses for more prolonged periods, are required to obtain definitive results.
In light of the relatively short trial time, and given the fact that stroke benefit from statins is not apparent until ~3 years of treatment (Byington et al., 2001; Pedersen et al., 1998), increasing the trial time beyond 1 year may reveal more robust positive effects on AD progression. However, a note of caution: Although a ~5-year administration of high-dose atorvastatin (similar to the dosage used by Sparks) to patients with stable coronary heart disease provided significant clinical benefit beyond that afforded by 10 mg per day atorvastatin, this benefit occurred with a greater incidence of elevated aminotransferase levels (LaRosa, et al., 2005). While Sparks and colleagues screened for adverse changes in liver function and monitored for muscle derangements and rhabdomyolysis following administration of 80 mg per day atorvastatin, it remains unclear as to why, out of the 63 patients considered evaluable after completing the 3-month visit, only 46 individuals completed the 12-month study.
In summary, while not all the large-scale epidemiologic studies have found a link between statin use and AD prevention, the small-scale, randomized clinical trial of Sparks et al. appears to support the notion that statin treatment may reduce AD progression. However, given the small sample sizes of both the Sparks and the Simons studies, the data from well-designed, large-scale multicenter clinical trials clarifying the safety and efficacy of long-term, high-dose statin treatment on AD progression is avidly awaited.
I must start by saying that it is quite gratifying that there has been such interest in our clinical trial (AD Cholesterol-Lowering Treatment—ADCLT trial; the Lipitor trial) (1). As the very first AD treatment trial testing a statin medication for clinical benefit, other shorter investigations were initiated and completed during the course of the ADCLT, including the Simons study. We initiated our study cognizant of many mechanisms by which atorvastatin could produce clinical benefit in AD, but deemed it more important to demonstrate clinical efficacy and argue over mechanism later. We now have shown clinical benefit, and discussions of the mechanism are clearly warranted. We, of course, respect and acknowledge each investigator’s opinion as to the mechanism of atorvastatin action, but must clarify certain issues and correct some factual errors.
As noted by Dr. Hartman, the Simons study was a 26-week study of simvastatin where stable performance on the Mini Mental State Exam (MMSE) in the treatment group was significantly different from the placebo group. This was because the placebo group showed a somewhat accelerated 4-point deterioration between baseline and the 6-month evaluation. At the 6-month time-point in the ADCLT, we observed stable performance on the MMSE in the atorvastatin-treated population that was not significantly different from the placebo group—as the placebo group only deteriorated two points. The differences between the studies on the MMSE are more due to the rate of deterioration in untreated participants with AD.
Dr. Wolozin indicates that we reported a significant difference on the Alzheimer’s Disease Cooperative Study Activities of Daily Living (ADCS-ADL) index, when in fact this was the one of six clinical instruments where there was no positive signal produced by atorvastatin. This investigator suggests that the Simons trial and the ADCLT were of similar size when actually the ADCLT included twice as many subjects with mild to moderate AD. We include data from 63 participants—32 on atorvastatin and 31 placebo with 25 treated subjects completing the 1-year investigation, whereas the Simons study included 37 subjects, with 13 of 17 simvastatin-treated subjects completing the 6-month study. Dr. Wolozin also suggests that we have reported a dose-related decrease in circulating Aβ levels with atorvastatin treatment, when in fact we have found slight gradual increases in both Aβ40 and 42 that were not significant. These data will be published in the near future (2).
We agree with Drs. Gandy and Petanceska that atorvastatin may partially reinstate clearance of Aβ from the brain. The observed gradual increase in circulating Aβ levels among subjects treated with atorvastatin may be associated with the gradual reductions of ceruloplasmin—the copper chaperone in the blood. We have shown that increased copper/ceruloplasmin levels promote central accumulation of Aβ in the cholesterol-fed rabbit model of AD, while reduced circulating copper/ceruloplasmin allows clearance to the blood and minimal accumulation in brain (3-5).
In reply to Dr. Fagan, as part of the preplanned design of the ADCLT, we explored many more circulating markers than initially published to assess multiple mechanistic avenues. As noted above, we have determined the effect of atorvastatin treatment on Aβ levels. We have also established that atorvastatin produces reduced circulating ApoE levels during the time-course of treatment (2). I would also note that in addition to cholesterol-fed animals, nondemented individuals with autopsy-confirmed critical coronary artery disease (>75 percent stenosis) exhibit increased accumulation of Aβ and ApoE in the brain (6-8), thus making it difficult to separate the interrelationships among cholesterol, ApoE and Aβ.
In answer to Dr. Cole's and Dr. Vassar's queries, as part of the investigation of possible mechanisms of atorvastatin action in association with observed clinical benefit, we will soon report the effect of treatment on circulating free radical load (superoxide dismutase and glutathione peroxidase activities) and HDL/LDL/VLDL levels (9). In addition, we will be reporting the effect of active treatment on volumetric alterations measured by MRI, assessment of clinical parameters during the 1-year “open-label" extension of the ADCLT (Geneva/Springfield Conference, 2006), and treatment-related changes in circulating levels of ApoA1, ApoB, copper, 24OH- and 27OH-cholesterol, CRP, and CD40. Furthermore, in two weeks we will be reporting at the Alzheimer’s Association meeting in Washington, DC, the influence of initial cognitive impairment, initial cholesterol levels, and ApoE genotype on the clinical benefit produced by atorvastatin.