Skip to main content

Clinical Trials

A New Parkinson’s Therapy – Possibly

There’s a new report of progress in Parkinson’s disease, and from an unexpected direction. Well, it was unexpected for me, anyway. Parkinson’s is, famously, a condition that is driven by the steady deterioration of dopamine-rich neurons in the brain, most particularly in the substantia nigra region. An impressive amount of research over the years has gone into the study of this part of the brain and to dopamine handling in general. What sets off Parkinson’s and how to interrupt its progress are still very much open questions. It’s almost certainly a combination of genetic and environmental causes, but the details of that tangle are yet to be made clear.

This latest work, though, involves dosing a known diabetes medication, the GLP-1 agonist exenatide. That’s a peptide drug, famous among those who follow these things (I’m one) as having been derived from an active protein found in Gila monster saliva. Type II diabetes does seem to be a risk factor for Parkinson’s (as well as for Alzheimer’s), and there’s been  a lot of work on the “gut-brain axis” and what’s connected to what. Exenatide has effects on appetite, crosses the blood-brain barrier to some degree, and has shown protective and growth-factor effects on neurons in vitro. So it’s certainly possible that it could have neuroprotective effects on a degenerative disorder like Parkinson’s, and this paper is the first double-blinded trial putting that idea to the test.

There’s some promise. This was not a large trial (about thirty patients in each group), but at 60 weeks (48 treatment and 12 weeks of washout) motor function did seem to improve slightly in the treatment group, while it continued to worsen in the controls. The same team had previously done an open-label proof of concept study, and that one showed motor improvement as well, along with an even more significant change in cognitive assessment scores. This one, though, showed no real change in the latter, which tells you something about open-label measures of cognition and mood. But based on the motor function data, this could lead to a completely new direction in Parkinson’s therapy.

It needs one. There’s not a lot that can be done for Parkinson’s patients, outside of dopamine replacement, which isn’t that effective in the long run. This study is encouraging, in that it shows that there could be a route to neuroprotection outside of what’s been tried already. But at the same time, it’s worth thinking about this recent item, on statistics in biomedical and social science data. There’s been a call to reduce the threshold for what’s considered significant from the traditional p=0.05 down to p=0.005. If that’s the threshold, then this paper’s results don’t make the cut: they have p=0.03. So when if I say that the data do seem to be real, keep in mind that that’s for certain values of “real”. There’s a limit to what you can get in a trial this small, of course (which takes us back to the concept of effect size). And that means that the first thing to do, when you get a result like this, is to see if it holds up in a larger population.

“Run it on a larger population” is the general answer for statistical doubts (and in fact has been proposed as an alternative to that p-lowering proposal), but that means that a lot of studies just wouldn’t be done in the first place for lack of funding. Some of them won’t be missed, but some of them surely will be, so it’s a cost/benefit analysis and a difficult one to get a handle on. So is the p=0.005 proposal, of course: that will mean more negative studies, some of which will be even-harder-to-overcome false negatives, and all of them will be harder to publish under the current editorial systems. Nothing gets improved for free.

Do I think that the benefits seen in this Parkinson’s study are real? Yeah, if pressed, I think so – probably. The overall mechanism makes sense, and there are several lines of evidence that point the same way from earlier work. But at the same time, I wouldn’t go wild until seeing it reproduced in a larger cohort of patients, which will take a lot more money and time (and remember, this is already the larger cohort as compared to the first preliminary trial). There were no particular adverse event flags in this study, but in the same way that you have to be cautious about the good parts, you have to be cautious about the bad parts, too. So anyone with Parkinson’s who’s thinking about trying this therapy free-lance should keep that in mind as well; it’s yet another trade-off.

This is why we run trials. And this is why the trials have to be big ones, and long ones, and expensive ones. I very much hope that this points the way to a new Parkinson’s therapy, but it’s honestly still too soon to say that it’s there yet. Things have to hold up, and we have to find out if they do.

35 comments on “A New Parkinson’s Therapy – Possibly”

  1. Lane Simonian says:

    I don’t really like beating a dead horse, but on the other hand you almost have to do it to try to achieve a paradigm shift.

    “Alleviation of Hyperglycemia Induced Vascular Endothelial Injury by Exenatide Might Be Related to the Reduction of Nitrooxidative Stress”

    “Nitrooxidative Stress and Neurodegeneration”
    Michal Fiedorowicz and Pawel Grieb
    Mossakowski Medical Research Centre, Polish Academy of Sciences, Warsaw

    This approach has also been tried for Alzheimer’s disease with hints of a few, slightly positive results.

    “In Alzheimer’s Disease, 6-Month Treatment with GLP-1 Analog Prevents Decline of Brain Glucose Metabolism: Randomized, Placebo-Controlled, Double-Blind Clinical Trial”

    1. Passerby says:

      “I don’t really like beating a dead horse”

      Oh I think you do.

  2. Wile E. Coyote, Genius says:

    Regarding the shift from 0.05 to 0.005. Doesn’t this cut both ways? The FDA wants to see both efficacy and safety in a clinical trial. By shifting to 0.005, do not a lot of adverse events become non-statistically significant, making it much easier to demonstrate safety?

    1. Eric says:

      No, not really. Adverse events don’t necessarily need to be statistically significant to be of concern to the FDA. A single SAE can be viewed as a safety risk regardless of the statistical significance.

      1. Wile E. Coyote, Genius says:

        suicidal thoughts or tendencies?

  3. Barry says:

    If the Blood-brain-barrier doesn’t stop exenatide (Molar mass 4186.6 g/mol 17 hydrophilic sidechains out of 38), what does it stop?

    Kastin AJ, Akerstrom V. Entry of exendin-4 into brain is rapid but may be limited at high doses. Int J Obes Relat Metab Disord. 2003;27(3):313–318.

    Does “limited at high doses” mean the authors believe it’s actively transported?

  4. Anon says:

    What I don’t understand is: Given that 5% of drugs tested will meet the p-Value criteria of 0.05 by random chance alone, how do we know that *any* of them that are eventually approved are actually effective?

    1. Isidore says:

      95% of them will be.

      1. shar says:

        Not necessarily. If you test 1000 homeopathic medicines in 1000 clinical trials at a P < 0.05 threshold you'd expect approximately 50 "significant" results. That doesn't mean that 95% of those 50 are actually effective; 0% are, because 100% of them are sugar pills. Given that the overall failure rate of clinical trials is pushing 90%, it's not a trivial concern that the successful 10% is contaminated by false positives.

        However, to Anon's question, some partial answers:
        – Some drugs show effects at a much higher significance threshold. Sofusbuvir (Sovaldi) managed P<0.001 for at least two of its Phase 3 trials.
        – I don't actually know that a single result with P<0.05 is sufficient for FDA approval. If that were the case, it would be hard to explain why the failure rate for Alzheimer's trials is ~99.9%.
        – Even after drugs are approved, post-approval studies can continue to measure the effectiveness, with fewer controls/randomization but larger N/longer timescales. Drugs that don't live up to their trial results can theoretically be pulled, given new warnings, etc.

        1. Patrick says:

          Well a single 0.05 result definitely isn’t enough for approval – phase 2 and phase 3 are enough to preclude that. If we take just the raw multiple, that’s now only a 1 in 400 chance of duds. A bit better.

        2. M. Welinder says:

          “Given that the overall failure rate of clinical trials is pushing 90%”

          For this purpose you don’t want to count failures due to, say, toxicity. That will make
          the situation look a little less grim.

    2. Anon says:

      The key metric is the false discovery rate, and the only way to keep that below a certain threshold is by reducing the critical p-value with each successive clinical trial according to corrections required for multiple hypothesis testing. By now (after tens of thousands of clinical trials), we should require a p-value of less than 0.0000005 to ensure that just 95% of approved medicines actually work!

      1. M says:

        The p-value you quote seems like an application of the Bonferroni correction, which is the adjustment for *any* false positives–that is what we’d want if we never wanted to make a mistake.

        The false discovery rate is much more relaxed, and would allow us to make a mistake 5% of the time. Not exactly, because for reasons others have explained that’s not actually what p=0.05 means. Regardless I’m too tired to sort out even to myself what FDR means in the context of 10,000 different independent experiments.

  5. Brian says:

    An ignorant question, but would this (and other cases of marketed drugs having potentially beneficial secondary effects) be a situation where giving biostatisticians access to health data from something like the UK’s National Health Service allow people to figure out if these trends are real or not? If there are 126,000 people with Parkinson’s disease in the UK currently and 3-5% of them use exenatide, is there some health metric that could be captured from current records? Obviously, if the effect is small or subtle, you might not see anything, but…?

    1. Alan Goldhammer says:

      One can get access to the NHS and other European databases. There are also some managed care databases in the US that can be made available. When I was at PhRMA we conceived of a project that could do just what you describe. It was the Observational Medical Outcomes Project, and the business plan conceived of doing both safety and efficacy studies across disparate databases in the US. We got funding from companies for several years and the project was turned over to the Reagan/Udall Institute for the FDA.

      It’s very difficult to do this type of research but it’s one of the few realistic ways to mine all the extant data.

      1. Scott says:

        You can probably get access to Department of Veteran’s Affairs clinical data, too, the VAMC have been running on computerized charts for longer than anyone else in the nation, IIRC.

        Though that does run you into trouble with comorbidities…

  6. Daniel Barkalow says:

    I wonder if it wouldn’t be best to have some sort of “demonstration of promise” level of evidence. Maybe if you can get a p<0.05 result, you can get approval for insurance to pay for the drug during your huge long study, and if that works, you get approval to use it to treat patients without them being in a study. It seems like it would be possible to arrange things so that the drug company makes a little profit on running the huge long study, but only makes back its initial investment if and when the study finds the drug worked (getting a period of exclusivity starting at that point).

    Right now, I think patients would rationally get this treatment if it's free to them, and the medical community wants to know the outcome enough that they'd be willing to pay the cost of finding out, so the drug company shouldn't be on the hook for costs, for as long as people think it seems to be working.

    1. tangent says:

      Our system has to notice how much benefit (and harm), please, not merely “we’re 95% confident the [net] benefit is greater than zero.” Yeah, it means some judgment, since there’s no reasonable blanket standard for how much benefit is enough. But using p < 0.05 alone is a choice of a blanket standard too — the standard is there is no standard — which is surely not a good one.

      (As an onlooker I wonder why every paper doesn't present a confidence interval on an actual figure of interest. "Relative risk of the cardiac endpoint is [0.66, 0.92]" Oh, many papers do do this, but why does anyone ever get away with the “decreases cardiac risk (p = 0.02)” presentation?)

  7. Lane Simonian says:

    Maybe not particularly helpful after all at least in regards to Alzheimer’s disease:

    “The GLP-1 receptor agonists exenatide and liraglutide activate Glucose transport by an AMPK-dependent mechanism”

    “AMP-activated protein kinase: a potential player in Alzheimer’s disease.”

  8. Chris Phoenix says:

    Am I missing something about basic statistics?

    p=0.05 is p=0.05 regardless of the population size: 1 in 20 false positives. So, how can increasing population size be an alternative to changing the required p value?

    1. Mister B. says:

      p = 0.05 is a consequence of 1 in 20 false positives.
      If you only have 1 false positive in a cohort of 200, you reach the p = 0.005.

      You can see a real difference if false positives don’t follow a linear increase. 1 in 20 does not mean 10 in 200, but maybe 5 or 15 in 200.

      I am no statisticiant at all, but it is the way I understand the sentence “go bigger” in clinical trials.

      1. Cellbio says:

        Bigger enables reaching significance in smaller effect sizes.

  9. Anders Carlsson says:

    You are right. The p-value is just the probability, given that no real effect exist, that you will get the same (or more extreme) result due to randomness. It is not the probability of the null hypothesis being true.

    Increased population size will not mitigate this. It’s pretty simple to run “simulations” with an increasing number of random numbers in Excel, and you’ll find that low p-values are just as common when using hundreds or thousands of samples as when using a handful (an experiment I recommend every new (and old) scientist to do).

    Population size is more coupled to effect size, i.e. if you are comparing mice to elephants you don’t need that many animals to find differences you believe in. Jaguars to leopards – you probably want to look at a few before feeling confident.

  10. dearieme says:

    At last, good news for Mrs Clinton.

    1. joe shmoe says:

      But unfortunately there is no helping Mr. Trump (or the world).

  11. steve says:

    There have been some recent interesting, but seemingly contradictory, data on Parkinson’s. The finding that Parkinson’s patients have T cells reactive with alpha-synuclein would suggest an autoimmune component. Other data suggest that Parkinson’s begins in the gut as misfolded proteins that migrate to the brain; vagotomy reduces the incidence of Parkinson’s by 40% and the gut microbiome has been shown to play a role as well. Now we have data on exenatide. How all these will play out I’m not sure but I doubt that the answer will be through peroxynitrites.

  12. jb says:

    Another week and something sugar related makes it’s way to the headlines.

    Parkinson’s has long been characterized by hypometabolism of glucose:

    Arch Neurol. 1992 Dec;49(12):1262-8.
    Cerebral glucose metabolism in Parkinson’s disease with and without dementia.

    J Neurol Neurosurg Psychiatry. 2017 Apr;88(4):310-316
    Cerebral glucose metabolism and cognition in newly diagnosed Parkinson’s disease: ICICLE-PD study

    Every single time a there’s a sugar related story about a possible link or role in a major disease: look to a phenomena called O-GlcNAc. Glucose is metabolized by cells to create N-AcetylGlucosamine (GlcNAc). GlcNAc gets added to proteins as a post translational modification (O-GlcNAc) in a way that’s almost the yin to the yang of protein phosphorylation and regulates virtually every aspect of the way proteins function, how they fold, where they’re trafficked…the point though being the fact that O-GlcNAc is probably THE fundamental link between sugar uptake and metabolism and direct influence on cellular physiology. The list of major diseases in which abnormal glucose metabolism that results in abnormal patterns of O-GlcNAc (and also where O-GlcNAc has been found on major, major proteins that are believed to be central in the pathophysiologies of the conditions below) :





    And the list goes on.

    What makes this study interesting is the fact PD is characterized by hypometabolism and that they appear to be able to treat PD by reversing this state by administering a diabetic drug in order to stimulate the uptake of glucose into cells. Recently, however, the role of O-GlcNAc in regulating alpha synuclein(a protein under intense investigation for its role in PD) has been described:

    Nat Chem. 2015 Nov;7(11):913-20.
    O-GlcNAc modification blocks the aggregation and toxicity of the protein α-synuclein associated with Parkinson’s disease.

    Chembiochem. 2012 Dec 21;13(18):2665-70
    O-GlcNAc modification prevents peptide-dependent acceleration of α-synuclein aggregation.

    So yes, this mechanism of treating PD with a diabetes drug may very well be plausible with the critical link being O-GlcNAc. By increasing glucose uptake in a diseased PD brain you might be increasing global levels of O-GlcNAc which could subsequently help prevent alpha synuclein aggregation. O-GlcNAc also has alllllllllll sorts of importance during co-translational events that profoundly effect protein folding, so the link between abnormal carbohydrate metabolism and PD, AZ, and other major diseases with misfolded protein phenomena could very well be O-GlcNAc.

    1. steve says:

      The O-glcnac pathway is interesting but you have to take into account recent results from an inducible knockout that saw little effect except in the pancreas ( Such is the problem with seeing one pathway as the end-all and be-all of biology, whether it be glycosylation or peroxynitrites. Life is complicated.

      1. Lane Simonian says:

        The story becomes a little murkier. O-GlcNAcylation by preventing phosphorylation may like nitration contribute to diabetes.

        “The O-GlcNAcylation of IRS1 inhibits phosphorylation required
        for its interaction with the p85 regulatory subunit of phosphatidylinositol-3-kinase.”

        It may further impair the neuroprotective Akt pathway:

        “The phospho-activation of protein kinase B (AKT) at T308 is inhibited by its O-GlcNAcylation, leading to downregulation of AKT function”

        The beneficial aspects of O-GlcNAcylation early on at least may be due to its role in activating AMPK.

        Thus studies as to whether O-GlcNAcylation is a positive or negative regulator of diseases are all over the place. But this conclusion may be a critical one for Alzheimer’s disease and other diseases.

        “One of the mechanism by which diabetes could accelerate the development of AD is increased ROS production. Insulin resistance in type 2 diabetes was also found to be associated with AD. How big of a role O-GlcNAc plays in these processes and how these processes are related to the regulation of tau phosphorylation still needs to be addressed. Nevertheless, we think that studying O-GlcNAc modification simultaneously with tau phoshorylation in oxidative stress-exposed neuronal cells should significantly contribute to the understanding of the development of Alzheimer’s disease.”

      2. Jb says:

        Steve – I think you are completely misunderstanding what I was trying to say. The idea isn’t to show that O-GlcNAc affects metabolism, but rather the idea is that metabolism and the metabolic fluxes of glucose affect patterns of O-GlcNAc since glcnac is a direct metabolite from glucose. That’s a phenomena that’s been well known and extensively published on. There have been oodles of papers describing reduction of global O-GlcNAc under periods of starvation in in-vivo models
        Brain. 2009 Jul;132(Pt 7):1820-32.
        Reduced O-GlcNAcylation links lower brain glucose metabolism and tau pathology in Alzheimer’s disease.

        Eur J Neurosci. 2006 Apr;23(8):2078-86
        Concurrent alterations of O-GlcNAcylation and phosphorylation of tau in mouse brains during fasting.

        The point here is simply that 1.) We have seen marked reduction in glucose metabolism in major diseases like PD and AZ 2.) O-GlcNAc is very sensitive to the metabolic state of glucose metabolism and 3.) O-GlcNAc is involved in all sorts of pathways and proteins involved central to diseases like AZ and PD,especially those characterized by misfolding phenomena of proteins since O-GlcNAc is absolutely critical for maintenance of protein turnover and folding during co-translation. Thus it could make sense that PD, which is characterized by a marked reduction of glucose metabolism could in theory be treated with drugs that stimulate glucose uptake and metabolism, with the fundamental link between the two being the effect that glucose uptake stimulating drugs have on patterns of O-GlcNAc (which again is vital for regulation of alpha synuclein).

  13. Lane Simonian says:

    The type of research and observations done by jb are quite helpful. Here is an article on the ties between O-GlcNaCylation and diabetes.

    And here is an article discussing the complex interplay between O-GlcNaCylation and oxidative stress.

    Here, too, is where diabetes drugs that work through AMPK activation come in. Under normal circumstances, AMPK activation would reduce oxidative stress via activation of the Akt pathway (which is neuroprotective), but under cases of oxidative stress this pathway is inhibited or blocked (such as in Parkinson’s disease and Alzheimer’s disease). Under these conditions, AMPK activity increases O-GlcNaCcylation which further increases oxidative stress. An AMPK-activating diabetes drug such as metformin thus can at a certain point increase the risk for both Parkinson’s disease and Alzheimer’s disease.

    1. Jb says:

      AMPK itself is O-Glcnac modified on its alpha and gamma subunits
      J Biol Chem. 2014 Apr 11;289(15):10592-606

      Cross-talk between two essential nutrient-sensitive enzymes: O-GlcNAc transferase (OGT) and AMP-activated protein kinase (AMPK).

      Bullen JW1, Balsbaugh JL, Chanda D, Shabanowitz J, Hunt DF, Neumann D, Hart GW

  14. Lane Simonian says:

    This one is the icing on the cake:

    The Role of Glucose Transporters in Brain Disease: Diabetes and Alzheimer’s Disease

    “Several studies done on AD patients [186–188] and rodent models of AD [189,190] support the role of dysfunctional insulin signaling in the pathogenesis of AD [180]. Liu et al. [180] found reduced levels and activity of many insulin-PI3K-AKT signaling pathway components which negatively correlated with the phosphorylation of tau. Also reduced insulin-PI3K-AKT levels positively correlated with reduced O-GlcNAcylation of protein. This suggests that through the downregulation of O-GlcNAcylation and/or the promotion of abnormal hyperphosphorylation of tau, neurodegeneration is a consequence of the impaired insulin-PI3K-AKT signaling pathway. The role of insulin/insulin resistance and its influence on impaired brain glucose transport and/or metabolism still remains to be established.”

    The inhibition of the phosphatidylinositol-3 kinase reduces neurogenesis, limits the flow of blood in the brain, reduces the transport of glucose in the brain (which leads to further oxidative stress and to the depletion of energy), and contributes to cell death. The disruption of this pathway is also implicated in Parkinson’s disease.

    Increased glucose levels in the brain contribute to Alzheimer’s disease as does the latter decline in glucose transport. Whereas being diabetic increases the risk for Alzheimer’s disease, the observation has been made that some diabetics progress more slowly during the disease as at least some of the excess glucose enters brain cells.

    “Predictors of progression of cognitive decline in Alzheimer’s disease: the role of vascular and sociodemographic factors”

    “Our findings suggest a slower disease progression in Alzheimer’s patients with diabetes. If confirmed, this result will contribute new insights into Alzheimer’s disease pathogenesis and lead to relevant suggestions for disease treatment.”

  15. Ana Aitawa says:

    We have published a research study on global deep brain stimulation devices used for Parkinson’s disease. The analysis estimates the global market to reach nearly US$ 135 Million by 2021, growing at 7.2% CAGR over 2017-2021 forecast period. For more details, visit

  16. Richard Lyon says:

    Re: p= 0.05, the authors’ comment “the size of the effect” is the crux of it. (Obviously, a larger population can statistically see even a very small effect at < 0.01 or less.)

    Is the effect large enough to be beneficial? If so, repeating the trial and or a a larger (more costly, population) study might be warranted.

    The phi-coefficient can help measure the efficacy of the effect and guide the size of N down the road.

    I see no benefit to drawing the "line" a p = 0.005. (my two cents)

Comments are closed.