This post aims to be a comprehensive review of the clinical trials that have been completed in humans with prion disease.
I’ve previously given a quick summary of all drugs tested in humans with prion disease, whether as part of a clinical trial or just a one-off compassionate use case, and I’ve also written specific histories or critiques of the trials for quinacrine, pentosan polysulfate, and doxycycline. But for months now I’ve had in my go-to slide deck a few slides summarizing what’s been done so far in clinical trials, and it’s time to turn those slides into a post.
FYI for people who’ve come upon this post through a Google search for CJD clinical trials: as of this writing, there are no ongoing therapeutic clinical trials for prion disease / CJD anywhere in the world that people can enroll in. There is only one currently ongoing trial, and it has a small, closed enrollment group (EudraCT 2010-022233-28, testing doxycycline in one family with fatal familial insomnia in Italy).
Summary of trials
The below table summarizes the clinical trials and observational studies that have been completed so far. I have included observational studies where there were at least several patients and some sort of systematic analysis, but I have omitted isolated case reports of just one or a few patients treated under compassionate use - those have been summarized in this post.
|pentosan polysulfate||U.K.||7||no||survival||Bone 2008|
|pentosan polysulfate||Japan||11||no||survival||Tsuboi 2009|
|doxycycline||France & Italy||121||yes||survival||Haik & Marcon 2014|
In all of the above trials, only symptomatic patients were included. In most of the above trials, the majority of patients had a diagnosis of sporadic Creutzfeldt-Jakob disease (sCJD), but all of them included one or a few individuals with variant CJD (vCJD) or genetic prion disease. The U.K. pentosan polysulfate study was unique in that all seven patients had either vCJD or genetic prion disease.
Trial designs and endpoints
Three studies were double-blind, placebo-controlled, randomized clinical trials. The others used a variety of different designs. The U.K. quinacrine trial was conceived as an open label patient preference trial. The French quinacrine study was a large series of compassionate use cases. The U.K. pentosan polysulfate study was not originally conceived as a clinical trial at all - the U.K. government was compelled by a court decision to make pentosan polysulfate available to patients who wanted it (see this history of pentosan polysulfate).
All but one of the studies considered survival time as a primary endpoint. This is possible in prion disease because the median disease course is only 5 months for the most common (sporadic) form of the disease [Pocchiari 2004]. In the randomized trials, the comparison was between survival curves of patients treated with test compound versus placebo. In the case of the French/Italian doxycycline trial specifically, it was calculated that the sample size gave 80% power to detect a two-fold increase in 1-year survival at the p = .05 threshold [Haik & Marcon 2014]. Similarly, the U.S. quinacrine trial was predicted to have 80% power to detect a doubling of mean survival time [Geschwind 2013].
Those randomized studies compared, as a primary outcome measure, the survival of all randomized patients. But these trials included people with sporadic CJD and people with genetic prion disease, people with various sporadic CJD subtypes, and people of both sexes and different ages, all of which are factors associated with significantly different survival time a priori [Pocchiari 2004]. In some cases, these differences are of a considerable magnitude - for instance, median survival for individuals with a codon 129MV genotype is 9 months, more than double the 4 month median for 129MM individuals [Pocchiari 2004]. This observation leads to the insight that additional statistical power might be gained by stratifying patients on these variables. Yet then you have the problem that the number of patients in each stratified subgroup will be very small.
One alternative might be to compare treated patients to historical data, where there are plenty of patients in most possible subgroups. But then you have the problem of bias according to which patients opted to participate in a clinical trial. For instance, early compassionate use studies of quinacrine at UCSF had suggested a benefit, yet the randomized trial saw no benefit [Geschwind 2013], suggesting that perhaps patients with an earlier diagnosis and/or slower progression were more likely to opt to be treated under the compassionate use model. Such is, of course, the value of double-blind, randomized, placebo-controlled trials. Another alternative has been proposed by MRC Prion Unit, where Simon Mead has developed a clinical dementia rating scale tailored to prion disease [Thompson 2013], which measures functional decline and for which codon 129 heterozygosity is apparently the only significant variable for stratification.
The flupirtine trial [Otto 2004] reported that flupirtine, an analgesic, significantly slowed cognitive decline as measured by ADAS-Cog, p = .02. However, in the decade since that time, it appears that flupirtine has never become standard of care for prion disease in the clinic - indeed, the drug is not even mentioned in the pages on CJD treatment by UCSF MAC, NINDS, the U.K. National Health Service or Mayo Clinic. Perhaps other neurologists weren’t convinced by the results reported in that trial - I would be curious to hear any comments as to why.
All of the other trials reported negative results. In some cases, the result was clearly negative [Haik 2004, Tsuboi 2009, Geschwind 2013, Haik & Marcon 2014] while in others there was nominally a positive result but this was attributed to a confounder rather than a genuine effect of the drug [Bone 2008, Collinge 2009].
It is remarkable just how low was the standard of evidence that led to these compounds being tested in clinical trials. Consider the following data:
|compound||effect in cultured cells||effect in mice||years that trials recruited patients|
|flupirtine||not done||not done||1997-2001 [Otto 2004]|
|pentosan polysulfate||effective [Caughey & Raymond 1993]||effective only early in disease course [Doh-Ura 2004]||2003-2004 [Bone 2008], 2004-2007 [Tsuboi 2009]|
|quinacrine||effective [Doh-Ura 2000, Korth 2001]||not effective [Collins 2002]||2001-2002 [Haik 2004], 2001-2006 [Collinge 2009], 2005-2009 [Geschwind 2013]|
|doxycycline||never reported as effective||not convincing [De Luigi 2008]||2007-2012 [Haik & Marcon 2014]|
Some of these drugs had never even been reported to reduce prion accumulation in cultured cells, and none of them had positive data from the most relevant animal models. Indeed, some trials continued, or even began, recruiting patients long after unencouraging data had already been reported from animal experiments. All of these trials took place before we even knew that a compound’s activity against mouse prions doesn’t necessarily predict its activity against human prions. Yet even given what we did know at the time that these trials were ongoing, the negative results of the trials could have been anticipated based upon available preclinical data, as I’ve argued for doxycycline.
Whether or not that means that trials like these should never have been conducted is debatable. Some will point to the monetary costs of these trials and the risk of losing the goodwill and trust of the patient community, while others will point to the rapid mortality coupled with the lack of any alternative treatment options. A positive outcome that I believe is important is that these trials have demonstrated feasibility and have gained us some very valuable institutional knowledge about how to run such a trial.
Indeed, an understanding of these trials is a jumping-off point to consider how we might design future trials. Here are some of the goals I think about as being important to enable better trials in the future:
- Setting standards as to what level of preclinical evidence qualifies a candidate drug to reach clinical trials. Perhaps trials should require efficacy against human prions in humanized mice (with possible exceptions to allow data from mice expressing mouse PrP, if the candidate therapeutic has a mechanism of action for which prion strain specificity would not be expected, or if the trial is for a specific genetic mutation in PRNP that has not been modeled in mice on a HuPrP background). Standards might also require that if a compound is only effective if given before onset of symptoms in mice, it will only be tested in asymptomatic humans.
- Achieving earlier diagnosis of symptomatic prion disease - perhaps using RT-QuIC [McGuire 2012, Orru 2014] - might make it possible to start treatment earlier when it might be more effective.
- Enabling clinical trials in asymptomatic people with PRNP mutations. This will probably require finding a biomarker in these individuals that is predictive of disease onset, and building a patient registry for recruitment.
- For trials in symptomatic patients, it would be great to find a way to gain additional statistical power through stratification, perhaps with the MRC rating scale or something like it, and/or perhaps with historical data if there is a way to use it without introducing ascertainment bias. A rating scale may have the added benefit of measuring quality of life rather than survival per se.