Marker development: anarchy vs plutocracy?

You’ve heard me whining that we need a diagnostic “trait marker” for PSP. In other words, we need to be able to accurately distinguish PSP – during life — from such mimics as Parkinson’s, multiple system atrophy, Alzheimer’s, corticobasal degeneration, normal-pressure hydrocephalus and others. Only in that way can we create “pure” groups of patients in which to study the disease and test specific treatments.
Right now, the best diagnostic test we have is the MDS-PSP Diagnostic Criteria, which requires only traditional history-taking and a hands-on neurological exam. Those criteria work well for PSP-Richardson syndrome after the first couple of years but not quite well enough for earlier-stage PSP-RS nor for the “minority” or “atypical” types, which together account for 60 to 75 percent of PSP.

The most promising markers using laboratory or imaging data are levels of phosphorylated tau and neurofilament light chain (NfL) in the spinal fluid and blood; and perhaps MRI measurements of the size of the midbrain, pons and related areas at the base of the brain. But these are far from ready for prime time. NfL is a protein component of brain cells that has been shown to occur at about a two-fold higher level, and to increase faster over time, in PSP and MSA than in PD, Alzheimer’s and other neurodegenerative diseases. MRI changes don’t occur in the early stages. Positron emission tomography is coming along, but won’t be ready for use in PSP for another few years, and even then will be costly and not widely available.

Last week, my routine surveillance of new PSP-related research papers in the literature yielded two interesting hits — both about PSP trait markers, both using new lab techniques, and both from Italy.

Corinne Quadalti and colleagues at the University of Bologna measured NfL and alpha-synuclein in spinal fluid and blood. They found that plasma NfL alone worked very well in distinguishing PD from PSP, with an accuracy of 0.94. (“Accuracy” in this context is the area under the receiver operating characteristic curve, which compares sensitivity with specificity. Perfect accuracy is 1.0 and a useless test’s accuracy is 0.5, where a coin flip would work as well.)

Alpha-synuclein is the main protein aggregating in Parkinson’s, dementia with Lewy bodies and multiple system atrophy. It is to those diseases what tau is to PSP and CBD. To measure it, they used a new technique called “real-time quaking-induced conversion” (RT-QuIC; pronounced, “R-T quick”), which measures that protein in its misfolded and aggregated forms. This prevents that abundant protein in its normal form from swamping the measurement. The result was positive in 91% of their patients with PD and in none of their 58 patients with PSP or CBD.

Now, if you have a nose for statistics, you’ll raise your hand and say, “But those 9% of PD patients with a negative test comprise more people in the general population than all the patients with PSP or CBD, so a negative test doesn’t mean much.” and you’d be right. So, while the sensitivity of the test for PD is excellent, the specificity is low, rendering the overall accuracy in a real-world situation insufficient.

For that reason, the authors combined two measurements – spinal fluid NfL and serum alpha-synuclein, with a resulting improvement in distinguishing PD from PSP/CBD to a sensitivity of 97.4% and specificity of 100%. That’s more like it, but keep in mind a few issues: They combined PSP and CBD into one group, and we don’t know if the results apply as well to each disease alone. They had no autopsy confirmation of the diagnoses, which means that these patients were already at a stage that was possible to diagnose using traditional clinical criteria; this means that patients with earlier-stage illness will be needed in a follow-up study. Finally, and as always, the results have to be confirmed at other centers using other techniques.

The other eye-catching paper was from Ida Manna and colleagues at the University Magna Graecia in Catanzaro, Italy. They use exosomal micro-RNA (miRNA) in blood to distinguish among PSP, PD and healthy controls.

Exosomes are tiny bubbles of brain cell membrane enclosing whatever cell contents were there when the bubble pinched itself off and floated free. They often find their way into the bloodstream. MicroRNAs are stretches of RNA averaging only about 22 nucleotides. They do not encode proteins as messenger RNA (mRNA) does, but instead bind to mRNAs to regulate their translation into protein. They are specifically encoded in the DNA of the genome and about 2,000 of them are known to exist.

Dr. Manna et al measured levels of 188 miRNAs for which there is evidence of association with some neurodegenerative disease. They found a set of 6 miRNAs that together yielded an accuracy in distinguishing PSP from PD of 0.91. The accuracy for distinguishing PSP from controls was 0.90.
Of course, many of the same caveats that I listed for the other paper apply to this one. Plus, PSP mimics other than PD were not included in the analysis. Just as important is that there were only 25 patients with PSP and they were a mixed bag of 20 with PSP-Richardson and 5 with PSP-Parkinson. In applying a marker for the purpose of excluding patients with PD from a study of PSP, it is critical to be able to distinguish PD from PSP-P. It is unlikely that those 5 patients with PSP-P constituted a statistically valid sample for that purpose. That will be a project for another day.

What do I take away from these two papers? Neither of them alone provides a marker just yet, and each has its drawbacks given the current early stage of work. But perhaps, with some refinement, combining them with other non-invasive markers could create a diagnostic panel with enough accuracy to distinguish PSP from all of its mimics. After all, in medicine in general, multiple diagnostic tests (several tests of body fluids, some imaging, a physiologic test such as an EKG) must be combined to produce an actionable diagnosis. Why should PSP be any different?

I think the problem (and it’s a good problem to have) is that new candidate markers are being identified all the time, as are ever more sophisticated technology for measuring them, with RT-QuIC, miRNA and exosomes as prime examples. That means as researchers turn their attention to early-phase development of newer ideas using newer technology, ideas that looked potentially useful if pursued further may be neglected and not developed into practical tests. What to do? Do we just let scientific nature take its course in its traditional, anarchical way, waiting for research groups to take techniques with good initial data to the next level? Or should a group of experts with an iron fist issue some sort of “white paper” listing which markers with good preliminary evidence, perhaps like the ones I describe here, should be nurtured with funding and collaborations? If so, who chooses those experts? And once the experts are chosen, how can we prevent them from favoring the ideas in which they’ve invested their own time, resources and reputations?

You know where the “comment” button is.

Markers: the longitudinal approach

We got plenty of candidate PSP treatments.

We got drug companies willing to risk their resources on trials for a rare disease.

We got clinical trial sites with proven records of efficiency. 

We got patients willing to make the sacrifices demanded by clinical trials. 

What ain’t we got? 

We ain’t got markers. 

(Deepest apologies to Rodgers and Hammerstein.)

Markers in this context are simply diagnostic tests, and there are two kinds – trait markers and state markers.  Trait markers allow us to distinguish people with from those without the disease, preferably in a very early stage, where treatments designed to prevent further decline would be most likely to occur and most useful to the patient.  Trait markers also allow us to exclude from a PSP trial any people who don’t actually have PSP.  State markers, on the other hand, quantify the amount of damage that’s already occurred and the degree of benefit of the experimental treatment. 

The best trait marker for PSP to date is purely clinical, meaning that it does not require any sort of imaging, automated measurements of movement, gene testing or chemical testing of body fluids.  That’s the MDS-PSP Criteria, published in 2017. Other types of tests can help exclude from consideration other conditions such as Alzheimer’s, Parkinson’s, MSA, normal-pressure hydrocephalus and vascular parkinsonism, but they are only helpful in those cases where a specific alternative diagnosis is plausible.  They don’t positively diagnose PSP; they only rule out other things.

Two up-and-coming trait markers for PSP are spinal fluid levels of tau with a phosphate group on amino acid 181 (Ptau181) and neurofilament light chain (NfL).  Ptau181 levels are below normal, on average, in all forms of PSP except for the gait-freezing type (PSP-PGF).  This contrasts with Alzheimer’s disease, where that marker is elevated, on average.  The average level of neurofilament light chain (NfL) in the spinal fluid is much higher in PSP and CBS than in controls or Alzheimer’s but is also elevated in many other neurodegenerative disorders.  So the ratio of NfL divided by Ptau181 in the spinal fluid is an good marker for PSP, but cannot distinguish it from CBD, and for PSP-Richardson, it may not be as accurate as the bedside clinical criteria. For ordinary clinical use, a blood test would be easier than a spinal tap, and the utility of these levels as a state marker has not been adequately studied, even for CSF.  That requires a longitudinal study over a period of at least a year.  So the NfL/Ptau181 ratio isn’t ready for prime time as a PSP trait marker, much less as a state marker.

The most widely used state marker for PSP is still the PSP Rating Scale, which is also purely clinical. (Disclaimer: I developed the PSPRS starting in 1995 and published it in 2005 along with my statistician colleague Pam Ohman-Strickland.)   It takes 15 minutes to administer and requires no equipment other than an armless chair, a cup of water to test swallowing — and the apparatus between the neurologist’s ears.  In recent years, modifications of the PSPRS have been shorter, easier to administer by laypersons, or more directly reflective of the patient’s daily activities.  Although all of these revisions are valid and have been shown to correlate well with the full, original PSPRS, none has been widely tested in the field, and the PSPRS remains the standard for now.  But it’s not good enough.  Its score is affected by common non-PSP conditions such as injuries, arthritis or strokes, or by PSP-related conditions; for example, orientation testing can be affected by apathy, gait testing by muscle rigidity, blepharospasm by Botox and everything by dehydration or malnutrition.  So there’s a lot of variance in the PSPRS as measured from one visit to the next.  This dictates that trials be large enough and long enough to cancel out the “statistical noise,” and that costs money.

A longitudinal study is observational – it includes no treatment.  It enrolls patients with the disease of interest, or sometimes also healthy people with histories suggesting a high risk of developing that disease.  Many longitudinal trials also enroll control subjects with no apparent risk for the disease — typically spouses, relatives or friends of those in the first two groups.  All of the subjects undergo tests at entry using whatever diagnostic procedures are being evaluated as markers, some of which are repeated periodically.  The study follows the patients through their course, at least with interim histories and physical exams.  If feasible and appropriate, autopsies are obtained to verify the diagnosis and to correlate specific autopsy features with diagnostic test results during life.  The goal is to identify which, if any, of the diagnostic tests prove able to accurately identify people with the disease in the earliest stages and which can track their subsequent course with precision.

There are presently at least 8 PSP longitudinal studies in progress: 2 in Germany and 1 each in India, Italy, Japan, Luxembourg, the US/Canada and the UK.

At the PSP Study Group meeting on October 4, James Rowe of Cambridge updated the group on the longitudinal PROSPECT-M-UK study, which is headed by Huw Morris of University College London. (“M” is for MSA, a late addition.) It now includes 21 academic clinic sites in the UK and about 700 patients, of whom about 100 have made more than the initial visit.  They have found that using MRI measures of atrophy of certain regions of the cerebrum is more precise than the PSPRS, reducing the number of patients needed for a treatment trial by nearly half.  The measures were atrophy of frontal and temporal lobes and enlargement of the lateral ventricles, an indirect sign of diffuse cerebral atrophy.  This confirms and extends the findings in the two trials of monoclonal antibodies that failed to help PSP, where MRI at the start and end of the studies provided a sharper picture of the patients’ progression than the PSPRS.  The reduction of the sample size was even more marked for CBD, but in fairness, the PSPRS was not designed for that disease.  One of the PROSPECT-M-UK study’s specimen collections is skin biopsies.  These can be used to look for tau aggregation in nerve endings, a potential early-stage, only slightly invasive trait marker.  Skin biopsies can also be used to create stem cells, which are then converted into neuronal cultures in which experimental treatments can be tested.  In this case, each such “brain in a dish” will come with a detailed, standardized clinical record.  Even more important, that lab model is not a mouse with a PSP-like condition, but a human being with real PSP.

LK Prashanth of Vikram Hospital in Bangalore described the longitudinal PSP study being conducted by the Parkinson Research Alliance of India. The Pan-India Registry for PSP (PAIR-PSP) includes 15 centers with 68 patients, with a goal of 1,000 over the next 2 years.  They are performing whole genome sequencing along with more conventional measures.  They have found that PSP-Richardson syndrome, the classic form, exists in only 25% of their group.  Next is PSP-parkinsonism with 22% and PSP-CBS with 18%.

Martin Klietz of Hannover Medical School updated the group on the two German studies, DESCRIBE and ProPSP.  The first has enrolled 400 patients, the second, 276.  Each study covers the entire country, although one is based in the north, at Hannover and the other in the south, at Munich.

Rejko Krüger of the University of Luxembourg mentioned that his institution’s longitudinal Parkinsonism study, which recruits from that small country as well as nearby areas of France, Germany and Belgium, has recruited 80 patients to date, and is collecting skin biopsies and spinal fluid in addition to the usual imaging and clinical markers.

Takeshi Ikeuchi of Niigata University, Japan, described the Japanese Longitudinal Biomarker Study in PSP and CBD (JALPAC).  It has accumulated 337 patients with at least one visit, of whom 257 have had at least two.  They found PSP-Richardson in a slightly higher percentage, 35%, than did the study in India.  They found a good correlation of the PSPRS with disease duration but, as expected, wide range of velocities of progression across patients. 

No one at the meeting provided an update on the US/Canada study, which focuses not specifically on PSP or CBD, but on a much more inclusive disease category called frontotemporal dementia (FTD).  PSP and CBD are often classified within the category of the FTD’s because they usually feature dementia of frontal lobe origin.  The protein aggregating in the brain cells is different in the various FTD diseases – tau, TDP-43 and FUS are the most common.  The study, called “ALL-FTD,” is headed by Brad Boeve at the Mayo Clinic Rochester and Adam Boxer and Howard Rosen at UCSF.  It presently includes 21 sites in the US and 2 in Canada. The longitudinal arm has a goal of 1,100 patients and the biofluid-focused arm, with just one visit apiece, aims for 1,000 patients.  I’ll let you know about current PSP enrollment once I can squeeze that out of someone, but for more info, try their website.

Gabor Kovacs of the University of Toronto described a project based in Japan to study “incidental PSP.”  This is early brain changes of PSP that had not yet started to cause symptoms by the time of death.  It is found in specimens donated by families whose loved one died without known neurological illness.  One such collection, at Banner Health in Arizona, found very mild PSP pathology in 5% of their autopsied brains.  This means that 5% of the elderly population may be incubating PSP.  Of course “may” is the critical word, but analyzing the medical, genetic, and toxin exposure backgrounds of such a large group of people, even in retrospect, could provide valuable clues to the cause of PSP.