One day in 7^th grade, my science teacher started a lesson by walking around the room with a big bag of old, cancelled postage stamps and dumped a large handful on each desk.  The assignment was to sort them into groups using any method we liked.  There were many features to choose from, especially if combined to produce finer groupings.  (That’s what obnoxious, smarty-pants Larry did.)  The point was to demonstrate that living things can be classified in many ways, too.  Now hold that thought.

We once thought all PSP was pretty much alike, with no more variation than any other neurological disease (“disease” being defined as a common autopsy or biochemical picture, typically with a common causality, if one is known).  But in 2005, a group at University College London led by David Williams and Andrew Lees reviewed the clinical records of all 103 patients in their files with autopsy-proven PSP.  For each, they tabulated a long list of clinical features and, as I did with my pile of stamps, created groups using combinations of features. 

They found that 54% conformed more or less to the original combination of features (called a “syndrome”) described by Steele, Richardson and Olszewski in 1963 and 1964, where the first and worst symptom was poor balance with falls, also with symmetric motor signs, prominent cognitive loss, poor response to levodopa, little or no tremor and rapid progression.  They called this combination PSP-Richardson’s syndrome. 

Another 32% had a different picture for the first few years, with general slowness and stiffness as the initial deficits, asymmetric feature, little cognitive loss, a useful response to levodopa, moderate tremor, and slower progression.  They recognized this type as similar in many ways to Parkinson’s disease and dubbed it PSP-Parkinsonism.  The other 14% of the patients of Williams et al didn’t conform well to either PSP-RS or PSP-P.  

The basic picture at autopsy for PSP-RS and PSP-P was identical, though subsequently, as one would expect, the tau aggregation of PSP-RS would be found to emphasize the brainstem, while that of PSP-P emphasizes the basal ganglia.  Williams et al found that the MAPT H1/H2 ratio (the most important genetic risk factor for PSP) and the tau 4R/3R ratio (a feature of the structure of the tau protein in the neurofibrillary tangles) were each higher in PSP-RS than in PSP-P, but I haven’t seen confirmation of this since the original 2005 paper.

This PSP-RS vs PSP-P differentiation by Williams et al rested on the results of a statistical procedure called “principal component analysis,” which tabulated which of a list of common PSP features tend to occur in the same patients.  It’s what I was doing in my head with the stamps in 7th grade, but in a much less sophisticated way.

Over the next decade, a variety of other PSP types were found to account for the last 14% of Williams et al.  Like PSP-RS and PSP-P, they all had the same set of autopsy abnormalities with minor differences in the areas of the brain involved corresponding to their specific, predominant symptoms.  However, their definitions relied only on a single feature occurring first and worst rather than on a more complex analysis of a long list of features as a principal component analysis would.  So, we can’t be sure that they represent biologically relevant differences that might, for example, be susceptible to different kinds of diagnostic markers or neuroprotective treatments.

A first step toward resolving that issue has now come from a group of researchers mostly in London, Cambridge and San Francisco led by William J. Scotton of University College London, with senior author Peter A. Wijeratne.  They analyzed existing MRI images in a group of 426 living patients with a variety of PSP subtypes and 290 control individuals without PSP.  They divided the PSP types into 3 categories:

1. PSP-Richardson’s syndrome (PSP-RS) (84% of the total)
2. A “cortical” group comprising PSP-behavioral variant frontotemporal dementia (PSP-F), PSP-corticobasal syndrome (PSP-CBS) and PSP-speech/language (PSP-SL) (12%)
3. A “subcortical” group comprising PSP-Parkinsonism (PSP-P) and PSP-primary gait freezing (PSP-PGF) (4%)

(A statistical detail, for those interested: Note that in this study the percentage of all PSP accounted for by PSP-P is much lower than in most surveys, where it’s about 30%.  This is explained by a new way of assigning a sub-type using a statistical approach called “multiple allocation extinction rules,” which helps avoid the frequent problem of individual patients satisfying criteria for multiple subtypes.) 

(Now a clinical detail, for those interested: In most referral centers, the fractions of PSP accounted for by these sub-types are roughly: PSP-RS 50%, PSP-P 30%, PSP-PGF 5%, PSP-CBS 4%, PSP-F 4%, PSP-SL 3%.  That makes 96%.  Four others not included in the Scotton et al series, each at about 1%, are PSP-cerebellar (PSP-C), PSP-primary lateral sclerosis (PSP-PLS), PSP-ocular motor (PSP-OM) and PSP-postural instability (PSP-PI).  In Japan, PSP-C is far more common for some reason: about 10-15% of all PSP.)

The result was that MRI in the cortical subtype showed atrophy starting in the:

1. frontal lobes and
2. insula (the surface of cortex on the side of the brain hidden by the temporal lobe), and in the brainstem, which of course is a subcortical area. 

The subcortical subtype’s atrophy was most prominent in the:

1. brainstem,
2. ventral diencephalon (the area of cerebrum just above the brainstem),
3. superior cerebellar peduncles (fiber tracts carrying most of the output of the cerebellum to the brainstem and cerebrum), and the
4. dentate nucleus (the part of the cerebellum where the fibers of the superior cerebellar peduncle originate, so called because its zig-zag shape resembles a row of teeth).

Here are some of their additional observations:

1. For both the subcortical and cortical patients, 82% conformed to the MRI pattern described above. 
2. The subcortical subtype had worse PSP Rating Scale scores after potential confounders were accounted for.   
3. The subtypes held up over a period of years in the patients in whom multiple successive exams were available, but the pattern of atrophy at the end stage was similar for the cortical and subcortical subtypes.
4. The PSPRS subtype behaved in these respects almost exactly like the subcortical subtype except that it progressed faster, on average.

What does this mean?  As I sometimes do in this blog (probably not often enough), I’ll let the authors speak for themselves:

“The results suggest that the PSP-RS and PSP–subcortical syndromes share a similar trajectory of atrophy, though the latter tends to be at an early stage at diagnosis and progresses at a slower rate. Being able to accurately subtype and stage PSP patients at baseline has important implications for screening patients on entry into clinical trials, as well as for tracking disease progression.”

A major issue right now for clinical trial design for PSP is how to include the non-PSP-RS subtypes.  The PSP Rating Scale, still the world’s standard primary outcome measure for trials, was designed for what would a decade later be named the PSP-RS subtype.  For that reason, and because the diagnostic criteria for non-RS sub-types aren’t as accurate, PSP treatment trials have excluded non-RS subtypes.  But by tracking how the PSP Rating Scale progresses in the other sub-types, the statistical analysis of the trials’ data could be adapted to include those patients.   Another conclusion might be that we should design trials to include the two subcortical sub-types along with PSP-RS, as all three have a similar pattern of progression, albeit at different rates.  Of course, that would throw the cortical subtypes under the bus, awaiting development of their own trial outcome measure.

So, just as postage stamps can be classified in different ways, so can PSP.  Understanding all the resulting sub-types, if they’re based on validating factors like patterns of atrophy on MRI, allows potential PSP preventatives to be tested more democratically across the PSP population.  It also eases patient recruitment into clinical trials, speeding their completion and reducing their cost.