Neurologic Disease Diagnostics: Special Report

Volume 8, Issue 24 | December 14, 2023

and

Dec 15, 2023

In This Issue

Distinguishing bipolar from depression with a finger stick
Pharmacogenomics helps schizophrenia patients
Using speech patterns to diagnose schizophrenia
Wearables improve Parkinson’s monitoring

Editors’ Note

Last week, we focused on Alzheimer’s disease. This week we branch out to other neurological conditions, where the landscape of diagnosis and monitoring is changing rapidly.

From both of us: Happy Hanukkah, Merry Christmas, Happy Kwanzaa, and best of wishes for whatever end-of-year holiday you celebrate. We hope you're taking some time to relax and enjoy yourself. We'll be off for the next two weeks, hopefully doing the same.

New and Noteworthy

A blood test to distinguish bipolar disorder from depression - better, not perfect

Psychiatric disorders in general are hard to diagnose. It’s particularly difficult to differentiate transiently depressed bipolar (BPD) patients from those suffering major depressive disorder (MDD), the latter being twice as common. As a result, BPD patients are often misdiagnosed with MDD.

A recent JAMA Psychiatry paper reports that a finger-prick blood-spot test of 17 metabolites (ceramide being the most important) was able to correctly recategorize 30% more MDD patients to BPD than standard clinical evaluation alone. However, overall diagnostic accuracy was still only moderate (68%), largely because of a high rate of false positives; positive predictive value was just 52%. Negative predictive value was higher: 80% of the time, the test was able to rule out BPD in depressed patients.

Commentary: Finding reliable biomarkers for psychiatric disorders is a major priority with many benefits: They would enable more accurate treatment for patients, give patients and physicians more confidence in diagnosis, and provide researchers with novel pathways for therapeutic development.

Pharmacogenomics speeds improved outcomes for schizophrenia patients

Almost all patients with schizophrenia benefit from antipsychotics. However, drug effectiveness and side effects vary greatly across patients (see a 2019 comprehensive review). That variation can come down to genetics.

A recent JAMA paper looked at whether using pharmacogenomics (PGx; knowing how someone’s genetic profile will affect the way they respond to a given medication or combination of medications) led to better schizophrenia treatment than standard care alone. The results were promising: 17.4% more patients achieved symptom remission after 12 weeks of PGx-guided treatment (62.8% with PGx versus 45.4% without it).

Commentary: The study did not detail what exact therapy or dosing decisions were most commonly involved in cases with improved outcomes. Cytochrome P450-driven metabolism differences were responsible for a very big part of the effect, so it is unclear to what extent a more general PGx test might suffice.

That said, anything that improves or accelerates remission for these patients is critically important. Without PGx, the worst case is all too common: Ineffective therapy with severe side effects (sedation, weight gain, etc.), driving patients to stop meds altogether and drop out of treatment. Physicians must adopt these increasingly effective and cost-effective PGx tests at the beginning of treatment.

Using AI, speech patterns can be used to diagnose schizophrenia

Schizophrenic patients exhibit disorganized event-to-concept reasoning and incoherent word choices (aka “formal thought disorder”). AI large language models (LLM’s) seem like they should be an ideal tool for detecting these disorders, since their very foundation is probabilistic prediction of “normal” word choice and sentence structure, which can then be compared to patient responses to the same questions.

A recent PNAS paper used AI to do exactly that. This particular investigation goes beyond diagnosis to connect various speech irregularities to specific brain regions, which can inform the development of better treatments.

Commentary: The patient-versus-control differences are subtle, and this paper did not examine alternative diagnoses of the speech patterns evaluated. While the results are more proof-of-principle than a diagnostic test, the path forward looks very exciting.

One of the most exciting developments in diagnostics is the rapidly growing use of non-invasive “collateral” symptoms to detect underlying disease (e.g., changes in gait as a diagnostic for Parkinson's), and this paper is a great example of how AI techniques can expand the application of these techniques to new fields.

Food for Thought

Parkinson’s monitoring progressing with wearables and watches

Monitoring the progression of Parkinson’s disease currently relies on three low-tech tools: a neurologist’s eyeballs and brain, plus the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale. The effectiveness of these tools varies with clinician experience and depends on the snapshot of the patient’s symptoms that the clinician got in the exam room.

Tech advances may be starting to address those issues. You’re probably already familiar with the tool that’s being used to solve the “snapshot” problem: It’s the Apple Watch. Since June 2022 the FDA has approved three different apps for the watch that tracks Parkinson’s symptoms: StrivePD (Rune Labs, June 2022), Parky (h2otherapeutics, November 2022), and NeuroRPM (March 2023). All are available by prescription and track instances of tremor and erratic movements that are characteristic of the disease; one (NeuroRPM) can also track slowness of movement. It’s then up to clinicians to analyze the devices’ output.

This fall, a paper in Nature described a tool that aims to solve the clinician-experience problem. In this small study (74 patients), applying machine learning to data from wearable sensors teased out which symptoms were the most characteristic of disease progression. Many of those symptoms involve variability in the way the patient walks - small inconsistencies that are difficult for clinicians to see, but which wearable sensors can pick up.

Quick Hits

Still looking for some last-minute gifts? Mara found just the thing for the genomics nerd in your life. Happy holidays!