NU670 Unit 5 Depression Biotypes Seminar Reflection Example

Four Biotypes, Then None: A Seminar Reflection on Reading a Famous Depression Imaging Study Beside Its Replication

[Student Name]

Purdue University Global

NU670: PMHNP Neuroscience and Psychopharmacology

Unit 5 Seminar Reflection

[Instructor Name]

[Date]

The seminar group and its members are composite; the two studies are real and cited.

What I Brought to the Seminar

I came to our fifth seminar already persuaded. I had read about the study by Drysdale and colleagues in a news summary and again in the paper itself. Using resting-state functional imaging from [1,188] people, the authors reported that patients with depression could be divided into four subtypes, or biotypes, based on patterns of connectivity in limbic and frontostriatal networks. Classifiers built on those patterns identified the subtypes with sensitivity and specificity of [82 to 93] percent in validation and replication samples, and in [154] patients the biotypes predicted response to transcranial magnetic stimulation (Drysdale et al., 2017).

For a future psychiatric nurse practitioner, this was exciting. Depression is diagnosed from symptoms that overlap and vary, and treatment is chosen largely by trial and error. A scan that sorted patients into biologically distinct groups and told us who would respond to a particular treatment would change practice. When our facilitator asked for first impressions, I said that this was the strongest case I had seen for a biomarker in psychiatry.

What Happened in the Discussion

Our facilitator had assigned a second paper, the replication attempt by Dinga and colleagues. They followed the original procedure as closely as they could in a different, more varied sample of [187] people with depression and anxiety. Like the original study, they used a canonical correlation analysis to link connectivity with symptoms and then clustered patients into subtypes. They also added tests the original had not reported: whether the canonical correlations and the clusters were statistically significant, and whether the results were stable when the data were resampled (Dinga et al., 2019).

Their findings were uncomfortable for my opening view. They, too, found extremely high canonical correlations between connectivity and symptoms, and a three-cluster solution looked best. But neither the correlations nor the clusters were statistically significant, and the procedure was unstable across resamples (Dinga et al., 2019).

A classmate, Tomas, put the problem in a way that stayed with me. With thousands of connectivity features and a few hundred patients, a method that searches for the combination most correlated with symptoms will find a strong correlation almost anywhere, even in noise. The high correlation in both papers might therefore be a feature of the method rather than of depression. Our facilitator added a broader point from a well-known analysis of neuroscience studies: when studies are underpowered, the findings that do reach significance tend to overstate the true effect, and fewer of them reflect a real one (Button et al., 2013). A striking number that appears whenever the method is run is not evidence that the brain has four kinds of depression.

I pushed back at first. The original study had a validation sample and an out-of-sample replication, which seemed to answer that concern. Our facilitator asked what exactly had been replicated: the classifier's ability to identify clusters defined by the same procedure, or the existence of separate groups. That question changed my reading. A classifier can learn to recognize clusters very accurately even when the clusters themselves are arbitrary divisions of a continuous spread of patients.

How My View Changed

When the seminar closed I had not concluded that the original study was wrong. The replication used a different, smaller and more mixed sample, and the authors acknowledged the limits of comparison. The idea that depression contains biologically distinct subgroups may still be true. What changed was my view of what the 2017 paper had shown. I now think it demonstrated a method that produces clear-looking subtypes and strong correlations, and a promising but untested link to treatment response in one group of patients. It did not show that the subtypes are stable features of people with depression.

I also noticed something about myself. I had accepted the first paper largely because its numbers were high and its conclusion was what I hoped to hear. I had not asked the basic questions I would ask of a drug trial: was the key test significant, and would the result survive in a different sample?

What Survived and What Did Not

Survived: the problem the study addressed. Depression is heterogeneous, and imaging may one day help separate patients who respond to different treatments. The use of large, multisite samples is also a lasting contribution.

Did not survive my reading: the claim that there are four biotypes, the confidence that they can be identified reliably from a scan, and the suggestion that they are ready to guide treatment. None of these has been independently confirmed.

What I Will Do Differently

1. When I read a study that claims to find subtypes, I will look for tests of whether the clusters are statistically distinct and stable, not only whether a classifier can recognize them.

2. When a result is very striking, especially in a field without established biomarkers, I will look for a replication before I describe it to a patient or a colleague as settled.

3. When a patient asks whether a brain scan can tell which treatment will work for them, I will say that researchers are working on it, that no scan is used to choose depression treatment today, and that their response to treatment will guide us instead.

The seminar did not make me distrust neuroscience. It made me read it the way I would read any other evidence, with the same questions whatever the technology.

References

Button, K. S., Ioannidis, J. P. A., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S. J., & Munafò, M. R. (2013). Power failure: Why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience, 14(5), 365-376. https://doi.org/10.1038/nrn3475

Dinga, R., Schmaal, L., Penninx, B. W. J. H., van Tol, M. J., Veltman, D. J., van Velzen, L., Mennes, M., van der Wee, N. J. A., & Marquand, A. F. (2019). Evaluating the evidence for biotypes of depression: Methodological replication and extension of Drysdale et al. (2017). NeuroImage: Clinical, 22, Article 101796. https://doi.org/10.1016/j.nicl.2019.101796

Drysdale, A. T., Grosenick, L., Downar, J., Dunlop, K., Mansouri, F., Meng, Y., Fetcho, R. N., Zebley, B., Oathes, D. J., Etkin, A., Schatzberg, A. F., Sudheimer, K., Keller, J., Mayberg, H. S., Gunning, F. M., Alexopoulos, G. S., Fox, M. D., Pascual-Leone, A., Voss, H. U., . . . Liston, C. (2017). Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nature Medicine, 23(1), 28-38. https://doi.org/10.1038/nm.4246