A single electrode placed at the C3 location on the scalp during a routine sleep study may be enough to distinguish untreated schizophrenia from healthy controls, according to a retrospective pilot study from Kurume University in Japan. In 7 antipsychotic-naive patients and 7 matched controls, fast spindle frequency and detection time at C3 achieved an area under the curve of 0.939. The sensitivity was 85.7% and specificity was 100%, meaning no healthy control was misclassified.
The finding rests on decades of converging evidence that sleep spindles, brief bursts of 12-15 Hz oscillatory activity that emerge during non-REM stage 2 sleep, are generated by thalamocortical circuits. These same circuits are known to be disrupted in schizophrenia, and multiple studies over the past 20 years have reported reduced spindle activity in patients. What distinguishes the present study is its use of standard clinical polysomnography rather than high-density research-grade EEG, and its focus on patients who had never received antipsychotic medication.
What they found
The team, led by Satoshi Mizuki, analyzed PSG data collected at Kurume University Hospital between 2001 and 2024 from 7 patients diagnosed with schizophrenia (DSM-IV criteria, mean age 27.6, 4 male) who were antipsychotic-naive, and 7 healthy controls (mean age 31.8, 3 male) from prior studies conducted at the same laboratory.
Spindles were detected using the Complex Demodulation method and classified as total (9-15 Hz), fast (12-15 Hz), and slow (9-12 Hz). The key results centered on the C3 electrode (left central region).
| Metric | Controls | Schizophrenia | p-value |
|——–|———-|—————|———|
| Total spindle frequency at C3 (events/h) | 112.96 | 45.2 | 0.0041 |
| Total spindle detection time at C3 (s) | 835.8 | 289.0 | 0.0041 |
| Fast spindle frequency at C3 (events/h) | 98.25 | 35.69 | 0.0041 |
| Sleep efficiency (%) | 91.3 | 77.9 | 0.0239 |
| Deep sleep (Stage 3+4, min) | 77.0 | 45.2 | 0.0313 |
| REM latency (min) | 109.2 | 60.8 | 0.0028 |
The ROC analysis for fast spindle frequency and detection time at C3 both yielded AUC 0.939 with Youden-index selected cutoffs producing 100% specificity and 85.7% sensitivity. Slow spindles (9-12 Hz) did not differ significantly in frequency or detection time, though their duration and amplitude were reduced.
The study also examined the acute effect of a single 2 mg dose of risperidone. After administration, the number of awakenings decreased significantly, stage 2 sleep increased, and REM latency lengthened. Spindle measures showed small changes that were statistically significant (fast spindle frequency and detection time both p = 0.0313) but of unclear clinical significance given the wide individual variability.
Why it matters
Schizophrenia affects roughly 0.7% of the global population, but no objective physiological biomarker currently exists for diagnosis. The disorder is diagnosed entirely through clinical interview and symptom assessment. Sleep spindle deficits have been proposed as a candidate endophenotype because they are present in first-episode drug-naive patients and in non-psychotic first-degree relatives, and appear minimally influenced by medication.
The present study extends this line of evidence in two ways. First, it shows that spindle abnormalities can be detected using a standard clinical PSG montage, not just research-grade high-density EEG. C3 is a standard electrode in routine sleep studies. Second, it is one of the few studies to measure spindles specifically in an East Asian population, confirming that the spindle deficit generalizes beyond predominantly Caucasian cohorts.
The finding that fast spindles were more affected than slow spindles is consistent with the known topography of thalamocortical dysfunction in schizophrenia. Fast spindles arise from broader cortical networks including the hippocampus and prefrontal cortex, making them more susceptible to the widespread circuit-level disruptions characteristic of the disorder.
Limits
This is an exploratory pilot study with a sample size of only 7 per group. The 95% confidence intervals around the AUC (0.755-1.000) and the 100% specificity estimate are wide, reflecting substantial uncertainty. Multiple comparisons were performed without formal correction. The healthy control data were aggregated from prior studies and may carry systematic bias. PSG did not include frontal derivations or follow AASM standard montage, potentially reducing sensitivity for slow spindle detection. The risperidone effect was assessed after a single administration only, not chronic treatment.
Bottom line
Fast sleep spindle frequency and detection time at the C3 electrode show strong discriminative potential between untreated schizophrenia and healthy controls in this small sample, but validation in large prospective cohorts is required before any clinical application.
Source
Mizuki S, Kotorii N, Hiejima H, et al. Sleep Spindle Measures in Japanese Patients With Drug-Naive Schizophrenia Using Standard Polysomnography Data: A Retrospective Pilot Study. Early Interv Psychiatry. 2026;20(8):e70229. DOI: 10.1111/eip.70229. PMID: 42503890.