For decades, a puzzling disconnect has frustrated both patients and clinicians: people with chronic insomnia often describe their sleep as fragmented, shallow, and unrefreshing, yet their overnight sleep studies can come back completely normal. The standard polysomnography (PSG) report shows normal sleep latency, normal sleep efficiency, normal time spent in each stage — nothing to explain why the patient feels so terrible.
A new study published in Sleep Medicine may finally resolve this paradox. Researchers led by Dr. Umaer Hanif at Universite Paris Cite and Hotel-Dieu Hospital in Paris, in collaboration with the Danish Center for Sleep Medicine in Copenhagen, examined the sleep records of over 1,000 participants and found that the key to the discrepancy lies not in the broad architecture of sleep but in its finer details — specifically, the microstructure of REM sleep.
What they found
The study analyzed PSG data from 1,049 individuals: 141 healthy good sleepers, 214 chronic insomnia patients with normal PSG macrostructural measures (classified as CIN-NP), and 694 chronic insomnia patients with detectable PSG abnormalities (CIN-AP). Using automated algorithms, the researchers quantified not just standard sleep stages but also REM-specific features including arousals, awakenings, REM density (rapid eye movements per hour), and the proportion of phasic versus tonic REM sleep.
The results were striking. Both insomnia subtypes — even those with completely normal PSG reports — showed significantly more REM arousals and awakenings per hour, higher REM density, and a greater proportion of phasic REM compared to good sleepers. These microstructural abnormalities were present in CIN-NP patients despite the absence of any abnormality on conventional sleep staging.
This finding offers a direct explanation for the subjective-objective sleep discrepancy that has long puzzled sleep medicine. Patients with normal-looking sleep studies are not imagining their poor sleep. Standard PSG simply does not capture the REM fragmentation that is happening at a finer timescale — the frequent micro-arousals, the bursts of phasic activity, the excessive eye movement density that disrupts the restorative quality of REM sleep even when total REM time appears normal.
The two insomnia subtypes, however, were not identical. CIN-AP patients had significantly longer REM awakening durations than CIN-NP patients — their awakenings lasted longer even though the frequency of arousals was similar. Using multinomial logistic regression, the researchers identified six independent predictors that distinguished CIN-AP from CIN-NP: male sex, older age, lower total REM duration, higher REM percentage relative to total sleep, increased phasic REM proportion, and longer REM awakening durations. A predictive model combining these factors achieved an AUC of 0.752 (65.6 percent sensitivity, 72.3 percent specificity).
The study also examined arousal dynamics using odds ratio product (ORP) analysis, a measure of how prone the brain is to wake up from any given sleep depth. CIN-NP patients spent more time in deeper sleep ORP deciles (deciles 2 through 4) but experienced more brief awakenings of 15 to 30 seconds. CIN-AP patients, by contrast, spent more time in lighter sleep deciles (9 and 10) and had more prolonged awakenings lasting 1 to 2 minutes, 5 to 10 minutes, and even 10 to 20 minutes. Adding ORP variables and awakening duration distributions to the predictive model yielded a marginal improvement to AUC 0.755 (68.4 percent sensitivity, 69.7 percent specificity).
Why it matters
These findings have several clinical implications. First, they provide a physiological basis for what insomnia patients have long reported: their sleep feels disturbed even when the standard sleep study says otherwise. Recognizing REM microstructure as a source of this discrepancy could change how clinicians interpret PSG results and how they communicate with patients whose tests come back negative.
Second, the study suggests that REM microstructural features may serve as objective biomarkers for distinguishing insomnia phenotypes. This matters because insomnia is a heterogeneous disorder — what looks like the same condition on the surface may have different underlying physiology. Patients with the CIN-AP profile (longer REM awakenings, lighter sleep, older age, male sex) may represent a subgroup at higher cardiovascular and metabolic risk. The researchers note that this group carried the highest burden of cardiometabolic comorbidity, potentially linking REM fragmentation to broader health outcomes.
Third, the ability to differentiate subtypes based on REM features opens the door to phenotype-specific treatment. A patient whose insomnia is driven by frequent but brief REM micro-arousals may respond differently than one whose sleep is dominated by prolonged awakenings and lighter overall sleep. Current first-line treatments like cognitive behavioral therapy for insomnia (CBT-I) may work across subtypes, but drug therapies that target arousal thresholds or REM stability could eventually be matched to specific profiles.
Limits
As a retrospective analysis, the study can identify associations but cannot establish causality — whether REM microstructural abnormalities cause insomnia or are a consequence of it remains unclear. The automated algorithms, while standardized, may not capture all clinically relevant features. Additionally, the two insomnia subtypes were defined by the presence or absence of PSG macrostructural abnormalities using established criteria, but other classification schemes might produce different groupings. The study also did not control for medication use, which could influence REM architecture.
Bottom line
Standard sleep studies may give a clean bill of health to insomnia patients whose sleep is, in fact, measurably fragmented at the microstructural level. REM sleep features — arousals, awakenings, density, and phasic activity — appear to capture clinically meaningful aspects of insomnia that routine PSG misses. For patients who have been told their sleep is normal when it clearly does not feel that way, this study offers both validation and a path toward more precise diagnosis.
Source
Hanif U, Crosbie F, Aloulou A, Romano F, Sauvet F, Jennum P, Solelhac G, Leger D. REM sleep fragmentation and arousability distinguish chronic insomnia subtypes and reflect subjective-objective sleep discrepancy. Sleep Medicine. 2026;147:109165. DOI: 10.1016/j.sleep.2026.109165. PMID: 42503244.