How the STOP-BANG, Berlin, and NoSAS Screening Scores Compare for European Patients

NoSAS score vs STOP-BANG in one answer

NoSAS score vs STOP-BANG comes down to one trade-off. NoSAS discriminates better in a European general population, while STOP-BANG is the safer tool when your priority is missing nobody. In a real-world clinical sleep-lab sample, NoSAS reached an area under the ROC curve (AUC) of 0.78, against 0.71 for STOP-Bang and 0.62 for the Berlin questionnaire (Herschmann et al., Sleep Medicine, 2021). Accuracy and usefulness are not the same thing.

The three tools answer different questions. STOP-Bang is tuned to catch nearly every case and accepts a flood of false alarms. NoSAS is tuned to avoid false alarms and accepts missing some real cases. Berlin sorts you into a high-risk or low-risk category from symptoms alone. For an item-by-item walkthrough of one tool, our free STOP-BANG risk test guide covers its eight questions, and the Berlin questionnaire explainer does the same for its three symptom categories.

What each questionnaire actually asks

The three tools score different things, which is why your results can disagree. The items side by side explain that faster than any accuracy figure.

Those NoSAS point values and its threshold of 8 come straight from the derivation study (Marti-Soler et al., The Lancet Respiratory Medicine, 2016).

Where the three scores came from, and why it matters

Provenance explains the whole comparison, and almost no page mentions it. NoSAS was derived from the HypnoLaus cohort in Lausanne, a Swiss general-population sample of 2,121 adults aged 40 to 85 (Marti-Soler et al., The Lancet Respiratory Medicine, 2016). STOP-Bang was developed in a surgical and pre-anaesthetic setting. Berlin came out of United States primary care.

That history is the mechanism. A tool built for people heading into an operating theatre must not miss anyone, so it is weighted for sensitivity. A tool built on a whole population must not send half a town for testing, so it is weighted for specificity. Neither design is wrong. They were solving different problems.

Two flaws follow from that, and they are everywhere in the guides that currently rank. The dominant calculator pages present STOP-Bang as though it were the only instrument, with no mention of a European-derived alternative. The big consumer sleep pages do note that women may need a lower threshold, then never publish a number you could act on.

In the HypnoLaus derivation cohort and the EPISONO validation cohort, NoSAS reached AUCs of 0.74 and 0.81, ahead of STOP-Bang at 0.67 and Berlin at 0.63 (Marti-Soler et al., The Lancet Respiratory Medicine, 2016). NoSAS was later proposed to French-speaking European general practice as a first-line screening tool for the primary care physician (Praxis, Bern, 2018).

NoSAS score vs STOP-BANG on accuracy, and what a low score really means

Sensitivity is the share of true cases a tool catches. Specificity is the share of healthy people it correctly clears. Pooled across 34 studies, the Berlin questionnaire showed sensitivity 0.80 and specificity 0.48, STOP-BANG sensitivity 0.89 and specificity 0.40, and the Epworth Sleepiness Scale (ESS) sensitivity 0.48 and specificity 0.73 (Sleep & Breathing, 2024).

Read that again. STOP-Bang wrongly flags roughly six in ten healthy people. The Epworth scale misses about half of real cases, which is why a sleepiness score on its own should never reassure you. A separate meta-analysis of 10 studies and 14,510 patients put NoSAS pooled sensitivity at 0.798 and pooled specificity at 0.582, with a summary ROC area of 0.77 (Sleep & Breathing, 2022).

Every figure in that table comes from Herschmann et al., Sleep Medicine, 2021. Here is the translation nobody gives you. A negative predictive value of 0.88 means that out of 100 people who screen negative, about 12 still have moderate-to-severe OSA at an AHI of 15 or above. At 0.92, roughly 8 in 100 do. A low score shifts your odds. It does not clear you.

The neck and BMI trap that breaks all three tools

Both STOP-Bang and NoSAS hinge on a neck circumference of 40 cm. Among 2,108 patients with OSA, the mean neck circumference was 39.4 cm in men but only 34.6 cm in women (Auris Nasus Larynx, 2022). The average woman with confirmed sleep apnoea therefore scores zero on that item in both tools. One repeated design choice becomes the largest single source of false negative sleep apnoea screening in women.

The BMI item compounds it. STOP-Bang awards its point only above BMI 35 kg/m2, while NoSAS awards 3 points at a BMI of 25 to under 30 and 5 points at 30 or above (Marti-Soler et al., The Lancet Respiratory Medicine, 2016). A slim adult with a BMI of 23 scores nothing on the heaviest-weighted item in either tool. Combine a BMI of 23 with a neck under 40 cm, and a woman with habitual snoring, witnessed apnoeas and an AHI of 25 can plausibly land at 2 out of 8 on STOP-Bang and 2 out of 17 on NoSAS.

This is the non-obese lean OSA phenotype, and it is common. In the same Swiss cohort NoSAS was built from, moderate-to-severe sleep-disordered breathing (SDB) at an AHI of 15 or above had a prevalence of 23.4% in women and 49.7% in men, with a median AHI of 6.9 per hour in women against 14.9 in men (Heinzer et al., HypnoLaus study, The Lancet Respiratory Medicine, 2015). Nearly one in four middle-aged and older women in a European general population met that threshold.

Sex-specific cut-offs that exist but are rarely quoted

Consumer pages tend to say that women may need a lower threshold, then supply no number. Published numbers do exist. In 3,129 adults, the optimal STOP-Bang cut-off was 4 for women against 5 for men, and the optimal NoSAS cut-off was 8 for women against 12 for men (Medical Science Monitor, 2025).

Sex-stratified data from 6,606 sleep-lab referrals, 46.2% of them women, showed STOP-Bang reaching an AUC of 0.731 in women against 0.743 in men for moderate-to-severe OSA, and NoSAS 0.699 against 0.704 (Duarte et al., Jornal Brasileiro de Pneumologia, 2020). Discrimination was statistically similar between the sexes. Sensitivity, though, was consistently lower and specificity higher in women. The No-Apnea score at an AHI of 15 or above, for example, reached 80.5% sensitivity in women against 94.0% in men.

The reading is uncomfortable but simple. These tools are not blind in women. They are systematically less likely to fire. Europe's own multicentre reference cohort, the ESADA European Sleep Apnoea Database, has been used to phenotype women with OSA through cluster analysis (Sleep Medicine, 2024), pointing to distinct clinical clusters rather than a milder version of the male picture. Our guide on why sleep apnoea in women is missed goes deeper into that presentation.

The insomnia interaction nobody screens for

Tools that lean on daytime function lose accuracy in people with comorbid insomnia and sleep apnoea (COMISA). Berlin and the Epworth Sleepiness Scale both ask about excessive daytime sleepiness and tiredness, and insomnia distorts those answers. Clinicians commonly report that people who cannot fall asleep at night describe themselves as wired rather than sleepy by day, so they score low.

In that group, the STOP-Bang sensitivity advantage becomes the practical argument for using it. Its questions lean on observed signs and body measurements, with a single item on daytime tiredness. NoSAS asks about no symptoms at all, which makes it immune to this particular distortion and blind to a patient whose only signal is how they feel.

Upper airway resistance syndrome (UARS) sits in a similar blind spot. Effortful, fragmented breathing without full apnoeas can leave the AHI low while the person still feels exhausted. None of the three questionnaires was built to detect it.

A five-step protocol for a low score with real symptoms

This is the question the research pages never answer. Use this sequence tonight.

Why your score has no legal standing in the EU

A questionnaire result carries no legal weight for driving fitness anywhere in the European Union. Only a sleep study does. The EU driving licence Directive 2014/85/EU defines moderate obstructive sleep apnoea syndrome as an apnoea-hypopnoea index between 15 and 29, and severe as an AHI of 30 or more (European Respiratory Journal, 2025).

All 25 of the 27 member states that responded to a 2025 European Respiratory Journal survey, covering around 450 million people, have implemented the directive, 63% of them largely unchanged. Greece and Bulgaria apply a stricter national threshold of an AHI of 5 or above combined with sleepiness. And 37% of member states impose a minimum treatment period of two weeks to two months before a driver returns to the road, which the directive itself does not require. Licence validity for treated OSA is typically three years for Group 1 and one year for Group 2 (European Respiratory Journal, 2025).

What happens after a confirmed mild-to-moderate result

Many people who chase down a false-negative screen end up somewhere awkward. A confirmed AHI between 5 and 29, real snoring, and a treatment conversation that stalls. Continuous positive airway pressure stays first-line for moderate and severe disease, but CPAP therapy adherence is genuinely hard at the milder end, and some patients are offered nothing at all.

Options for this group include positional therapy, a custom-fitted mandibular advancement device, weight management where relevant, and nasal-airway approaches such as an intranasal stent or nasal dilator. Back2Sleep makes one of these: a CE-certified Class I soft silicone intranasal stent that keeps the nasal airway open during sleep, for snoring and mild-to-moderate obstructive sleep apnoea. It needs no prescription, no electricity and no tubing, and the starter kit contains four sizes.

Two boundaries matter here. A nasal stent is not a substitute for testing, and nothing in this article should push you to skip a sleep study. And a driver with a confirmed AHI of 15 or above under Directive 2014/85/EU needs treatment their national authority accepts as effective at that severity, which an intranasal stent is not certified to provide.

Frequently Asked Questions