How Dogs Detect Seizures & Low Blood Sugar Before They Happen

By Sarah Bennett, Certified Animal Nutritionist

Imagine having a four-legged companion who taps you on the leg, sits rigidly, or refuses to let you walk out the door — and fifteen minutes later, you experience a diabetic crash or an epileptic seizure. For thousands of people living with epilepsy or Type 1 diabetes, this is not a comforting fantasy but a documented daily reality. Alert dogs trained — or, in many cases, spontaneously able — to predict medical episodes are transforming lives in ways that no wearable sensor or pharmaceutical intervention has yet matched.

Seizure Alert Dogs: What Do They Actually Detect?

The question of precisely what seizure alert dogs detect has challenged researchers for decades. The dominant hypothesis is chemical: just as dogs can identify cancer-related volatile organic compounds (VOCs), they may be detecting subtle biochemical shifts in sweat, breath, or skin secretions that precede a seizure. Changes in serotonin, adrenaline, or other neurotransmitter-linked compounds may alter body odour in the minutes or hours before a seizure, and a dog's 300-million-receptor nose may be sensitive enough to register those changes.

Alternative theories include detection of micro-behavioural changes invisible to human observers — subtle alterations in gait, facial muscle tension, or breathing patterns that precede the electrical cascade in the brain. Some researchers believe it may be a combination of both olfactory and visual cues. The Guardian reported on a 2019 study from the University of Rennes that found, in a carefully controlled design, dogs detected a specific smell associated with epileptic seizures with up to 100% accuracy — suggesting the olfactory route is at least a significant component.

A key study published in Science Daily in 2019 (Science Daily coverage here) confirmed that dogs trained to identify the scent of seizure-positive sweat samples could distinguish them reliably from non-seizure samples. Crucially, they were also able to distinguish between seizure samples from different individuals, suggesting the scent signature is generalised across epilepsy rather than unique to any one person — which has major implications for training standardisation.

The Research on Seizure Dogs

Despite widespread use of seizure alert dogs in clinical and home settings, rigorous peer-reviewed data is still relatively sparse. This is partly because seizure prediction — as opposed to seizure response — is extraordinarily difficult to test under controlled conditions. A dog cannot be conveniently placed beside a person and timed for its alerts in a lab environment in the way cancer detection can be evaluated with stored urine samples.

What research does exist is encouraging. Brown et al. (PMID 22985386) examined owners' reports of seizure prediction in family dogs who had not undergone formal training. A significant proportion of owners reported that their dogs exhibited alert behaviours — becoming clingy, agitated, or vocalising — consistently before a seizure event. While owner-reported data carries methodological limitations, the consistency across studies is striking.

Importantly, some dogs who live with people with epilepsy appear to develop seizure alert behaviour spontaneously, without any formal training. This suggests that the signal is sufficiently strong and distinctive that a motivated, bonded dog can learn to associate it with an impending medical event through everyday experience.

Diabetes Alert Dogs: A Stronger Evidence Base

If the science of seizure alert dogs is still developing, the evidence for diabetes alert dogs (DADs) detecting hypoglycaemia is considerably more robust. Low blood sugar (hypoglycaemia) produces a well-characterised chemical change: as glucose falls, cells switch to burning ketones, and isoprene — a specific VOC — increases measurably in breath. Dogs trained on this compound have demonstrated reliable and impressive detection rates.

Research by Rooney et al. (PMID 25637371) examined whether trained diabetes alert dogs could reliably distinguish hypoglycaemic samples from euglycaemic (normal blood sugar) samples in a controlled setting. The results confirmed above-chance detection, and owner surveys found that trained DADs reduced Dangerous">dangerous-dog-toys" title="10 Dog Toys That Are Actually Dangerous">Dangerous">Dangerous">Dangerous (And What to Use Instead)">dangerous hypoglycaemic episodes significantly for their handlers. The Medical Detection Dogs charity, which trains some of the UK's most rigorously evaluated DADs, reports that their dogs alert to low blood sugar events in real-world conditions — sometimes before a glucometer would flag a problem.

The BBC covered a compelling case in which a trained Labrador alerted his owner to dangerous low blood sugar levels during the night — a scenario where many dangerous hypoglycaemic events go undetected because the person is asleep. Nocturnal hypoglycaemia is a significant cause of mortality in people with Type 1 diabetes, and a reliably alerting dog may provide a life-saving safety net that continuous glucose monitors sometimes miss.

Hyperglycaemia: Can Dogs Detect High Blood Sugar Too?

While most research has focused on hypoglycaemia detection, some trainers and owners report that their dogs also alert to high blood sugar (hyperglycaemia). The chemical basis for this is less well characterised — glucose itself is not highly volatile — but ketoacidosis, the dangerous condition that can accompany very high blood sugar in diabetics, produces acetone and other VOCs in breath that dogs may learn to associate with danger. This area warrants further controlled investigation.

How Alert Dogs Are Trained

Both seizure and diabetes alert dogs undergo specialised positive-reinforcement training, often starting with scent discrimination tasks using samples collected during actual medical events. For diabetes alert dogs, training typically involves:

• Collecting saliva or sweat samples during verified hypoglycaemic episodes (confirmed with a glucometer).
• Presenting those samples alongside euglycaemic samples in a scent discrimination exercise.
• Rewarding the dog for correctly identifying the hypoglycaemic sample.
• Gradually training the dog to alert its handler using a specific, reliable behaviour (pawing, barking, sitting, nudging a button).

The process takes 6–18 months for a skilled trainer. Importantly, the dog's alerting behaviour must be consistent — a dog that alerts only sometimes is potentially worse than no alert dog at all, because it may create false reassurance on missed events.

Limitations and What the Science Can't Yet Tell Us

It is important to be clear about what this field of research does not yet conclusively establish. Alert accuracy varies enormously between individual dogs and training programmes. Not all dogs are suitable — this is a high-drive, high-focus task that suits specific temperaments and requires sustained motivation to work. There is also no standardised certification system for alert dogs in most countries, which means the quality of animals supplied by different organisations varies widely.

Medical alert dogs should be viewed as a complementary tool — not a replacement for continuous glucose monitoring, anti-epileptic medication, or medical supervision. They are, however, a tool of remarkable biological sophistication that is changing lives in measurable, documented ways.