If your dog has just eaten something, stop reading and call. ASPCA Animal Poison Control on (888) 426-4435, or Pet Poison Helpline on (855) 764-7661. Both run 24/7 and both charge a fee. Have your dog’s weight, what it ate, and roughly how much. Do not make your dog vomit unless you are told to. Why that matters is below.
Now the part you can read at leisure.
Popular “foods toxic to dogs” articles run to thirteen items, or twenty, sometimes thirty. The Merck Veterinary Manual’s food hazards section holds exactly seven: garlic and onion, avocado, bread dough, chocolate, macadamia nut, grape/raisin/tamarind, and xylitol. The main peer-reviewed review, Cortinovis and Caloni (2016), covers eight. It drops avocado and adds ethanol and hops.
Of those eight, five carry any dose figure for dogs. Read them closely and only two have a real dose-response: a number where more food reliably means worse harm. The rest give a lowest dose from a poison-centre file, or no number at all.
That gap is what this article is about.
If your dog has just eaten something
The most important correction we can offer: do not make your dog vomit on your own judgement. Pet Poison Helpline says it plainly: “do NOT induce vomiting without consulting a vet or Pet Poison Helpline”.
Sometimes the vomiting is what does the damage. Merck lists when not to do it. It is wrong “when the swallowing reflex is absent” or “the patient is not neurologically stable”, because a dog that cannot swallow may breathe in what comes up. It is wrong when what went down includes “corrosive agents, volatile hydrocarbons, or petroleum distillates”: drain cleaner, petrol, lamp oil. And it is wrong when “the risk of aspiration pneumonia is high”. Add flat-faced (brachycephalic) breeds, dogs with heart or breathing disease, and any dog already vomiting, drowsy or unsteady.
Hydrogen peroxide is not the harmless kitchen remedy it is sold as. Niedzwecki and colleagues (2017) put a camera into six healthy dogs given 3% peroxide, plus one given the vet drug apomorphine. All six peroxide dogs had stomach lesions by four hours, worse at 24 hours, and one had grade III oesophagitis (a badly inflamed gullet) a week later. The apomorphine dog had minimal lesions and normal tissue. Merck treats peroxide at home as what you use “when veterinary care or prescription options are not available”: a fallback, not a first move.
Timing depends on what was eaten, in a way no owner can be expected to know. For a pill that dissolves fast, the window may be 15–30 minutes. For chocolate it runs up to 12 hours, because chocolate slows the stomach down. That alone argues for phoning rather than guessing.
And do not wait for symptoms. Three of the worst entries below announce themselves late. Xylitol liver injury can arrive with no low blood sugar before it, onion damage to red blood cells peaks around three to five days, and grape kidney failure builds over 24 to 72 hours.
Tier 1: dose-response established
Two entries only.
Chocolate and other methylxanthines. Methylxanthines are the stimulants in chocolate, theobromine and caffeine. A dog clears them far more slowly than you do. The thresholds cover the two added together: mild signs from 20 mg/kg, severe and cardiotoxic signs, meaning harm to the heart, at 40–50 mg/kg, seizures at 60 mg/kg or above. The LD50, the dose that kills half the animals given it, is 100–200 mg/kg for each compound. Merck and the ASPCA APCC’s own brief give identical figures, and the Frontiers review reproduces them.
Content varies hugely by type, which is why “how much” is the wrong question without “what kind”:
| Type | Theobromine + caffeine |
|---|---|
| Dry cocoa powder | 807 mg/oz |
| Baking / unsweetened | 440 mg/oz |
| Semi-sweet chips, dark | 150–160 mg/oz |
| Cocoa bean hulls (garden mulch) | 255 mg/oz |
| Milk chocolate | 64 mg/oz |
| White chocolate | 1.1 mg/oz |
Signs usually show within 6–12 hours, and “may persist for up to 72 hours in severe cases”.
One note on where those numbers come from. The 20 / 40–50 / 60 set traces back to the ASPCA’s poison-centre database, which the source itself calls unpublished. Everyone reproduces it and it is the accepted clinical standard. It still came out of case records, not a controlled trial that gave measured doses.
Xylitol, low blood sugar only. Merck describes “a rapid, dose-dependent insulin release that can result in profound hypoglycemia” (a flood of insulin that drops blood sugar hard) at doses “greater than approximately 100 mg/kg”. The Frontiers review cites effects “as low as 0.03 g/kg”, which is 30 mg/kg. Signs can start within 30 minutes, or be “delayed up to 12–18 hours”. We have not settled which lower bound is better supported, so we quote both.
Tier 2: real toxicity, no reliable dose
Xylitol, liver injury. A separate problem from the blood sugar, and the place popular content most often overstates the evidence. The familiar 0.5 g/kg figure is the lowest dose ever seen alongside liver failure, in a case series of eight dogs, not a line below which the liver is safe. Merck hedges to match: “some dogs … may develop severe hepatic insufficiency”, which means the liver failing. A 2025 look back at 95 dogs found half developed a hepatopathy, liver damage of some kind, while “the dose of xylitol ingested … was not significantly associated with the development of hepatopathy”. Six of the original eight dogs never had low blood sugar first. Treat it as idiosyncratic, striking some dogs and not others for reasons nobody can predict, and do not read a normal blood sugar as the all-clear.
Grapes and raisins. For roughly two decades vets gave firm warnings with no known mechanism and no reliable dose. Wegenast and colleagues (2022) proposed a mechanism, from six dogs poisoned by cream of tartar and tamarind: tartaric acid is “identified as the likely toxic component”. Note the hedge. It is inference from things lining up, not proof, and the authors say the way the kidney tubules die “was not described in the literature”. Merck has adopted it, and gives a rule of thumb rather than a threshold: “more than one grape or raisin per 4.5 kg (10 pounds) of body weight may contain enough tartaric acid to pose a risk”. Vomiting comes at 6–12 hours, kidney failure at 24–72 hours, with little urine or none, and “anuric dogs are unlikely to survive” without dialysis. Naming the toxin did not produce a dose, which is why grapes stay in Tier 2.
Onion, garlic and other alliums. Sulfur-containing oxidants damage red blood cells. They throw up Heinz bodies, clumps of damaged blood pigment, “within 24 hours”, peaking “in approximately 72 hours”, and the cells then break apart “typically 3–5 days after exposure”. For onion, 15–30 g/kg has produced clinical signs; for garlic, Merck gives no figure beyond “3–5 times more toxic than onion”. The usual framing is stronger than the evidence. In the experimental study, dogs got the equivalent of 5 g of whole garlic per kg daily for seven days. They developed Heinz bodies and eccentrocytes, red cells with the pigment shoved to one side, but “no dog developed hemolytic anemia”. Separately, some dogs inherit a red-cell type that leaves them more open to harm, reported to be more common in Japanese breeds including Akita, Shiba and Jindo. We found no reliable figure for how common, and we are not guessing one.
Macadamia nuts. Three separate sources agree the mechanism is unknown. In a 48-case series the mean amount eaten was 11.7 g/kg, range 2.2–62.4; signs showed within 12 hours in 79% of dogs and cleared in 24–48 hours. Merck describes it as mild and self-limiting, with most dogs recovering without specific treatment. Real, unexplained, and not fatal.
Hops. The entry that deserves far more attention than it gets. A look back at 71 dogs found 77% survived, a 23% death rate. Temperature ran high in 96% of them, and the median time to death was 10.7 hours. Every degree of temperature rise came with a 78% higher chance of dying, and none of the treatments looked at (washing out the gut, cooling, drugs) improved survival. The toxic principle is unknown, and per the Frontiers review, “the dose required to induce toxicity has not been established precisely”. Signs can start 30–60 minutes in. If you brew beer at home, this is the item on this page to worry about.
Bread dough and ethanol. Two problems at once, per Merck. The mechanical one: “dough expansion causes gastric distension and resultant vascular compromise to the gastric wall, similar to that in gastric dilation and volvulus”. In plain terms, the dough rises, the stomach stretches, and the blood supply to the stomach wall is choked off, as in bloat. The chemical one is ethanol, the alcohol yeast makes as it ferments, and “death is usually due to metabolic effects of ethanol rather than to gastric distention”. Signs start 30–60 minutes in. We could not find a confirmed ethanol dose for dogs in any source we could reach, so we are not publishing one.
Cannabis. A survey of 251 North American vets found case numbers rose significantly after Canada legalised it, with 82 of 85 vets reporting a rise. The striking finding is how rarely it kills on its own: a cited study “was unable to determine a lethal dose”. Sixteen deaths were reported, most of them second-hand: inhaled vomit in the lungs, breathing stopping. Edibles were the commonest way in, which means the real combined risk is THC plus chocolate in the same brownie. Note the design limit: this is vets recalling their own caseload in bulk, not counted case records. The direction is well supported. The size of it is not.
Nicotine. The APCC brief reports a minimum lethal dose of “20 to 100 mg, which may consist of one to five regular cigarettes”. That figure is not adjusted for bodyweight, which makes it far more dangerous to a small dog than it sounds. Cigarettes hold 9–30 mg each, patches 8.3–114 mg. Signs start within an hour, among the fastest here. The brief predates vaping, and we found no source covering e-liquid strengths.
Tier 3: mechanism or single case reports only
| Food | What the evidence actually is |
|---|---|
| Avocado | Merck: dogs “seem relatively resistant”; one case report of two dogs. Pet Poison Helpline: “Avocado is non-toxic to dogs and cats.” The real risks are the stone blocking the gut, and fat. |
| Tomato, green potato | ASPCA lists the plant, for solanine, and says “ripe fruit is non-toxic”. No dose for dogs, no case series, no Merck entry. |
| Apple seeds | Cyanide kills at roughly 2 mg/kg, but apple is not among Merck’s cyanide plants, poisoning is overwhelmingly a farm-animal problem, and we found no published dog case. The risks are a blockage and cracked teeth. |
| Nutmeg | Myristicin is a real toxin, but recipe amounts are “very unlikely to cause serious toxicity” per Pet Poison Helpline. |
| Cinnamon | Same source: “generally well tolerated”. Risk needs over a teaspoon of powder, or the essential oil. Almost no list makes that distinction. |
| Star fruit | No dog evidence of any kind. The research is all human, mostly dialysis and kidney patients. |
| Raw eggs | Avidin, a protein in raw egg white, does bind biotin. We found no published dog case of biotin deficiency from eating eggs. If you want to avoid raw egg, Salmonella is the better argument. |
Not toxicity at all
Milk. Lactose upsets the gut in proportion to how much goes in. It pulls water into the bowel and feeds the bacteria there. There is no toxin, no target organ, no lethal dose. Listing milk beside xylitol is a category error, and it makes the lethal entries easier to wave away.
Bones. A mechanical hazard. In a series of 129 dogs with bone stuck in the gullet or stomach, one thing predicted a trapped bone: its width against the size of the dog, or cross-sectional area relative to bodyweight. Most bone that reached the stomach was left there to dissolve, “with no reported complications”. That study draws no line between cooked bone and raw, so the universal claim that cooked bones splinter and raw ones are safe is not something we can pin on it.
What actually poisons dogs
The ASPCA’s 2025 figures cover over 376,000 exposure cases. Medicines and supplements bought over the counter came first at 16.9%. Food and drink came second at 16.4%. Human prescription medicines were third, and chocolate fourth at 13.6%. Within the food group, “grapes, raisins, xylitol gums, onions and garlic make up the largest part”.
Two honest readings of that. Food really is about 30% of cases, so that part of the popular story holds. But the biggest single driver is medicine, not food (pills on the floor and in handbags), while content aimed at owners is nearly all food lists. And the foods named are exactly the Tier 1 and Tier 2 entries. Nutmeg, star fruit, apple seeds, raw eggs and milk appear in none of the years we checked. One newer entry does: protein drinks and bars, a sugar-alcohol route in that no traditional list mentions.
These are call volumes at a poison centre, not rates of illness, and not weighted by how bad the case was. They tell you what generates calls, not what kills.
The asymmetry, in one line
Hops kills roughly one dog in four that turns up at a clinic. It has no known toxin, no established dose, and no treatment shown to help, and it appears on almost no popular list. Avocado appears on nearly every one, and the poison centre most likely to get the call about it says it is non-toxic to dogs.
A list that treats those two as comparable is not informing anyone. That is why we have sorted this page by evidence rather than A to Z, and why some entries say plainly that nobody knows.
This article is not a substitute for a phone call. If your dog has eaten something, the numbers are at the top.