Elimination diet for dogs: how to run a food trial that actually works

If your dog will not stop scratching, somebody has told you it might be the food. They are often right. The difficulty is that there is no test for it, and the thing that does work is eight weeks long and fails quietly if you get it slightly wrong.

Written for owners running a trial at home. Reviewed against the veterinary literature cited at the foot of the page. This is information, not veterinary advice.

Roughly a quarter of itchy dogs are reacting to something in their food. The rest are reacting to pollen, dust mites, fleas or something else entirely, and in plenty of dogs it is more than one thing at once. Nothing on a label and nothing in a swab will tell you which of those your dog is, so the only way to find out is to take the food away and watch.

That process is called an elimination diet, or a food trial. Done properly it gives you a real answer. Done loosely it gives you eight weeks of effort and a shrug, which is why so many owners start one, lose faith around week three and never finish.

First, the tests that do not work

Hair and saliva allergy kits are sold heavily to owners of itchy dogs. They are cheap, they arrive quickly and they come back with a long list of ingredients to avoid. They are also not a reliable way to diagnose a food reaction, and the veterinary profession has been saying so for years.

The problem is not that the laboratories are dishonest. It is that there is no blood, hair or saliva marker that reliably separates a dog with a food reaction from one without. Send the same dog's sample twice and you can get two different lists. Send a sample from a stuffed toy and you can get a list back for that too.

If you have already bought one and changed the food on the strength of it, you have not wasted your money entirely, because you have a food the dog has recently eaten. That is useful information for picking a trial diet. Just do not treat the list as a diagnosis.

How long it really takes

This is the number most pages get vague about, and it is the one that decides whether you give up early.

The best evidence comes from a review of 209 dogs on elimination diets. About half had a marked reduction in signs by the end of week three. More than 85 per cent were back to normal by week five. More than 95 per cent by week eight. The authors concluded that a trial should run a minimum of five weeks, and that extending it to eight weeks raises the chance of catching a food reaction to over 90 per cent.

Dogs with gut signs tend to improve sooner than dogs with skin signs. Diarrhoea can settle in a fortnight. Skin takes longer because inflamed skin has to calm down and then regrow, and because a secondary infection often needs treating alongside.

Choosing the food

There are two ways to build a trial diet and the right one depends on how much you know about what your dog has eaten.

A novel protein

One protein and one carbohydrate the dog has never had before. Venison, rabbit, duck, goat and insect are the usual candidates. The logic is simple: the immune system cannot react to something it has never met.

The catch is that it only works if you genuinely know the whole history, and most of us do not. Puppy food, training treats, the chew from the pet shop, two years of a food whose recipe changed quietly. If there is doubt, a protein you believe is novel may not be.

A hydrolysed diet

These are prescription diets in which the protein has been broken into fragments too small for the immune system to recognise. It does not matter whether the dog has eaten that protein before, which makes them the safer choice when the history is uncertain. They cost more and some dogs find them less appetising.

Whichever you choose, it has to be nutritionally complete if the dog is eating it for two months. A home-cooked single protein is fine for a fortnight and a problem by week six unless a nutritionist has balanced it.

What breaks a trial

This is where most trials fail, and they fail silently. The dog eats the trial diet for eight weeks and one other thing, and at the end you cannot tell what the result means.

Everything that goes in the dog's mouth has to be the trial diet. That means no treats of any kind, no dental chews, no rawhide, no bully sticks. It means checking that the wormer and any tablets are not the flavoured sort, because most palatable medications are flavoured with pork or beef. It means no flavoured toothpaste. No scraps from plates, nothing from the neighbours, nothing from the school run, nothing from whoever walks the dog on a Tuesday.

The hardest part is usually other people rather than the dog. Tell everyone in the house, tell the dog walker, tell the grandparents. A single sausage from a well-meaning visitor at week six can undo the whole thing.

Measuring the itch, which is the part everybody skips

Almost every guide tells you to keep a diary. Almost none tells you what to write in it, and "a bit better this week" is not something you can compare with anything.

Veterinary dermatologists use a numbered itch scale for exactly this reason. The owner scores the dog from 0 to 10 each day, with each point tied to a description of what the dog is actually doing, so that your 4 in week eight means the same thing as your 4 in week one. Scales of this kind were developed and validated precisely because owners' impressions drift, and they are what drug trials for itchy dogs use to decide whether a treatment works.

Here is the scale, in plain language. Score once a day, at roughly the same time, thinking about the last twenty four hours as a whole rather than the worst moment of it.

Daily itch score. Pick the description that fits best, not the number that feels right.

Score	Level	What it looks like
0	None	Normal dog. No scratching, licking, chewing or rubbing at all.
1	Very mild	Very occasional. Barely noticeable, easy to forget about.
2	Very mild	Occasional brief episodes, a few times a day.
3	Mild	Noticeable but only now and then. Stops on its own.
4	Mild	Regular episodes but not during sleep, eating or play.
5	Moderate	Frequent episodes. Interrupts play and settling, but not sleep.
6	Moderate	Frequent, and starting to interrupt sleep or feeding.
7	Severe	Prolonged bouts. Wakes at night. Hard to distract from it.
8	Severe	Most of the time. Sleep clearly disturbed, skin looks sore.
9	Very severe	Almost constant. Cannot settle. Broken skin or hair loss.
10	Extreme	Constant, nothing stops it. Distressed and in obvious discomfort.

Score for a fortnight before you change the food. That fortnight is your baseline and without it you have nothing to compare against, which is the single most common reason a finished trial still cannot be interpreted.

Then read the trend rather than the day. Itch moves about on its own, so a dog that is 6 on Monday and 4 on Tuesday has told you nothing. What you are looking for is the fortnightly average drifting down and staying down. A fall of a point or two that holds for two weeks is worth paying attention to. A good day is not.

The rechallenge, and why you cannot skip it

Say your dog improves. That feels like the answer and it is not, yet.

Eight weeks is a long time. The pollen season turned over. You gave a flea treatment. A skin infection was treated and cleared. Any of those could explain the improvement just as well as the food can, and if you stop here you will never know which it was, and you will have put the dog on a restricted diet for life on the strength of a coincidence.

The rechallenge settles it. Once the dog is clearly better, go back to the old food and watch. If food was the cause the signs usually come back within one to three days, although it can take up to four weeks. If they come back, you have your answer and you return to the trial diet. If nothing happens after a month on the old food, the improvement was something else.

After that, if you want to know which ingredient was responsible rather than simply avoiding the old food forever, you add one protein at a time to the trial diet. A small amount at each meal, roughly a tablespoon for a small dog and up to a quarter of a cup for a large one, for four weeks each before you judge it. It is slow, and most owners stop once they have a food that works, which is a perfectly reasonable place to stop.

If nothing changes

About three quarters of itchy dogs are not itchy because of food. If you have run eight clean weeks with a sensible diet and the score has not moved, that is a useful result and worth taking to your vet, because it narrows things down considerably.

The usual next suspects are environmental allergies, which tend to be seasonal at first and year round later, fleas, which can cause severe itching from a tiny number of bites, and mites. All of those have their own treatments, and a dog can easily have two of these problems at once, which is why a food trial sometimes helps a bit without solving it.

Common questions

Can I just switch to a grain-free food instead?

Grain-free is not hypoallergenic. The proteins dogs react to most often are beef, dairy and chicken rather than wheat or maize, and a grain-free food usually still contains a meat the dog has eaten before. Switching to one is not a trial.

My dog is on a hypoallergenic food already and still itches.

Hypoallergenic has no legal definition on a pet food label, so the word on its own guarantees nothing. Check the actual ingredient list for a protein your dog has eaten before. If it has one, the food is not a trial diet whatever the packaging says.

Can I do a trial on raw food?

Yes, if it is a single novel protein, complete, and nothing else is fed alongside. The usual difficulty is that raw ranges often share production lines and recipes change, so check each batch rather than assuming.

How soon should I expect to see something?

Gut signs can settle within a fortnight. Skin is slower and often looks worse before it looks better, particularly if there is an infection being treated at the same time. Half of dogs that are going to respond show a clear reduction by week three.

What if I cannot get through eight weeks?

Five weeks catches most cases, so a five week trial that is genuinely clean is far more useful than eight weeks with treats in the middle. If you know a clean eight weeks is not realistic, aim for five and be strict about it.

A food trial is a diagnostic process and it works best alongside your vet, particularly if your dog has broken skin, an infection, or is on medication. Nothing here replaces that. If your dog is in obvious distress, do not wait eight weeks to ask for help.