How We Calculate

Most calculator sites show you a number and nothing else. That has always struck me as backwards. In a clinic, a number without a source is not a result — it is a claim. So this page exists to make every claim on this site checkable.

Here is what I promise. Every value lives in one file, pet-data.js, which every calculator reads from. Nothing is hard-coded anywhere else. Each entry carries its source, the population it applies to, and a verification date. Where a source gives a range, we store the range and show the range. Where sources disagree, we store both and tell you they disagree — we never quietly average them. And where the evidence does not exist, the tool refuses to produce a number rather than inventing one.

The sources we use, and why these

In order of how much weight they carry here: the Merck Veterinary Manual, the AAHA guidelines (Nutrition and Weight Management 2021, Weight Management 2014, Canine Life Stage 2019, and the AAHA/AAFP Feline Life Stage Guidelines 2021), WSAVA's Global Nutrition resources, the ASPCA Animal Poison Control Center, the Pet Nutrition Alliance, the Ohio State University Veterinary Medical Center, Tufts Cummings School (Petfoodology), the AKC breed standards, International Cat Care, and named peer-reviewed papers where they are the best available evidence.

We do not use pet-food brand marketing, blogs, or other calculators as sources. Where I could only find a number on a blog, I left it out and said so.

Energy: RER and MER

Where our sources genuinely conflict — inactive dogs

This is the biggest disagreement in the whole source set, and it lands squarely on the population this calculator exists for. For an inactive, overweight-prone dog: AAHA 2021 says 1.0–1.2. Merck says 1.4. PNA and Ohio State say 1.2–1.4. AAHA's entire range fails to reach Merck's single value — they do not overlap at all. That is a 40% spread on the same dog.

We store the full 1.0–1.4 span and default to 1.2. AAHA's own footnote under that table is worth reading: "Sedentary and/or indoor pets may require less caloric intake than indicated above." AAHA is telling you its own numbers skew high. We take that seriously.

Corrections we made to values we started with

These were in our own working draft before verification, and every one of them is wrong. They are all still in wide circulation elsewhere, which is why I am listing them rather than quietly fixing them.

A modelling caveat we could have hidden

WSAVA does not use the RER × multiplier model at all. Its cat chart is built on NRC 2006 equations with different exponents entirely — lean adult 100 × BW^0.67, obese-prone 130 × BW^0.4. A RER × factor model cannot reproduce that curve; the shape differs, not just the scale. So our numbers will diverge from the WSAVA chart at the weight extremes. That is expected, and it is not something we intend to fudge away by nudging our factors. Likewise, AAHA/AAFP 2021 use the linear RER for adult cats, so our exponential default sits about 1–3% off their chart.

Dog age: the AKC size-adjusted chart

The chart is the AKC's. It is not the AVMA's. The AVMA publishes no dog age chart — its 2023 brochure contains zero occurrences of the word "human". "The AVMA dog age chart" is a misattribution repeated across most of this niche, and we would rather be correct than consistent with everyone else.

Verified 2026-07-16 against the AKC's published chart. Two honest notes. The chart states no derivation and no evidence base — it is authoritative by convention, not a study output. And it stops at 16 years; there is no "16+" row, so anything past 16 on this site is our own extension of the chart's final step, and we label it.

The structure is worth understanding: small, medium and large are identical for years 1–5 and only diverge at year 6. Giant breeds start lower (12 at year one, against 15) and then cross over at year three. Full table on the dog age calculator.

Dog age: the 2020 UCSD epigenetic formula

human age = 16 × ln(dog age) + 31 — from Wang et al., Cell Systems, 2020.

We show it because it is real science and it is what people have read about. We show it alongside the AKC chart, rather than instead of it, because of what the paper actually says about itself:

• The cohort was 104 dogs (95 after quality control), primarily Labrador retrievers, aged 0.1–16 years, compared against 320 humans from previously published data.
• The authors' own words: "Distinct breeds exhibiting widely varying lifespans could yield different age-translation functions." They make no all-breed claim.
• It fits worst in mid-life, by their own admission: "For adolescent and mature stages, the correspondence was more approximate."
• The published equation is a blend of two fits (17ln+33 and 16ln+30), not a single regression.
• It states no valid age range. Below 0.144 years — about 53 days — it returns a negative human age. Our calculator guards that boundary and returns nothing rather than nonsense. Several others do not.
• At one year it says 31. The AKC chart says 15. Both are on our page.

Cat age

Year 1 = 15 human years, year 2 = 24 (+9), then +4 per year thereafter, to 25 years = 116. Life stages: Kitten (0–6 months), Junior (7 months–2 years), Adult (3–6), Mature (7–10), Senior (11–14), Super Senior (15+).

Source: International Cat Care. It is not AAFP/AAHA's. We text-searched both the 2021 and the 2010 AAFP/AAHA feline life stage guidelines; neither contains any human-year conversion at all. The chart comes via the Feline Advisory Bureau, which is why the older AAFP stages line up with it. Like the AKC chart, it states no derivation — authoritative by convention.

Chocolate toxicity

The correction underneath our chocolate tool

Merck publishes a single mg/oz column that most calculators copy and label "theobromine". It is not theobromine. It is methylxanthines — theobromine plus caffeine. The identity holds six for six: 0.25 + 0.85 = 1.1. 58 + 6 = 64. 393 + 47 = 440. 737 + 70 = 807. Merck's column is exactly the ASPCA's two columns summed.

This matters because the published thresholds are methylxanthine thresholds too. So the correct maths is to total both stimulants and compare against them — which is what we do, and which is why our number runs slightly higher than tools that count theobromine alone. A calculator using Merck's figures while calling them theobromine is double-labelling: right number, wrong name, and no way for a reader to tell.

One more thing you deserve to know, and I went back and forth on whether to say it. ASPCA attributes these thresholds to "(ASPCA/APCC Database: Unpublished data)." The most-repeated numbers in this entire subject terminate in an internal database that nobody outside can inspect. Merck reprints them without that attribution. They are still the best figures available and we use them — but "best available" and "independently verifiable" are not the same thing, and you should know which one this is.

We also carry Merck's formula for chocolate of a known cocoa percentage — 15.5 mg/g × cocoa fraction — which beats a category guess when the wrapper states a percentage.

Two further corrections. Merck reversed its activated-charcoal guidance: the current position is a single low dose, for lethal exposures only, with hypernatraemia risk. The "repeat doses for enterohepatic recirculation" line that competitors still repeat is the 2001 position. And on inducing vomiting, the ASPCA is unambiguous: "Trying to induce vomiting in your pet at home can be dangerous." We publish no home-emesis instructions.

Grapes and raisins

There is no established safe dose, and our data file stores that as a permanent null. It is not a placeholder to fill in later. It is the finding.

What the evidence establishes: grapes and raisins cause acute kidney injury and death in dogs, and since 2021 it points strongly — but not conclusively — to tartaric acid as the toxic principle. What it does not establish: any dose below which they are safe, a threshold, or a dose–response curve.

Four independent reasons, each sourced, and they compound. The tartaric acid content of grapes runs roughly 0.35%–1.1% and can reach 2%, varying with cultivar, growing conditions and locality — so two identical-looking grapes are not the same dose. Dogs differ in individual sensitivity. The clinical case series show no relationship between dose and outcome. And the poison services say so outright.

The number we use for the tool is Merck's conservative risk threshold: more than one grape or raisin per 4.5 kg (10 lb) of body weight. It is derived from how much tartaric acid a grape could contain, not from observed cases, and for a consumer-facing tool that is the right basis.

We also show the "lowest observed toxic doses" — 0.7 oz/kg for grapes, 0.11 oz/kg for raisins — labelled for exactly what they are: the smallest amounts recorded in dogs that got sick. Not thresholds. Note the roughly 20× gap between Merck's figure and these: for a 10 kg dog, more than 2 grapes versus about 40. They answer different questions, and the honest reading is that the true threshold is unknown.

On prognosis, the numbers differ tenfold depending on the denominator, and the difference matters enormously to a frightened owner. Across all 139 known ingestions in Reich 2020: 6.7% developed acute kidney injury, and 138 of 139 survived. The frequently quoted 47% mortality from Eubig is among dogs already in renal failure. Quoting "47% die" at someone whose dog just ate three grapes is a serious distortion. Neither figure licenses complacency — the reason the first number is small is early treatment.

We do not publish the circulating "11 g/kg" or "3 g/kg" figures — provenance untraced. And we will not mention calcium carbonate or Tums as an option: the 2025 JAVMA paper on it is an explicitly unvalidated hypothesis, and since hypercalcaemia is a hallmark and a negative prognostic indicator, that suggestion could be actively dangerous.

Growth curves

This is the part of the site I am most pleased with and the part that needs the most honest labelling.

The gold standard is Salt et al. 2017, PLOS ONE — growth standards built with GAMLSS/BCPE modelling from Banfield Pet Hospital records, 1994–2013, across roughly 900 US hospitals, screened to healthy dogs in ideal body condition. The final model used about 106,000 data points across some 33,000 individual dogs. Charts cover 12 weeks to 2 years, with nine centiles plotted.

The paper publishes those standards as graphs only. There are no numeric percentile tables in the paper, in its supplement, or in its public dataset. Nobody can look these numbers up. That is presumably why every competitor fell back on multiplier rules instead.

So we got them another way. We downloaded the published WALTHAM chart PDFs — male and female, all five categories — and programmatically extracted the vector path coordinates of the nine centile curves, then calibrated against the printed axes. The calibration self-validates, which is why I trust it: every x-axis gridline resolves to exactly 1-week spacing, every y-axis gridline to exact round kilogram steps, and all nine curves on all ten charts begin at precisely 12.0 weeks — independently reproducing the paper's stated range.

Size classes

We use WALTHAM's five categories as the engine's internal model, because they were derived by cluster analysis of how dogs actually grow rather than by picking round numbers. Our toy/small/medium/large/extra-large/giant labels are a convenience mapping we define openly: toy under 6.5 kg, small 6.5–9, medium 9–15, large 15–30, extra large 30–40, giant over 40.

The AKC uses four classes by imperial weight and disagrees with all of this. It has no toy class at all; its "Small" (0–20 lb) swallows WALTHAM's category I and II and part of III. Sources genuinely differ, and we publish both rather than pretending one is canonical.

No standard exists for giant breeds

Salt et al. could not model category VI. Growth patterns among breeds over 40 kg were too inconsistent to combine into a single curve, so they published five categories and left the sixth blank. Royal Canin's own veterinary teaching material repeats this.

Which means every Great Dane, Mastiff and Newfoundland growth prediction on the internet is an extrapolation with no evidence base underneath it — including ours. We extend the category V curve, blend it toward the AKC's X-Large percentages, and flag the result on screen as having no published standard behind it. I have not found another calculator that tells you this.

Our accuracy bands, and where they come from

The bands we publish (±22% toy, ±15% small, ±16% medium, ±10% large, ±9% extra large, ±20% giant) are the width of one centile line either side of the median, measured from our digitization.

What that means in practice comes from Salt et al. 2020, which measured how much healthy dogs actually drift: 42% of healthy dogs cross more than one centile line during growth, but fewer than 5% cross more than two. Read it both ways. A puppy's track is stable enough that curve-based prediction works at all — and unstable enough that one weigh-in is a weak anchor. WALTHAM's own guidance is three serial measurements about two weeks apart, after 12 weeks, before trusting a track.

The same paper gives the clinically useful contrast: in dogs with abnormal body condition, 68% of obese dogs crossed two or more centile lines upward and 49% of underweight dogs crossed two or more downward. The signal is not "off the average" — it is crossing lines, consistently, in one direction.

Why we do not adjust for neutering

Because the best available evidence says the effect is smaller than our error bars, and the people who built the curves declined to model it.

Two effects get conflated here, and almost everyone has them backwards. Neutering before growth-plate closure makes a dog taller — Salmeri et al. 1991 (32 dogs, gonadectomy at 7 weeks or 7 months against intact controls) found the growth rate unaffected but the growth period extended, producing greater final bone length. The effect on weight is real at population scale but, in Salt's words, "dog-to-dog variability was large compared with the magnitude of the post-neutering change" — so WALTHAM concluded that "separate curves for neutered dogs were not needed."

You will also see the Golden Retriever and Labrador neuter studies cited here. Those are about joint disorders and cancers. They are important and not evidence about adult size. Different question.

Kitten growth

The cat side of this turned out to be the mirror image of the dog side, and the contrast is interesting enough to be worth spelling out.

A published feline growth standard does exist. Salt et al. 2022 in PLOS ONE is the feline sibling of the 2017 dog paper — same institute, same Banfield data, same modelling approach, 8 to 78 weeks, sexes separate, nine centiles. So unlike giant-breed dogs, kittens are not an evidence desert.

It has the same defect, though: graphs only, no numeric table. We checked three ways, including extracting the text layer out of the AAFP-hosted chart PDFs — axis labels and nine percentile labels, and not a single weight value. And it has one gap the dog paper doesn't: it is not validated for neutered cats. The authors excluded them by design. Most pet cats are neutered by four to six months. That gap is larger than it sounds.

So rather than digitize curves that exclude most of our readers' cats, we use the one measured table that exists: DiGangi et al. 2019 in the Journal of Feline Medicine and Surgery — birth to six weeks from a 246-cat colony with known dates of birth, and eight to twenty weeks from 1,310 owned kittens. It runs to 20 weeks. Beyond 20 weeks, no numeric kitten weight table exists anywhere, so that is where our tool stops rather than extending someone's curve past its evidence.

Two kitten rules that don't survive checking

"A kitten weighs a pound per month." A convention with no traceable originator — DiGangi call it "the clinical guideline" and cite nobody. But it was actually tested, and it falls apart exactly where people use it. Among privately owned kittens whose ages had to be estimated — the situation an owner or a shelter is actually in — it lands within a week of the truth for 77.4% at 6 weeks, 57.0% at 10 weeks, and 17.9% at 20 weeks. The usual phrasing, "up to about six months", is precisely the range where it stops working. The authors' own conclusion: body weight predicts age well through 10 weeks, and "factors other than body weight should be considered when estimating kitten age beyond that time point." ASPCApro independently caps it at 10 weeks. Our tool shows the rule's answer alongside its accuracy at the age you entered, because the accuracy is the interesting part.

The widely-reproduced ASPCA kitten weight chart is arithmetic, not data. Its whole eight-week table reconstructs exactly from 100 g + (7–15 g × days) — we verified that at eight independent checkpoints. It is a formula published by a welfare body, which means it cannot corroborate anything. It also disagrees with measurement: its day-56 range (485–925 g) sits below DiGangi's measured 8-week mean of about 920–1,020 g. The ASPCA does enormous good and this is a chart rather than a scandal — but a reader deserves to know which of two numbers came from weighing kittens.

The related "100 g per week" rule conflates two figures: a quarter pound is 113 g, not 100 g. The 113 g/week number is traceable to UC Davis Koret Shelter Medicine and UW Madison.

Neutering and kitten growth — the cat evidence is better than the dog evidence

And it kills a claim the internet repeats constantly. Root et al. 1997 found that gonadectomy delays physeal closure and significantly increases final radial length — bone, not weight. But in the authors' own words: "Age at gonadectomy had no effect on age and radial length at time of the growth plateau." Neutered at seven weeks or at seven months, the outcome was identical.

So "don't neuter early or he'll grow too tall" fails the one study built to test it. Early versus late makes no difference. Neutered versus intact does. Weight gain after neutering is a separate mechanism entirely — an obesity-risk question, per AAHA/AAFP — and conflating the two is how this myth survives.

What we won't publish about kittens

Fading-kitten mortality statistics (the source we found is an image-only PDF with no text layer, so we could not verify it). The "cats reach 75% of adult weight by six months" figure attributed to Cornell — it is on no Cornell page we could fetch, and looks like a content-farm invention laundered through a search summary. And the by-sex or by-breed adult weights from Taylor 2022, because two extraction passes disagreed by about 0.9 kg; only the overall median (4.16 kg, IQR 3.30–4.99) is safe.

Also worth saying: "a typical adult cat is 8–10 lb" is a convention, not a veterinary standard. No veterinary body publishes a normal adult range. Cornell deliberately defines obesity relative to the individual cat's own normal and uses body condition score instead — which is a better approach, and the one we point people to.

Raw and homemade feeding

Every number on that page is labelled either vet source or community convention, and the split is stark.

Community convention: the 2–3%-of-body-weight rule (traces to Ian Billinghurst's BARF, 1993), the puppy 8–10% figures (sourced only to raw-food retailers), and the 80/10/10 ratio (a prey-model derivation). We checked Tufts, ACVN, WSAVA, AVMA, FDA, Merck, PNA and UC Davis: no veterinary nutrition body publishes a percentage-of-body-weight rule. Billinghurst is himself a vet, so "no vet ever said it" would be too strong — but no nutrition body or peer-reviewed source endorses it.

We give you the arithmetic anyway, because you asked for it, and we show the cross-check nobody runs: energy needs scale as body weight^0.75 while the percentage rule is linear, so one percentage cannot fit both ends of the size range. It also applies a mass rule to foods spanning roughly 30–80 kcal an ounce.

The quality of life scale

The HHHHHMM scale is Dr. Alice Villalobos's. Seven factors, 0–10 each, with a total above 35 described as acceptable for continuing with hospice care.

Three things nearly every page about it gets wrong. First, the 0–10 scale and the ">35" threshold come from the scale created for her 2007 book (revised for IVAPM in 2011) — not from the 2004 Veterinary Practice News article that everyone cites, which uses a 1–10 scale and a per-item rule. Different instruments, different years. Second, >35 licenses continuing hospice care — it does not mean a pet is fine, and below 35 is not a euthanasia trigger. Third, the "2:1 ratio" widely attributed to her under "More Good Days Than Bad" does not exist in either primary source. We do not publish it.

On validation, the honest answer is neither yes nor no: it was an expert clinical heuristic for about 19 years, then validated once — Testoni et al., Animals 2023, Cronbach's alpha 0.78, with the original one-factor model fitting only "partially adequate" and an Italian, 81.8%-female sample. The >35 cut-point itself has never been validated against any outcome.

The scale is copyrighted — the canonical PDF carries "Reprinted with permission from Dr. Alice Villalobos & Wiley-Blackwell." The factor names, the range and the threshold are facts and we state them. Her descriptive text is her work, and we have not reproduced it; the scoring guidance on our page is written from scratch. And the 2016 AAHA/IAAHPC End-of-Life Care Guidelines never mention Villalobos, Pawspice, HHHHHMM or any named scale — so we do not imply they endorse it.

Poison control numbers and fees

Verified directly against each organisation's own site on 2026-07-16. These are the most perishable facts here, so they get re-checked quarterly — next around October 2026.

You will see "$95" quoted widely as the ASPCA's fee. That figure comes from third parties — ASPCA Pet Health Insurance, Embrace, GoodRx — not from the ASPCA, whose own pages say only that a fee may apply. We are not going to print a price tag on someone else's behalf.

There is also a second number circulating on third-party veterinary sites — including VCA — presented as the Pet Poison Helpline's. It does not appear anywhere on the Helpline's own site. We are not going to reprint it here even to warn you off it, because a phone number in an article about phone numbers is exactly the kind of thing that gets screenshotted without its caption. If you have a number for them that isn't the one above, check it against their own site before you dial in an emergency.

What we deliberately do not publish

Medication dosing. It is the single biggest traffic opportunity in this niche — Benadryl, trazodone, gabapentin, meloxicam, carprofen — and we will not publish it without a named DVM reviewer on the byline. A dosing page is not a content decision. If we get that reviewer, those pages will show vet-typical ranges with contraindication warnings, and never a single directive number.

Any claim that a toxicity exposure is safe. Our toxicity tools output urgency states. The lowest state on the chocolate tool is "monitor at home", and it still tells you what to watch for and when to call. The grape tool has no state below "call your vet", by design.

Home instructions for inducing vomiting. The ASPCA is unambiguous that this is dangerous without veterinary direction, and never appropriate in cats.

Numbers we could not source. Kcal ranges for 3 oz and 5.5 oz cans (every figure traced to brand labels or blogs). Per-chocolate-chip weights (the ubiquitous "311 chips per cup" has no traceable origin and is internally inconsistent). Theobromine in chocolate-covered espresso beans or Oreos. A published spread of adult weight within a breed. We would rather have a gap than a guess.

Change log

Found something wrong? Tell me. A correction with a source beats a compliment, and it will be logged here with the date.