How each question is assessed
Every question on this map carries two marks, and they answer two different questions. Certainty says how much the evidence can be relied on. Direction says which way that evidence points. A question can score low on the first and still be clear on the second, and the two are recorded separately for that reason.
Take atenolol in preclinical hypertrophic cardiomyopathy as an example. The certainty is Low, because the only study is an open-label observational cohort in 63 cats rather than a randomised trial. The direction is No benefit shown, because that cohort looked for a survival difference over five years and did not find one. Reporting a single combined score would lose one of those two facts.
Certainty: how far the evidence can be relied on
These five labels follow GRADE, the system used for the same purpose in human medicine. The starting point is the study design, and the certainty is then rated down for risk of bias, for imprecision, for evidence that is indirect, and for findings that disagree between studies. It may be rated up where an effect is large and consistent. GRADE has four levels, and the fifth label used here, No evidence, marks a question nobody has studied at all.
Direction: which way the evidence points
Three of these are routinely run together, and the distinctions matter at the bedside. No benefit shown means somebody looked and found nothing, which is a finding. Inconclusive means somebody looked and the study cannot answer the question, which is not the same as a null result. Untested means nobody looked at all.
Inconclusive describes what the studies show rather than how good they are. A question can be marked inconclusive on a well-conducted study that was simply built to answer something else, and the certainty mark is where the quality of the work is recorded.
Why the two marks are kept apart
Study design determines where an assessment begins, and it does not determine where it ends. A randomised trial that measured a surrogate outcome, was funded by the manufacturer of the drug under test and has never been replicated may warrant less confidence than a well-conducted observational study. A single ranking that puts every trial above every cohort would get that case backwards, which is why design here sets the starting point and the specific problems with each study then move the mark.
The reasons why certainty was not rated higher are listed under each entry rather than summarised, so that any judgement on this map can be checked against the studies it came from and disputed.
One rule applied without exception
Evidence resting only on a surrogate outcome cannot be rated above Very low, whatever its design. A surrogate is a measurement that stands in for something the animal experiences: wall thickness in place of heart failure, ectopic counts in place of sudden death, a blood marker in place of survival. A surrogate can improve while the animal is no better off, and in human cardiology drugs have improved the surrogate and increased mortality at the same time. Entries resting on a surrogate are marked Surrogate endpoint only beneath the question, and Surrogate only in the table above. The rule is this site's own, adapted from SORT, the grading scheme used in human primary care.
Summary of findings
The table lists every entry assessed so far, together with the evidence that exists behind it and the assessment that follows from that evidence. Each row links to the full entry below, where the verdict and the specific limitations are set out.
DomainMyxomatous mitral valve diseaseDilated cardiomyopathyFeline cardiomyopathyCongestive heart failureArrhythmias
| Clinical question or practice | Stage | What exists | Assessment |
|---|---|---|---|
| Myxomatous mitral valve disease (dogs) | |||
| Pimobendan in preclinical stage B2 disease | Asymptomatic | 1 randomised trial, 360 dogs, placebo-controlled and double-blind, clinical endpoint | Moderate certaintyPoints to benefit |
| Does preclinical pimobendan shorten survival once heart failure develops? | Asymptomatic | 2 retrospective cohorts, 137 and 143 dogs, no randomisation | Low certaintyInconclusive |
| ACE inhibitors before heart failure | Asymptomatic | 2 randomised placebo-controlled trials, 229 and 124 dogs | Moderate certaintyNo benefit shown |
| Spironolactone with benazepril to delay the onset of heart failure | Asymptomatic | 1 randomised trial, 184 dogs, placebo-controlled | Moderate certaintyNo benefit shown |
| Pimobendan in congestive heart failure | Symptomatic | 1 randomised trial, 260 dogs, active comparator, clinical endpoint | Moderate certaintyPoints to benefit |
| Spironolactone added to standard heart-failure therapy | Symptomatic | 1 randomised trial, 212 dogs, placebo-controlled and double-blind | Low certaintyPoints to benefit |
| Furosemide dose and dosing frequency in stage C disease | Symptomatic | No studies | No evidenceUntested |
| Dilated cardiomyopathy (dogs) | |||
| Pimobendan in preclinical dilated cardiomyopathy (Dobermanns) | Asymptomatic | 1 randomised trial, 76 dogs, placebo-controlled and double-blind | Low certaintyPoints to benefit |
| Screening intervals in at-risk breeds | Asymptomatic | No studies | No evidenceUntested |
| Diet change and taurine in diet-associated dilated cardiomyopathy | Symptomatic | 1 retrospective review and 2 uncontrolled prospective cohorts, 24 to 71 dogs | Very low certaintyInconclusive |
| Feline cardiomyopathy | |||
| Atenolol in preclinical hypertrophic cardiomyopathy | Asymptomatic | 1 prospective observational cohort, 63 cats, not randomised or blinded | Low certaintyNo benefit shown |
| Rapamycin (sirolimus) in subclinical hypertrophic cardiomyopathy | Asymptomatic | 1 randomised trial, 43 cats, surrogate endpoint only Surrogate only | Very low certaintyInconclusive |
| Any therapy that delays the onset of heart failure in hypertrophic cardiomyopathy | Asymptomatic | No study with a clinical endpoint | No evidenceUntested |
| Do glucocorticoids precipitate congestive heart failure? | Asymptomatic | 1 case series, 3 dosing studies, 1 cohort of 1042 cats | Low certaintyPoints against |
| Clopidogrel to prevent recurrent arterial thromboembolism | Symptomatic | 1 randomised trial, 75 cats, double-blind, active comparator, clinical endpoint | Moderate certaintyPoints to benefit |
| Pimobendan in feline congestive heart failure | Symptomatic | Retrospective records only | Very low certaintyInconclusive |
| Congestive heart failure: management | |||
| Torasemide as an alternative to furosemide (dogs) | Symptomatic | 1 randomised non-inferiority trial, 366 dogs, active comparator, 3 months | Low certaintyNo benefit shown |
| Sacubitril–valsartan (ARNi) in congestive heart failure | Symptomatic | 1 retrospective cohort, 50 dogs, no control group | Very low certaintyInconclusive |
| Dietary sodium restriction | Symptomatic | No studies in dogs | No evidenceUntested |
| Exercise restriction after diagnosis | Symptomatic | No studies | No evidenceUntested |
| Arrhythmias | |||
| When to treat ventricular arrhythmias in asymptomatic dogs | Asymptomatic | No outcome data | No evidenceUntested |
| Holter thresholds that predict sudden death | Asymptomatic | Screening cohorts, 28 Dobermanns on seven-day monitoring | Low certaintyInconclusive |
| Rate control in atrial fibrillation: diltiazem with digoxin versus monotherapy | Symptomatic | 1 randomised crossover study, 18 dogs, surrogate endpoint Surrogate only | Low certaintyPoints to benefit |
| Sotalol versus mexiletine–atenolol in Boxer arrhythmogenic cardiomyopathy | Symptomatic | 1 randomised four-arm study, 49 Boxers, surrogate endpoint Surrogate only | Low certaintyPoints to benefit |
Scroll the table sideways for the assessment column.
Table 1. Certainty and direction are recorded as separate marks, and neither of them is derived from study design alone.
The same drugs, in people
Much of small-animal cardiology is extrapolated from human cardiology, and the extrapolation is seldom stated. Human and animal cardiac disease differ enough that a human trial rarely settles a veterinary question. Drug classes, however, travel further than diseases do, and several of the classes used here have been tested in people at a scale veterinary medicine will never reach.
Oral inotropes increased mortality in two placebo-controlled trials. Suppressing asymptomatic ventricular ectopy after myocardial infarction more than doubled it, with no warning on any intermediate measure. Beta blockade in non-obstructive hypertrophic cardiomyopathy reduced exercise capacity against placebo. Glucocorticoids are given in human heart failure deliberately, to improve diuresis, rather than withheld.
A companion page sets out each intervention class in turn, describing what was tested in people, what the trials showed, and how close the analogy to the veterinary question is judged to be. The judgements range from close, through partial, to no analogue at all.
Clinical questions and practices
Showing 24 of 24 entries
Cardiology24 questions
Myxomatous mitral valve disease (dogs)
AsymptomaticScreening, and treating before signs appear
1 randomised trial, 360 dogs, placebo-controlled and double-blind, clinical endpoint
Points to benefit — in stage B2 dogs only, on a composite endpoint, in a trial stopped early and never repeated
EPIC, the single trial of pimobendan in preclinical stage B2 disease, supports delaying heart failure in dogs meeting its entry criteria, and nothing else on this map is better evidenced. It is also more fragile than the headline suggests: nine reclassified dogs remove significance, and on proportions alone the result is not significant. A median difference of 380 days survives the removal of the first 23 weeks, so the benefit cannot be attributed to early failures alone. Whether dogs at the mild end of stage B2 benefit is not answerable from the published data.
EPIC, 2016 · FDA NADA 141-273, 2025 · full appraisal
2 retrospective cohorts, 137 and 143 dogs, no randomisation
Inconclusive — both cohorts report shorter survival, and confounding by indication explains it equally well
Two retrospective cohorts report shorter survival after heart failure in dogs that had been treated beforehand, and the question they raise is a reasonable one to ask. Neither study is able to answer it, because in both of them the pretreated dogs already had larger hearts at the point at which failure developed, and that difference is sufficient on its own to explain the survival gap. The finding argues for a prospective study, not for a change in practice.
McAulay, 2026 · Park, 2025
2 randomised placebo-controlled trials, 229 and 124 dogs
No benefit shown
Two placebo-controlled trials set out to show that ACE inhibition delays heart failure in preclinical mitral valve disease, and neither of them did so. The result should therefore be read as a reasonably firm negative rather than as an absence of evidence. VETPROOF is nonetheless cited as a positive trial on the strength of secondary endpoints that its design cannot carry. The same drug behaved the same way in the equivalent human trial.
SVEP, 2002 · VETPROOF, 2007
* Full appraisal in preparation
1 randomised trial, 184 dogs, placebo-controlled
No benefit shown
DELAY, which tested spironolactone with benazepril before heart failure, failed its primary endpoint, and time to heart failure was numerically worse in the treated dogs. The authors state this plainly; the abstract then rests its conclusion on echocardiographic and NT-proBNP changes, which are surrogates. Read as it was designed, this is evidence that the combination does not delay the onset of heart failure.
DELAY, 2020
SymptomaticOnce the animal is clinically affected
1 randomised trial, 260 dogs, active comparator, clinical endpoint
Points to benefit — compared with benazepril, which was itself never shown to beat placebo in heart failure
The evidence shows pimobendan outperforms benazepril in dogs already in heart failure, and that comparison is sound. It does not show pimobendan outperforms placebo, and benazepril itself was never established against placebo at this stage, so the size of the true benefit is unknown. Owners were not blinded, and part of the composite endpoint was judged by the investigators.
QUEST, 2008
* Full appraisal in preparation
1 randomised trial, 212 dogs, placebo-controlled and double-blind
Points to benefit — on 39 events in a trial two thirds of dogs did not finish
Spironolactone added to standard heart-failure therapy points to benefit, but the result rests on 39 events in a trial that around two thirds of dogs did not complete, and part of the composite endpoint was defined by the investigators. The manufacturer funded the study and employed four of the seven authors. The much larger human trials of the same drug are the stronger reason for using it.
Bernay et al., 2010
* Full appraisal in preparation
No studies
Untested
No study has compared doses or dosing intervals of furosemide in dogs with heart failure. Practice derives from convention and from what worked in the individual animal. This is among the most frequently made decisions in cardiology and among the least examined, although the same gap exists in human medicine, so it is not a failing peculiar to this field.
Dilated cardiomyopathy (dogs)
AsymptomaticScreening, and treating before signs appear
1 randomised trial, 76 dogs, placebo-controlled and double-blind
Points to benefit — on time to event, not on the primary endpoint as originally analysed; Dobermanns only
PROTECT, the preclinical pimobendan trial in Dobermanns, is routinely cited as positive. The proportion of dogs reaching its primary endpoint did not differ; the significant result comes from a time-to-event analysis at an interim look that was not pre-planned. Seventy-six Dobermanns were enrolled from approximately a thousand screened, so the finding is narrow as well as fragile, and it applies to a single breed.
PROTECT, 2012
* Full appraisal in preparation
No studies
Untested
No study has compared screening intervals against any outcome. The intervals in current guidelines are expert judgement, offered as such by their authors and cited afterwards as though established. Whether annual screening detects disease earlier than biennial screening, and whether earlier detection alters the outcome for the dog, both remain unanswered.
SymptomaticOnce the animal is clinically affected
1 retrospective review and 2 uncontrolled prospective cohorts, 24 to 71 dogs
Inconclusive — the improvement cannot be separated from the heart drugs started at the same time
In the diet-associated dilated cardiomyopathy literature, every dog in every study changed diet, so no comparison group exists. Improvement is confounded with the pimobendan and furosemide started at the same time, with taurine supplementation, and with regression to the mean in animals enrolled at their worst measurement. The echocardiograms were read unblinded, by the authors' own account. The one study without industry funding is also the one that states plainly that causation cannot be shown.
Kaplan, 2018 · Freid, 2021 · Freeman, 2022
* Full appraisal in preparation
Feline cardiomyopathy
AsymptomaticScreening, and treating before signs appear
1 prospective observational cohort, 63 cats, not randomised or blinded
No benefit shown
The atenolol evidence in preclinical hypertrophic cardiomyopathy is frequently described as a trial. It is in fact a prospective observational cohort, open-label and not randomised, so the reason a cat was treated cannot be separated from what subsequently happened to it. Within those limits it found no difference in all-cause or cardiac mortality over five years. The human evidence in non-obstructive disease points against treating.
Schober et al., 2013
1 randomised trial, 43 cats, surrogate endpoint only
Inconclusive — one trial reported a wall-thickness difference, on an endpoint the FDA records as selected after exploratory analysisSurrogate endpoint only
RAPACAT is a dose-finding study, described as such by its own authors, that is now carrying a marketing approval. Wall thickness at six months was 7.41 mm on low dose against 8.40 mm on placebo, but the overall treatment effect was not significant and alpha was set at 0.10 across three uncorrected contrasts. The FDA record states that the endpoint was selected after exploratory analysis, which the published paper does not disclose. Wall thickness is in any case a proxy for heart failure rather than a measure of it.
RAPACAT, 2023 · FDA FOI summary 141-604, 2025 · full appraisal
No study with a clinical endpoint
Untested
No study with a clinical endpoint has tested whether any drug delays the onset of heart failure in feline hypertrophic cardiomyopathy. Every trial so far has measured wall thickness instead. Until a trial measures heart failure, survival or quality of life, this question should be regarded as unanswered rather than as answered in the negative, since the two are not the same thing.
1 case series, 3 dosing studies, 1 cohort of 1042 cats
Points against
The teaching that glucocorticoids precipitate congestive heart failure in cats is widespread, and the published evidence does not establish it. The experimental literature is mixed rather than negative: plasma volume expansion has been demonstrated after methylprednisolone acetate, and one study found larger atrial and ventricular dimensions after short oral courses, while two others found nothing. No cat in any of them developed heart failure, and none of them enrolled a cat with cardiomyopathy, which is the population in which the effect is supposed to occur. In the only cohort with a control group, heart failure occurred in 0.82% of treated cats against 1.90% of untreated ones.
Smith, 2004 · Ployngam, 2006 · Khelik, 2019 · Dutch, 2023 · full commentary
SymptomaticOnce the animal is clinically affected
1 randomised trial, 75 cats, double-blind, active comparator, clinical endpoint
Points to benefit — compared with aspirin, in cats that had already thrown one clot
Clopidogrel after arterial thromboembolism carries the strongest independently funded evidence on this map: a double-blind randomised trial, paid for by the Morris Animal Foundation, showing longer time to recurrence than aspirin. Seventy-five cats gives wide confidence intervals, and the comparator was aspirin rather than placebo. In human medicine this drug class is used as the fallback rather than as the preferred option, and no comparison against anticoagulation has been made in cats.
FAT CAT, 2015
* Full appraisal in preparation
Retrospective records only
Inconclusive
Pimobendan in feline congestive heart failure rests on retrospective records alone, with no prospective study and no control group. Cats given the drug were selected by their clinicians for reasons the records do not capture. The usual justification is extrapolation from dogs, and the human trials of oral inotropes are a reason to be careful about extrapolating this particular drug class. The drug is therefore in wide use on the basis of almost no direct evidence.
Congestive heart failure: management
SymptomaticOnce the animal is clinically affected
1 randomised non-inferiority trial, 366 dogs, active comparator, 3 months
No benefit shown — not much worse than furosemide over three months; superiority was not shown
TEST was designed to show that torasemide is not appreciably worse than furosemide, and it showed that, against a non-inferiority margin of minus 20% chosen by the investigators. The survival hazard ratio that is widely quoted from it was a secondary endpoint, analysed after the fact and pooled across two trials that used different dosing rules. Follow-up in both trials extended to three months only.
TEST, 2017
* Full appraisal in preparation
1 retrospective cohort, 50 dogs, no control group
Inconclusive
The evidence for sacubitril–valsartan in dogs is fifty animals, retrospective, with no control group, cases selected by cardiologists, dosing extrapolated from human medicine, and survival compared against expectation rather than against a comparator. No design of that kind can separate a drug effect from case selection. The human evidence for this drug is among the strongest in cardiology, which is a reason to run the trial rather than to skip it.
Carlson & Stern, 2026
No studies in dogs
Untested
No trial in dogs has tested sodium restriction against any outcome. The practice is extrapolated from human medicine, where a randomised trial has since found no effect on death or hospitalisation. Restriction also reduces palatability in animals that are frequently eating poorly already, so the advice carries a cost even if the benefit were real.
No studies
Untested
Exercise restriction is advised almost universally after a cardiac diagnosis, on no evidence in either species. Human patients with heart failure are prescribed supervised aerobic training instead, and it is safe in advanced disease. Restricting exertion in order to prevent sudden death in an arrhythmogenic cardiomyopathy is a different question from restricting it in stable heart failure, and both are equally untested.
Arrhythmias
AsymptomaticScreening, and treating before signs appear
No outcome data
Untested
No study links antiarrhythmic treatment of asymptomatic ventricular ectopy in dogs to survival, in either direction. The same question was asked directly in human medicine and answered against treatment: mortality rose while the Holter recordings improved. Practice rests on the assumption that fewer ectopic beats is better, and that is precisely the assumption which failed when it was tested in people.
Screening cohorts, 28 Dobermanns on seven-day monitoring
Inconclusive — the same dog can fall either side of a cutoff depending on how long it was recorded
The Holter cutoffs used to predict sudden death are screening conventions rather than thresholds validated prospectively against outcome. Whether a dog falls above or below one depends partly on how long it was recorded: six of eleven positive Dobermanns in a seven-day study were identified only after the first day. The same animal can therefore be classified either way by the same test, depending only on the recording duration.
Wess et al., 2017 · Gunasekaran et al., 2020
SymptomaticOnce the animal is clinically affected
1 randomised crossover study, 18 dogs, surrogate endpoint
Points to benefit — for heart rate, which is not the same as feeling better or living longerSurrogate endpoint only
Diltiazem combined with digoxin controls ventricular rate in atrial fibrillation better than either drug given alone, and that much is demonstrated in eighteen dogs studied for two weeks per arm. Whether controlling rate in this way alters clinical signs, survival, or anything an owner would notice has not been studied, because the endpoint used here is a measurement rather than an outcome.
Gelzer et al., 2009
1 randomised four-arm study, 49 Boxers, surrogate endpoint
Points to benefit — for ectopy counts; syncope did not differ between the two effective armsSurrogate endpoint only
Sotalol and mexiletine–atenolol both reduce ventricular ectopy in Boxers, across four arms and forty-nine dogs over three to four weeks. Whether either reduces sudden death is untested, and syncope did not differ between the two effective arms. Suppression of ventricular ectopy is the intervention that increased mortality in the human trials, which is why the distinction between the count and the outcome matters here.
Meurs et al., 2002
Anaesthesia and analgesiaIn progress
Protocol selection, multimodal analgesia and the monitoring thresholds in routine use, across species and procedure types.
AntimicrobialsIn progress
Empirical drug selection, duration of therapy, and the evidence behind the first-line recommendations made in common presentations.
Nutrition and supplementsIn progress
Therapeutic diets, joint supplements and nutraceuticals, a field in which marketing claims and trial evidence diverge more often than in most others.
DermatologyIn progress
Atopic disease, otitis, and the long-term management protocols that rest largely on convention.
OncologyIn progress
Protocol comparisons, adjuvant therapy, and the outcome measures that are used to judge both.
Status
This map is incomplete and remains under revision. Every mark on it is a judgement, and the reasons behind each are listed rather than summarised so that they can be checked and disputed. Corrections and citations that have been omitted are welcome, and the about page explains how to send them.