The same drugs, in people

What the human trials show

Two large placebo-controlled trials tested oral phosphodiesterase-3 inhibition in chronic heart failure and both were stopped for harm. Milrinone increased all-cause mortality by 28% in 1,088 patients with severe heart failure. Vesnarinone increased mortality dose-dependently in 3,833 patients, 22.9% against 18.9%, while improving quality-of-life scores in the same trial. The excess deaths in both were predominantly sudden.

How far it carries across

Milrinone and vesnarinone are phosphodiesterase-3 inhibitors alone. Pimobendan combines phosphodiesterase-3 inhibition with calcium sensitisation, which is argued to raise contractility with a smaller rise in intracellular calcium and myocardial oxygen demand. The trial populations differ as well: those patients had predominantly ischaemic systolic dysfunction, in which the arrhythmic substrate thought to drive the excess mortality is already present. Myxomatous mitral valve disease is a volume-overload lesion in a non-ischaemic ventricle. Those differences are real and they may well be sufficient, but they do not dispose of the human result. The mechanism of harm was never established in the human trials, so the mechanistic argument for exemption remains an assertion rather than a finding. Neither EPIC nor QUEST was designed to detect a mortality difference of that size, since death was one component of a composite primary endpoint in both trials.

What the human trials show

This is the best-populated area of the comparison. Enalapril in 4,228 asymptomatic patients with an ejection fraction of 35% or less cut the combined incidence of death or heart failure by 29%, and heart-failure hospitalisation by 20%, but did not reduce mortality (8% risk reduction, p = 0.30). Spironolactone reduced mortality by 30% in severe heart failure, 35% against 46%, and eplerenone reproduced the effect in mild disease. Sacubitril–valsartan reduced cardiovascular death or heart-failure hospitalisation by 20% and all-cause death by 16% against an active comparator in 8,442 patients.

How far it carries across

Neurohormonal activation is the shared final pathway, and the analogy holds better here than anywhere else on this page. The analogy is not exact, because the human trials enrolled patients with reduced ejection fraction of ischaemic or idiopathic origin, whereas canine heart failure is usually valvular with systolic function preserved until late. Read together, the two literatures say the same thing. Blockade before symptoms alters progression at best and has never shown a survival benefit in either species, which is what SVEP and VETPROOF found in dogs and what SOLVD-Prevention found in people. Blockade after symptoms appear is where benefit has been demonstrated. The difference between the two literatures is one of strength rather than of direction, since the canine spironolactone result rests on 39 events and the human one on 1,663 patients.

What the human trials show

Torsemide and furosemide were compared head to head in 2,859 patients discharged after a heart-failure hospitalisation, with no difference in all-cause mortality (HR 1.02). Dose and route have been tested once, in hospital: DOSE randomised 308 patients with acute decompensated heart failure to high-dose against low-dose intravenous furosemide and to bolus against continuous infusion, and both co-primary endpoints were null: no difference in symptom relief or creatinine for bolus versus infusion, and only a nonsignificant trend favouring the high dose (P = 0.06). Outpatient dose and dosing interval, which is what most prescribing actually is, have not been settled by trial in either species.

How far it carries across

Both species are being asked the same pharmacological question about the same class of drug acting on the same segment of the nephron, and the answers agree with each other, in that the veterinary non-inferiority result says what a trial ten times larger says in people. The gap in dosing evidence is shared between the two fields rather than specific to veterinary medicine, which is worth stating plainly rather than treating as a failure peculiar to this field.

What the human trials show

CAST randomised 1,727 patients with asymptomatic or minimally symptomatic ectopy after myocardial infarction to encainide, flecainide or placebo. The drugs suppressed ectopy as intended. Total mortality was nonetheless 7.7% against 3.0%, a relative risk of about 2.5, and the trial was stopped early. No intermediate measure identified the harm, because the Holter recordings improved throughout.

How far it carries across

At the level of the drug the analogy is weak. Encainide and flecainide are class Ic sodium channel blockers acting on scarred post-infarct myocardium, and the proarrhythmia is attributed to conduction slowing in that substrate. Sotalol and mexiletine belong to other classes, and arrhythmogenic cardiomyopathy in the Boxer is a different substrate again. At the level of method the analogy is exact, and that is the part which transfers. Ectopy count is a surrogate outcome, it was suppressed, it moved in the intended direction, and the patients nonetheless did worse. Any protocol judged on Holter suppression, in any species, is exposed to the same failure, and no amount of suppression can serve as evidence of benefit on its own.

What the human trials show

In obstructive disease, metoprolol reduced outflow gradients and improved symptoms against placebo, and remains a comparator in current trials. In non-obstructive disease a placebo-controlled triple-crossover trial found the opposite, in that bisoprolol reduced exercise capacity and quality of life while increasing NT-proBNP and left atrial size. No trial in either phenotype has tested a hard outcome.

How far it carries across

The diseases are not equivalent. Human hypertrophic cardiomyopathy is a sarcomeric genetic disorder with an established mutation in around half of patients; feline hypertrophic cardiomyopathy is heterogeneous and, outside two breed-specific mutations, of unknown genotype. The distinction that does carry across is haemodynamic rather than genetic. Beta blockade helps the obstructive phenotype by slowing the heart and reducing the gradient, and it appears to harm the non-obstructive phenotype by limiting the rate response on which a stiff ventricle depends. Most cats given atenolol are subclinical and non-obstructive, which is the phenotype in which the human evidence is negative rather than merely absent.

What the human trials show

Clopidogrel with aspirin was compared directly with oral anticoagulation for stroke prevention in atrial fibrillation and was inferior, the trial stopping early (RR 1.44 favouring anticoagulation). Across the wider trial literature, antiplatelet therapy reduced stroke by about 22% and adjusted-dose warfarin by about 64%. In atrial fibrillation, antiplatelet therapy is inferior to oral anticoagulation and is not considered an equivalent substitute for it where anticoagulation is indicated.

How far it carries across

This is the weakest analogy on the page. The human evidence concerns cerebral embolism from the left atrial appendage in atrial fibrillation. Feline arterial thromboembolism is usually distal aortic, arises from an atrium enlarged by cardiomyopathy, frequently in sinus rhythm, and occurs in a species in which anticoagulant monitoring is impractical and the bleeding trade-off is therefore different. What transfers here is a question rather than an answer. The comparison that settled the human field, which set antiplatelet therapy against anticoagulation, has not been made in cats. FAT CAT compared clopidogrel with aspirin, and that is a comparison between two members of the weaker class.

What the human trials show

Lenient rate control, defined as a resting rate below 110 bpm, was not inferior to strict control for a composite of cardiovascular events in 614 patients, and was far easier to achieve.

How far it carries across

The arrhythmia is the same in both species, and the setting is not. Human atrial fibrillation is frequently lone or hypertensive with preserved ventricular function, whereas canine atrial fibrillation usually accompanies advanced structural disease at much higher rates. The targets are not interchangeable either, being derived from Holter averages in dogs and resting rate in people. What survives translation is the burden of proof. Aggressive rate targets were assumed to be better until the assumption was tested, at which point it did not hold.

What the human trials show

Both have been tested in humans and neither supports restriction. Reducing dietary sodium below 100 mmol per day made no difference to death or hospitalisation (HR 0.89, not significant) in 806 randomised patients. Supervised aerobic training in 2,331 patients with chronic heart failure missed its primary endpoint (HR 0.93, p = 0.13) and reached significance only after adjustment for prespecified prognostic variables; it was safe in advanced disease.

How far it carries across

Sodium intake is not comparable between a free-living human adding salt at table and a dog eating a formulated diet, and the achievable range of restriction differs accordingly. Exercise advice needs to be separated in the same way. Where restriction is advised in a dog with an arrhythmogenic cardiomyopathy the aim is to prevent sudden death during exertion, which is a different question from functional capacity in stable heart failure and one that the human exercise trials do not address. In the ordinary case, meaning a dog with stable treated heart failure that is told to stay quiet and fed a sodium-restricted diet, neither practice has a trial behind it in either species, and the human trials that do exist are either neutral or point the other way.

What the human trials show

Two separate literatures exist here, and they point in different directions. Acute exposure: a self-controlled case series of 2,623,327 adults given a single oral burst of fourteen days or less found a heart-failure incidence rate ratio of 2.37 (2.13–2.63) at days 5 to 30, attenuating thereafter, at an absolute rate of 1.3 per 1000 person-years. The commonest indications were skin disorders and respiratory infections. Therapeutic use: prednisone is given deliberately in heart failure to augment diuresis, and a randomised pilot in acute heart failure with raised C-reactive protein reported fewer events at 90 days (HR 0.31). The mechanism has been measured. Six normal subjects given five days each of prednisolone, methylprednisolone, triamcinolone and dexamethasone showed no weight gain, increased urinary sodium excretion, and plasma volume unchanged on radiolabelled albumin. The volume effect belongs to cortisol, not to the synthetics, although methylprednisolone pulses do suppress urinary sodium for up to 32 hours, with escape.

How far it carries across

The acute association is the closest human analogue to the feline claim, and it is confounded in exactly the same way, because people receive steroid bursts on account of being acutely unwell, and respiratory infection is itself a common precipitant of decompensation. The drug and the precipitant arrive together, just as illness, handling and stress arrive together in the cat. The one population where that confounding was tested directly settles it as far as human data can. Among 11,356 patients in acute heart failure, those newly started on a corticosteroid had worse crude outcomes “essentially because corticosteroid-treated patients were sicker”, and the difference disappeared on adjustment. Where allocation is random, nothing appears: 584 patients randomised to methylprednisolone or placebo for severe pneumonia had cardiovascular complications, on a definition explicitly including new or worsening congestive heart failure, in 4% against 5% by day 28. The transfer is therefore partial in a specific way. The mechanism proposed in cats is not supported, because plasma volume was measured after the same drug classes used in cats and did not change. The association transfers completely, and it carries its confounding with it.

What the human trials show

There is no human trial of rapamycin or an analogue in hypertrophic cardiomyopathy. The randomised human data concern everolimus and drug-induced left ventricular hypertrophy in transplant recipients, where the hypertrophic stimulus is a calcineurin inhibitor that is being withdrawn at the same time. The results are small and mixed and the design confounds the question.

How far it carries across

There is nothing here to extrapolate from in either direction. This is the one question on the map on which the veterinary evidence leads the human literature, which is a statement about human medicine and not an endorsement of the veterinary trial. A single 43-cat study reporting an effect below the measurement noise of its own endpoint is weak evidence whether or not anyone else has done better.