Pet Connection has an update. Their database of self-reported cases now stands at over 1500 sick or dead pets, half of whom are deceased. What does this mean? How useful are these numbers?
The data are self-reported, so we can’t be as confident about their significance as we could if they were collected by a randomized survey or in some other more rigorous manner. Potential problems with self-reported data include
- omissions (One primary problem with surveys in which participants are self-selected is a bias in the survey results toward those who have strong feelings about the issue being surveyed. People who believe their pets have died because of the recalled food will be far more motivated to participate in the Pet Connection survey than those whose pets merely got sick. And would participation even occur to people whose pets ate recalled food but didn’t get sick? These factors need to be considered when evaluating the Pet Connection data.)
- exaggeration (People with multiple pets eating multiple foods might have reasoned that because all their pets had access to affected food that all of them necessarily ate affected food. Such people would have reported in good faith, but their reports are potentially exaggerated.)
- mistakes and faulty assumptions (On Tuesday I had an e-mail from a client wondering if her cat’s death late last year might have been caused by now-recalled food. She thought that her cat had died of kidney failure. I checked the cat’s history and found that we had, in fact, treated it for liver disease and that the cat had died in late November, before the recalled food was even manufactured. Had this client gone straight to Pet Connection instead of contacting us, she would have entered erroneous data based on what she thought had happened and her faulty memory of exactly when it happened. I’m not suggesting that most or even many pet owners did this, but the possibility is good that some did, and we should keep that possibility in mind. A more likely confusion will be between chronic and acute renal failure. CRF is quite common, especially among older cats. Some percentage of pet owners will hear ARF when their veterinarian has found CRF. Some veterinarians might mistake CRF for ARF in new patients that already have CRF but that are having an acute episode or have decompensated. Some pet owners will have pets with undiagnosed CRF that decompensate and die at home, and the owner will conclude that the death was from ARF caused by eating recalled foods.)
- fabrication (I don’t think this is likely a large problem, but it should be considered.)
That said, the data are striking even if the collection method is not perfect. But even if we accept (for argument’s sake) that the data collected are flawless, we need to ask a different set questions as we attempt to assess their significance:
- What are the baseline rates of ARF in cats and dogs? In other words, in a given week how many animals typically become ill with acute renal failure? My wife indicates that ARF is considered uncommon in cats. Still, veterinarians have only limited data to support that conclusion. Morbidity and mortality data are not routinely collected in veterinary medicine. Often the best guesses about how frequently a disease or illness occurs are nothing more than guesses. So, unfortunately, we really don’t have a good baseline with which to compare the Pet Connection data. Based on what we think we know (ARF is uncommon), the data are stiking. But with no real point of comparison, we can’t be sure that the apparent significance of the data isn’t really an illusion caused because we tend to look at new data from within our existing frame of reference. Here’s an example: My wife might see two or three acute renal failures per year, or one every four to six months. By comparison, the numbers reported by Pet Connection seem shocking. But let’s do some math. If my wife’s rate of two to three per year is extended across all the veterinarians in North America, you have 80,000 to 120,000 cases (I’m assuming a conservative 40,000 vets; the U.S. Department of Labor says there were 61,000 in 2004, but that figure includes industry, government, and other vets who don’t practice on small animals), or approximately 225 to 325 cases each day! In the five days since the recall was announced, we’d therefore expect to see 1,100 to 1,600 cases. In the real world, cases aren’t spread out that evenly, though, which leads to the next question.
- When during the year do we normally see spikes in ARF? Are the data in line with or in excess of those spikes? In cats, we often see a spike in acute renal failure in late winter and early spring. This is probably related to the fact that Easter lilies start showing up in people’s houses around this time. All parts of the lily are toxic to cats and cause ARF. In both cats and dogs, rates of ethylene glycol–related kidney failure increase in the winter as people use more antifreeze in their vehicles. I have no information on how large the winter/early spring spike in acute renal failure is, but for argument’s sake let’s say it’s 10 percent. Returning to my earlier example, that would mean an additional 110 to 160 cases over the five-day period since the recall began. That brings us to 1,110 to 1,760 cases.
- What percentage of acute renal failure patients die? The prognosis is generally regarded as poor. This communication from Tufts claims a survival rate of 41 percent for dogs with ARF that received dialysis, the gold standard for treatment. In human beings estimates of the mortality rate for ARF range widely, from 25 to 90 percent. The estimate for hospitalized patients is 40 to 50 percent. Applying the optimistic 40 percent mortality rate to my previous example, we would expect to see between 440 and 700 deaths over the five day period since the recall began.
As you can see, these numbers are in line with what Pet Connection is reporting. Does this mean their findings aren’t significant? Not necessarily.
If anything, the Pet Connection numbers are an underreport. Many pet owners and veterinarians will have been unaware of Pet Connection. Gina has several times requested that site visitors who have told their pets’ stories on the blog also add the info to the registry. Undoubtably some of those people never did so. I think we can feel confident that the Pet Connection numbers are only a piece of the pie. How small a piece is the question.
Second, the assumptions I’ve used in trying to compare the Pet Connection data with what might be expected to occur normally could be way off. Is my wife’s experience of two to three ARF cases per year typical, higher than average, or lower than average? Estimates of poor prognosis in renal failure are mainly based on the fact that presenting cases are often already quite ill. Awareness of the recall and prophylactic measures by veterinarians and pet owners may have caught many potentially serious ARF cases early enough that the exisiting mortality estimates don’t apply.
The point of all this is not to criticize the good folks at Pet Connection. The work they have done to compile the data and to promote awareness of the recall is greatly appreciated and will doubtless save pets’ lives by making people more aware that a serious problem exists. Their data will prove of great value down the road, too, as researchers try to answer what happened, how widespread the problem was, and other questions. Finally, the data do indicate that a serious problem is occuring and that pets are dying. But members of the veterinary commuity have an obligation to think critically, to step back and try to assess the situation. I am far from the best person to do that, but I hope these initial efforts on my part will encourage others who are more qualified to share their thoughts.