Given the lack of meaningful regulation in the area of veterinary diagnostic testing, the burden of ensuring clinical usefulness falls on the manufacturers and the end user (i.e. the veterinarian). It will come as no surprise to readers of this column that I have concerns about how the veterinary profession develops and uses diagnostic tests. I have published several journal articles1,2 and columns3-5 discussing the principles of testing and the evidence concerning specific tests on the market. I believe the following principles should, but sadly often do not, guide our use of diagnostic tests: A prioritized list of realistic differential diagnoses based on history and physical exam should inform the selection and interpretation of any lab tests. Results should be understood as shifting the probability of a diagnosis up or down, not as determining the diagnosis. Screening tests (used in individuals with no clinical signs of symptoms to detect occult disease) should be used sparingly and cautiously, only when there is good evidence to demonstrate that early detection will benefit the patient. Too much testing is at least as harmful as too little. All vets should understand the concepts of positive and negative predictive value and overdiagnosis. All too often, companies develop and promote diagnostic tests in the veterinary field with no reasonable evidence that using them will benefit patients. Given the lack of meaningful regulation in this area, completely useless tests based on pseudoscience are freely sold to vets and pet owners.4 Even tests that work as claimed may become widespread well before there is evidence to believe their use will improve patient outcomes. The measurement of symmetric dimethyl-arginine (SDMA), for example, clearly detects the decreased glomerular filtration rate of chronic renal disease (CKD) in dogs sooner than measurement of creatinine. Widespread adoption of this test and inclusion of it in clinical practice guidelines has almost certainly resulted in a higher rate of diagnosis of dogs with early (IRIS Stage 1) CKD. However, despite the use of this test over several years, it is still unclear whether detection and treatment of dogs with Stage 1 CKD improves survival or other meaningful clinical outcomes. Are we giving dogs with early-stage CKD longer, healthier lives, or are we just diagnosing more dogs with CKD? More research is needed to answer this question. Figure 1: Lead time bias increases perceived survival following earlier diagnosis even without a true improvement in the time from disease onset to death. Looking closer The phenomenon of lead-time bias inevitably makes it look like patients diagnosed with a fatal disease live longer if diagnosed earlier, but this can be a statistical illusion (Figure 1). The length of time between diagnosis and death is necessarily longer if we make the diagnosis earlier, regardless of whether the patient lives longer because of diagnosis and treatment. For most tests in veterinary medicine intended for early or pre-symptomatic diagnosis, the necessary research to prove a true improvement in patient outcomes has rarely been done. I recently looked at a validation study of a test intended for both screening asymptomatic dogs and aiding in diagnosing symptomatic dogs with cancer.6yet there are no established screening paradigms for early detection. Liquid biopsy methods that interrogate cancer-derived genomic alterations in cell-free DNA in blood are being adopted for multi-cancer early detection in human medicine and are now available for veterinary use. The CANcer Detection in Dogs (CANDiD This study demonstrates both some of the problems associated with novel and more sensitive diagnostic tests in veterinary medicine and some of the potential solutions to these problems. The test is a “liquid biopsy” for detecting circulating biomarkers of cancer. In this example, cell-free DNA with genomic alterations characteristic of neoplastic cells is detected in blood. Such a test could allow early, more accurate, and less-invasive diagnosis of neoplasia in dogs. This has significant potential benefits to patients and their owners, though, as always, the devil is in the details. The design of this case-control study was a reasonable balance between the ideal and the practical, and the size of the sample was quite large. Of course, single studies are virtually never sufficient in themselves to support significant changes in clinical practice or adoption of radical new approaches. The authors indicate additional studies are underway, which is encouraging. The overall sensitivity of this test (the proportion of dogs with cancer testing positive) in detecting cancer of any type was 54.7 percent. This improved to more than 80 percent for some common types of cancer (e.g. hemangiosarcoma and lymphoma). In general, the test was better able to detect aggressive and disseminated cancers, as one would expect given such cancers are more likely to shed DNA than more indolent neoplasms. The specificity (proportion of dogs without cancer correctly testing negative) was 98.5 percent. Though one might initially be disappointed the sensitivity was so low, this is actually exactly what we want in a test intended to screen asymptomatic dogs for a serious disease. We would much rather miss a case, which we might be able to catch later when the dog develops symptoms, than decide a dog has cancer when it does not, which could lead to dangerous and expensive unnecessary testing and treatment or even euthanasia. I was particularly encouraged by the authors’ discussion of positive and negative predictive value (PPV and NPV). As a clinician, I care a lot less about the sensitivity and specificity of a diagnostic test than about how likely it is a dog testing positive actually has cancer and one testing negative does not, and this is what PPV and NPV tell me. This depends not only on how accurate the test is, but on the prevalence of the disease in the population I am testing. The more common a disease, the more reliable a positive test is, and the less common a disease, the more likely a negative test means the dog really doesn’t have the disease. Unfortunately, we often do not know how common diseases are in our patient population (the prevalence), and the authors had to rely on some “cocktail napkin” calculations. The actual numbers they came up with are reasonable, but also very likely not accurate for many individual veterinary practices and their local populations. Still, if we use these guesses to estimate what will happen if veterinarians start using this test in practice, we can see some of the potential problems. The authors estimate an overall cancer prevalence of 5.5 to 9.25 percent in the dog population in the U.S., which is about 65 to 77 million dogs. This gives a PPV of 75 to 80 percent and an NPV of 95 to 96 percent. This means dogs testing negative very likely do not have cancer at the moment, and they almost certainly do not have aggressive or disseminated cancer. However, it also means up to 25 percent of dogs testing positive for cancer do not actually have it! If we test even 10 percent of dogs, hundreds of thousands of them could test positive that do not actually have cancer! This risk only increases when the test is repeated yearly, as the company recommends. The cost, anxiety, and unnecessary treatment based on this test could be huge. Avoiding this harm would require better data and very judicious use of the test. The company does offer some guidance on how to interpret and react to results of its test, but, of course, they cannot guarantee veterinarians and owners will respond appropriately to the information provided by their test. Issues with overdiagnosing Figure 2: Possible outcomes of cancer screening. From Løberg et al (2015)8 Another issue not discussed in this report is the subject of overdiagnosis. Overdiagnosis is different from a false positive or mistaken diagnosis. It is the correct diagnosis of disease that will never cause clinical harm or death. Screening of asymptomatic individuals turns out to find a lot of disease, including cancer, which is indolent and clinically irrelevant. The authors of this study refer to mammography and prostate specific antigen (PSA) testing as examples of successful cancer screening in humans. They fail to mention use of both has been significantly reduced due to recognizing overly aggressive use of these tests led to significant overdiagnosis and often more patient harm than benefit. The nature of a liquid biopsy is it is more likely to detect aggressive or disseminated cancer since those are the concerns more likely to shed cell-free DNA. This makes the risk of overdiagnosis lower than with other methods, but does not eliminate it entirely.7 Given the widespread lack of awareness of overdiagnosis in the veterinary profession, this risk should be further investigated and addressed. Figure 2 provides an illustration of both positive and negative outcomes that can result from screening for cancer. A potentially greater problem with early detection of cancer through liquid biopsy is there is little to no effective treatment for many canine neoplasms. Early detection of diffuse B-cell lymphoma, for example, might lead to better outcome than later treatment or it might not, but at least we have some reasonably effective therapies. How much benefit is early detection of hemangiosarcoma, heart base tumors, acute leukemia, histiocytic sarcoma, etc., going to provide to our patients? We not only do not know if earlier treatment will improve outcome compared to treatment after symptoms occur, we often do not have any truly effective treatments at all! Do we want to tell owners of these dogs they have incurable cancer before a pet is even sick? Will this make anyone’s life better? Bottom line Given the near complete lack of meaningful regulation in the area of veterinary diagnostic testing, the burden of ensuring tests are accurate and clinically useful falls on the manufacturers and the end users—veterinarians in practice. Proper validation and assessment of the overall benefits and harms of testing, especially screening, is a lengthy and expensive process. The economic reality is we can rarely expect companies to invest years and millions of dollars in proving their products do (or do not) work before marketing when they are not required to do so. Veterinarians, unfortunately, often do not have the training or necessary data to effectively evaluate the risks and harms of specific tests. The economic and psychological incentives are nearly all aligned to encourage more testing, whether or not this is in the best interests of patients. Liquid biopsy is a promising technologic with real potential benefits. The company offering this test has done more than most to generate needed data and to encourage appropriate and responsible use of their test. However, much more data is needed to determine how such testing benefits and harms patients and what the balance between these impacts actually is. More evidence is needed to determine when earlier therapy is more effective and when it is not for particular cancers. And better therapies, of course, are needed to make diagnosis of many cancers more than an inescapable death sentence passed a bit earlier, before the patient is actually ill. The desire to find cancers earlier and save dogs who might otherwise die if not diagnosed until their disease was more advanced is an understandable and laudable one. This new test will almost certainly have this benefit for some dogs. However, the chance it might lead to harm for as many or more dogs is not one we can ignore or dismiss without stronger evidence than we currently have. While this evidence is being produced, veterinarians should be very cautious about screening apparently healthy dogs, and about subjecting patients to aggressive and invasive testing and treatment based on screening test results. Brennen McKenzie, MA, MSc, VMD, cVMA, discovered evidence-based veterinary medicine after attending the University of Pennsylvania School of Veterinary Medicine and working as a small animal general practice veterinarian. He has served as president of the Evidence-Based Veterinary Medicine Association and reaches out to the public through his SkeptVet blog, the Science-Based Medicine blog, and more. He is certified in medical acupuncture for veterinarians. Columnists’ opinions do not necessarily reflect those of Veterinary Practice News. References McKenzie BA. Rational use of diagnostic and screening tests. J Small Anim Pract. June 2021:jsap.13393. doi:10.1111/jsap.13393 McKenzie BA. Overdiagnosis. J Am Vet Med Assoc. 2016;249(8). doi:10.2460/javma.249.8.884 McKenzie B. Why do we run diagnostic tests? Vet Pract News. February 2018:38. McKenzie B. Is there a gold-standard test for adverse food reactions? Vet Pract News. February 2019:40. McKenzie B. Fecal microbiome testing: Is the research reliable to ensure a healthy pet? Vet Pract News. May 2020:32. https://www.veterinarypracticenews.com/fecal-microbiome-may-2020/. Flory A, Kruglyak KM, Tynan JA, et al. Clinical validation of a next-generation sequencing-based multi-cancer early detection “liquid biopsy” blood test in over 1,000 dogs using an independent testing set: The CANcer Detection in Dogs (CANDiD) study. Bauer JA, ed. PLoS One. 2022;17(4):e0266623. doi:10.1371/journal.pone.0266623 Welch HG. Liquid biopsy: misplaced faith in early cancer detection? STAT News. August 2021. Løberg M, Lousdal ML, Bretthauer M, Kalager M. Benefits and harms of mammography screening. 2015. doi:10.1186/s13058-015-0525-z
