# Likelihood Ratios in Veterinary Medicine ## Quick Answer - Likelihood ratios (LRs) convert a diagnostic test result into a quantitative update of disease probability, moving interpretation beyond simple positive or negative classifications. - Calculate the positive LR as sensitivity divided by (1 minus specificity) and the negative LR as (1 minus sensitivity) divided by specificity, then apply these values to pretest probability using a Fagan nomogram. - LRs are population independent, but they do not reflect absolute test accuracy, so sensitivity and specificity must still be reported alongside them for complete test evaluation. ## Understanding Likelihood Ratios in Veterinary Diagnostics Veterinary clinicians routinely face the challenge of interpreting diagnostic test results in patients where disease prevalence, test accuracy, and clinical presentation all influence the final diagnosis. Traditional binary thinking, where a test is simply positive or negative, fails to capture the diagnostic information contained in the full spectrum of possible results. Likelihood ratios address this limitation by quantifying how much a particular test result changes the odds of disease in an individual patient. The diagnostic likelihood ratio represents a fundamental shift in how test accuracy is measured and applied. According to a 2021 review in the Revue Scientifique et Technique, diagnostic sensitivity and specificity report classification accuracy in infected and non-infected individuals separately, but they do not directly compare the likelihood of true versus false results across both groups. Predictive values combine these likelihoods but depend heavily on disease prevalence in the tested population, which prevents generalization across different clinical settings. The likelihood ratio balances the likelihoods of true versus false results and remains population independent, making it a more universal measure of diagnostic test accuracy. For veterinary practitioners, this means that a likelihood ratio calculated from a published study can be applied to their own patient population even when the prevalence of disease differs from the study population. This property distinguishes LRs from predictive values, which shift dramatically as disease prevalence changes. A test with a positive predictive value of 90 percent in a referral hospital may have a positive predictive value of only 50 percent in a general practice setting, but the likelihood ratio remains constant across both settings. The practical value of LRs extends beyond simple test interpretation. In feline infectious peritonitis (FIP) diagnosis, the [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) emphasize that understanding each diagnostic test's sensitivity, specificity, predictive value, likelihood ratio, and diagnostic accuracy is essential when building a case for FIP. The guidelines note that FIP can be challenging to diagnose because of the lack of pathognomonic clinical signs or laboratory changes, especially when no effusion is present. Nearly every small animal veterinary practitioner will see cases of FIP, and the disease accounts for an estimated 0.3 to 1.4 percent of feline deaths at veterinary institutions, with young cats under two years of age especially vulnerable. The clinical reasoning process in FIP diagnosis illustrates the value of sequential likelihood ratio application. The guidelines describe building the index of suspicion "brick by brick," where each diagnostic test result updates the probability of disease. This iterative process is precisely what likelihood ratios enable in quantitative form. Instead of simply accumulating positive and negative results, the clinician can calculate how each result shifts the probability of FIP in an individual cat. ## The Mathematical Foundation of Likelihood Ratios ### Defining Sensitivity and Specificity Before applying likelihood ratios, the clinician must understand the component measures from which they are derived. Diagnostic sensitivity (DSe) represents the proportion of truly infected or diseased individuals that test positive. Diagnostic specificity (DSp) represents the proportion of truly non-infected or disease-free individuals that test negative. These measures describe test performance in each population separately, but they do not directly compare the likelihood of a true result versus a false result. Consider a hypothetical veterinary test for a metabolic disease. If the test has 90 percent sensitivity, it correctly identifies 90 of every 100 diseased animals. If it has 85 percent specificity, it correctly identifies 85 of every 100 healthy animals. These figures describe the test's behavior in each group independently, but they do not tell the clinician how to interpret a positive result in an individual patient. The interpretation depends on the pretest probability of disease, which varies with signalment, history, physical examination findings, and population prevalence. ### Calculating Positive and Negative Likelihood Ratios The positive likelihood ratio (LR+) answers the question: how much more likely is a positive test result in a diseased animal compared to a healthy animal? It is calculated as: LR+ = Sensitivity / (1 - Specificity) The negative likelihood ratio (LR-) answers the question: how much less likely is a negative test result in a diseased animal compared to a healthy animal? It is calculated as: LR- = (1 - Sensitivity) / Specificity A positive likelihood ratio of 1.0 indicates that a positive test result provides no diagnostic information, because the result is equally likely in diseased and healthy animals. Values above 1.0 increase the probability of disease, with higher values providing stronger evidence. A negative likelihood ratio of 1.0 similarly indicates no diagnostic information. Values below 1.0 decrease the probability of disease, with lower values providing stronger evidence against disease. The [2021 review in the Revue Scientifique et Technique](https://pubmed.ncbi.nlm.nih.gov/34140723) notes that as a relative measure, LR ignores the absolute accuracy of tests, and two tests with different accuracy profiles may have the same LR. This limitation can be mitigated by using complementary measures of accuracy, including diagnostic sensitivity and specificity, or ancillary selection criteria. For example, a test with 99 percent sensitivity and 90 percent specificity has an LR+ of 9.9, while a test with 50 percent sensitivity and 95 percent specificity has an LR+ of 10.0. Despite similar positive likelihood ratios, these tests behave very differently in clinical practice, and the choice between them depends on whether missing a diagnosis or generating false positives carries greater clinical risk. ### Converting Probability to Odds and Back Likelihood ratios operate on odds instead of probabilities, which requires the clinician to convert between these two scales. The relationship is: Odds = Probability / (1 - Probability) Probability = Odds / (1 + Odds) The post-test odds are calculated by multiplying the pretest odds by the likelihood ratio: Post-test Odds = Pretest Odds x LR The post-test probability is then derived by converting the post-test odds back to probability. While these calculations can be performed manually, the Fagan nomogram provides a graphical method that avoids arithmetic errors. The nomogram consists of three vertical scales: pretest probability on the left, likelihood ratio in the center, and post-test probability on the right. A straight line drawn from the pretest probability through the likelihood ratio intersects the post-test probability scale at the updated value. ## Applying Likelihood Ratios with the Fagan Nomogram ### Constructing and Reading the Nomogram The Fagan nomogram transforms the mathematical calculation into a visual tool that can be used at the point of care. The left vertical axis represents pretest probability, ranging from 0.1 percent at the bottom to 99 percent at the top. The center axis represents the likelihood ratio on a logarithmic scale, ranging from 0.01 to 100. The right vertical axis represents post-test probability on the same scale as the left axis. To use the nomogram, the clinician first estimates the pretest probability of disease based on signalment, history, physical examination findings, and population prevalence. A straight edge is then placed from this point on the left axis through the likelihood ratio value on the center axis. The intersection with the right axis gives the post-test probability of disease. For example, consider a dog suspected of hyperadrenocorticism (HAC) based on polyuria, polydipsia, panting, and a distended abdomen. The clinician estimates the pretest probability at 50 percent based on the strength of these clinical signs. A low-dose dexamethasone suppression test (LDDST) is performed, and the result shows lack of suppression. According to a [2021 study in Veterinary Clinical Pathology](https://pubmed.ncbi.nlm.nih.gov/33728722), the positive likelihood ratio for the lack of suppression pattern to diagnose HAC was infinite, meaning that this pattern essentially confirms the diagnosis. On the nomogram, a line from 50 percent pretest probability through an infinite LR+ intersects the post-test probability scale at essentially 100 percent. If the same dog instead shows a partial suppression pattern, the study reported an LR+ of 8.09. Drawing a line from 50 percent pretest probability through 8.09 on the likelihood ratio axis gives a post-test probability of approximately 89 percent. This quantitative update helps the clinician decide whether additional testing is warranted or whether treatment can proceed. ### Interpreting Likelihood Ratio Magnitudes Clinical interpretation of likelihood ratio values follows established conventions. An LR+ greater than 10 provides strong evidence to rule in disease, while an LR+ between 5 and 10 provides moderate evidence. An LR+ between 2 and 5 provides weak evidence, and an LR+ between 1 and 2 provides almost no change in probability. For negative likelihood ratios, an LR- less than 0.1 provides strong evidence to rule out disease, while an LR- between 0.1 and 0.2 provides moderate evidence. An LR- between 0.2 and 0.5 provides weak evidence, and an LR- between 0.5 and 1.0 provides almost no change. These thresholds are interpretive conventions instead of absolute rules. The clinical context determines whether a given likelihood ratio is sufficient to change management. In a disease with a grave prognosis and effective treatment, the clinician may accept a lower LR+ threshold for initiating therapy. In a disease with significant treatment side effects, a higher LR+ threshold may be required. ### Case Example: Canine Hyperadrenocorticism The [2021 Veterinary Clinical Pathology study](https://pubmed.ncbi.nlm.nih.gov/33728722) examined LDDST response patterns in dogs suspected of HAC. The study population included 115 dogs diagnosed with HAC (54 percent) and 62 dogs with non-adrenal illness (46 percent). Cortisol concentrations of at least 27.59 nmol/L (at least 1 microgram/dL) eight hours after dexamethasone administration were considered positive results regardless of the pattern observed. The study calculated likelihood ratios for different response patterns. The lack of suppression pattern had an infinite positive likelihood ratio, strongly supporting a diagnosis of HAC. The partial suppression pattern had an LR+ of 8.09 with a 95 percent confidence interval of 2 to 32.72, moderately increasing the likelihood of HAC. The escape pattern had an LR+ of 3.23 with a 95 percent confidence interval of 0.75 to 14, but the study found no association between the escape pattern and a diagnosis of HAC, which does not support its integration into decision making. The inverse pattern had an LR+ of 0.2 with a 95 percent confidence interval of 0.06 to 0.73, meaning that this pattern decreases the likelihood of HAC. This study demonstrates the value of moving beyond binary test interpretation. A dog with an inverse pattern on LDDST is less likely to have HAC than a dog with a partial suppression pattern, even though both results would be classified as "positive" under the standard cortisol threshold. The likelihood ratio captures this diagnostic information that binary classification loses. ## Likelihood Ratios in Feline Infectious Peritonitis Diagnosis ### The Diagnostic Challenge of FIP Feline infectious peritonitis represents one of the most important infectious diseases and causes of death in cats, with young cats less than two years of age especially vulnerable. The [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) note that FIP is caused by a feline coronavirus (FCoV) and that the disease can be challenging to diagnose because of the lack of pathognomonic clinical signs or laboratory changes, especially when no effusion is present. The diagnostic approach to FIP requires the clinician to answer two questions before proceeding with any diagnostic test or commercial laboratory profile: why this test, and what do the results mean? The guidelines emphasize that the approach to diagnosing FIP must be tailored to the specific presentation of the individual cat. The clinician must consider the individual patient's history, signalment, and comprehensive physical examination findings when selecting diagnostic tests and sample types. ### Building the Index of Suspicion Brick by Brick The guidelines describe building the index of suspicion "brick by brick," which is a qualitative description of the sequential probability updating that likelihood ratios enable quantitatively. Each diagnostic test result, whether positive or negative, adds information that shifts the probability of FIP in the individual cat. Consider a young cat presenting with fever, lethargy, and abdominal distension. The clinician estimates a pretest probability of FIP at 30 percent based on signalment and clinical signs. A serum albumin-to-globulin ratio below 0.6 has been associated with FIP, and if this result is obtained, the likelihood ratio for this finding shifts the probability upward. If the cat develops effusion, analysis of the effusion fluid provides additional diagnostic information. The guidelines note that a good understanding of each diagnostic test's sensitivity, specificity, predictive value, likelihood ratio, and diagnostic accuracy is important when building a case for FIP. The sequential application of likelihood ratios allows the clinician to determine when the probability of FIP is high enough to justify treatment or when additional testing is needed. Given that the disease is fatal when untreated, the ability to obtain a correct diagnosis is critical. The guidelines note that research has demonstrated efficacy of new antivirals in FIP treatment, but these products are not legally available in many countries at this time. ### Limitations of Likelihood Ratios in FIP Diagnosis The FIP diagnostic process illustrates several limitations of likelihood ratios in clinical practice. First, the likelihood ratios for individual tests are often derived from studies with specific patient populations, and these values may not apply directly to other populations. Second, the pretest probability estimate is subjective and depends on the clinician's experience and the specific clinical context. Third, the sequential application of likelihood ratios assumes that test results are conditionally independent given disease status, which may not hold when multiple tests measure related biological pathways. The guidelines emphasize that the clinician must be able to answer the questions of why this test and what do the results mean before proceeding with any diagnostic test. This emphasis on test selection and interpretation reflects the reality that likelihood ratios are only as useful as the clinician's understanding of the test and the clinical context in which it is applied. ## Likelihood Ratios in Cardiac Risk Stratification ### The MINE Score in Preclinical Mitral Valve Disease The application of likelihood ratios extends beyond infectious disease diagnosis into prognostic assessment. A [2025 study in the Journal of Veterinary Internal Medicine](https://pubmed.ncbi.nlm.nih.gov/40865020) examined the Mitral INsufficiency Echocardiographic (MINE) score for risk stratification in dogs with preclinical myxomatous mitral valve disease (MMVD). The study included 749 dogs with preclinical MMVD in a retrospective, multicenter, cohort study. The primary aim was to verify the efficacy of the MINE score in stratifying cardiac risk in preclinical MMVD. The study evaluated the association between the MINE score and median time to cardiac event using Cox proportional hazards regression. Based on multivariate analysis, a simplified version of the MINE score was redefined to include only the left atrium-to-aorta ratio, the left ventricular end-diastolic diameter, and the E-wave velocity. The study found that mild cases had a longer median time to cardiac event of 2604 days with a 95 percent confidence interval of 2344 to 2604 days, compared to moderate cases at 1216 days with a 95 percent confidence interval of 998 to 1882 days, and severe cases at 718 days with a 95 percent confidence interval of 599 to 980 days. Among stage B2 dogs, severe cases had a shorter median time to cardiac event of 718 days compared to moderate cases at 1141 days and mild cases where the median was not available. ### Risk Stratification as Probability Updating The MINE score approach demonstrates how likelihood ratios and related probability measures can be applied to prognostic questions. Instead of asking whether a dog has disease, the clinician asks what the probability is of a cardiac event within a given time frame. The echocardiographic measurements that comprise the MINE score update this probability in a manner analogous to likelihood ratio application. The study proposed a definition of "advanced B2" for dogs in stage B2 classified as severe. This classification has direct management implications, as dogs with advanced B2 may warrant earlier intervention or more frequent monitoring. The simplified version of the MINE score was clinically effective for risk stratification of preclinical MMVD in the study cohort. The prognostic application of likelihood ratios requires careful attention to the outcome being predicted. In the MINE score study, the outcome was time to cardiac event, and the likelihood ratios would be calculated for the probability of experiencing an event within a specified time frame given a particular score category. This application differs from diagnostic likelihood ratios, which address the probability of current disease. ## At a Glance: Likelihood Ratio Interpretation | Likelihood Ratio Value | Clinical Interpretation | Clinical Action | |------------------------|------------------------|-----------------| | LR+ greater than 10 | Strong evidence to rule in disease | Consider definitive diagnosis or treatment | | LR+ between 5 and 10 | Moderate evidence to rule in disease | Consider additional confirmatory testing | | LR+ between 2 and 5 | Weak evidence to rule in disease | Interpret in context of other findings | | LR+ between 1 and 2 | Minimal change in disease probability | Limited diagnostic value | | LR- less than 0.1 | Strong evidence to rule out disease | Consider alternative diagnoses | | LR- between 0.1 and 0.2 | Moderate evidence to rule out disease | Consider additional testing if suspicion remains | | LR- between 0.2 and 0.5 | Weak evidence to rule out disease | Interpret in context of other findings | | LR- between 0.5 and 1.0 | Minimal change in disease probability | Limited diagnostic value | ## LDDST Response Patterns and Likelihood Ratios in Canine Hyperadrenocorticism | Response Pattern | LR+ (95% CI) | Diagnostic Implication | |------------------|--------------|------------------------| | Lack of suppression | Infinite | Strongly supports HAC diagnosis | | Partial suppression | 8.09 (2 to 32.72) | Moderately increases HAC likelihood | | Escape | 3.23 (0.75 to 14) | No confirmed association with HAC | | Inverse | 0.2 (0.06 to 0.73) | Decreases HAC likelihood | Data from the [2021 Veterinary Clinical Pathology study](https://pubmed.ncbi.nlm.nih.gov/33728722) on LDDST response patterns in dogs suspected of hyperadrenocorticism. ## Practical Workflow for Applying Likelihood Ratios ### Step 1: Estimate Pretest Probability The first step in applying likelihood ratios is estimating the pretest probability of disease for the individual patient. This estimate integrates signalment, history, physical examination findings, and knowledge of disease prevalence in the relevant population. The [Merck Veterinary Manual](https://www.merckvetmanual.com/) provides authoritative background on disease prevalence and clinical presentation that informs these estimates. For example, a clinician evaluating a middle-aged dog with polyuria, polydipsia, and a distended abdomen might estimate the pretest probability of hyperadrenocorticism at 40 to 60 percent based on the strength of these clinical signs. A clinician evaluating a young cat with fever and lethargy might estimate the pretest probability of FIP at 20 to 40 percent depending on risk factors such as age, origin, and exposure history. The pretest probability estimate is inherently subjective, but it should be informed by available data. The [World Organisation for Animal Health](https://www.woah.org/en/what-we-do/animal-health-and-welfare) provides official animal-health and surveillance context that can inform prevalence estimates for specific diseases in specific regions. The [American Veterinary Medical Association](https://www.avma.org/resources-tools/pet-owners) provides general pet-owner education context that may be relevant when discussing diagnostic approaches with clients. ### Step 2: Select the Appropriate Test Test selection should be guided by the clinical question and the test's accuracy profile. The clinician should consider the sensitivity and specificity of available tests, the likelihood ratios associated with different result categories, and the practical considerations of sample collection, cost, and turnaround time. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) emphasizes that to select, interpret, and assess the fitness-for-purpose of diagnostic tests, we need to compare the likelihoods of test results being true versus false across both infected and non-infected individuals. The review illustrates the applications and benefits of LR using three assays certified by the World Organisation for Animal Health as serological tests for bovine tuberculosis. When multiple tests are available for the same disease, the clinician should compare their likelihood ratios and consider how the tests complement each other. A highly sensitive test with low specificity may be useful for screening, while a highly specific test with lower sensitivity may be useful for confirmation. The likelihood ratios for each test should be interpreted in the context of the clinical question being addressed. ### Step 3: Perform the Test and Determine the Result Category The test result should be classified into the appropriate category for likelihood ratio application. Some tests have binary results, where the likelihood ratio for a positive or negative result is applied. Other tests have multiple result categories, such as the LDDST response patterns in the hyperadrenocorticism study, where each pattern has its own likelihood ratio. The clinician should ensure that the test was performed correctly and that the sample was handled appropriately. Pre-analytical errors can invalidate test results and make likelihood ratio application meaningless. The [Cornell University College of Veterinary Medicine](https://www.vet.cornell.edu/) provides university veterinary education and animal-health context that may inform understanding of proper sample collection and handling. ### Step 4: Apply the Likelihood Ratio Using the Fagan Nomogram The likelihood ratio is applied to the pretest probability using the Fagan nomogram or direct calculation. The post-test probability represents the updated probability of disease given the test result. This updated probability becomes the pretest probability for any subsequent diagnostic test. The sequential application of likelihood ratios allows the clinician to integrate multiple test results into a coherent probability estimate. Each test result updates the probability, and the process continues until the probability is high enough to justify treatment or low enough to rule out disease, or until the costs and risks of additional testing outweigh the expected diagnostic benefit. ### Step 5: Make a Clinical Decision The post-test probability informs the clinical decision, but it does not determine the decision automatically. The clinician must consider the consequences of false positive and false negative results, the availability and risks of treatment, and the client's preferences and resources. For example, a post-test probability of 90 percent for hyperadrenocorticism may justify treatment in a dog with moderate clinical signs, while the same probability may warrant additional testing in a dog with mild signs or significant comorbidities. The [American Animal Hospital Association](https://www.aaha.org/resources) provides companion-animal preventive care, life-stage, nutrition, and practice guidance context that may inform these decisions. ### Step 6: Document the Reasoning The clinical reasoning process should be documented in the medical record, including the pretest probability estimate, the test results, the likelihood ratios applied, and the post-test probability. This documentation supports continuity of care, facilitates communication with clients and other veterinarians, and provides a basis for reviewing diagnostic decisions. The [World Small Animal Veterinary Association](https://wsava.org/global-guidelines) provides global companion-animal nutrition, welfare, vaccination, and clinical-guideline context that may inform documentation standards and clinical decision-making frameworks. ## Common Failure Patterns in Likelihood Ratio Application ### Misinterpreting Likelihood Ratios as Probabilities A common error is treating likelihood ratios as if they were probabilities or probability changes. A likelihood ratio of 8 does not mean that the probability of disease increases by 8 percentage points. The actual probability change depends on the pretest probability. At a pretest probability of 10 percent, an LR+ of 8 increases the post-test probability to approximately 47 percent. At a pretest probability of 50 percent, the same LR+ increases the post-test probability to approximately 89 percent. At a pretest probability of 90 percent, the same LR+ increases the post-test probability to approximately 99 percent. The Fagan nomogram makes this relationship visually apparent. The same likelihood ratio produces different probability changes at different pretest probability levels. Clinicians who ignore this relationship may overestimate or underestimate the diagnostic value of a test result. ### Applying Population-Specific Likelihood Ratios Inappropriately Likelihood ratios are population independent in the sense that they do not depend on disease prevalence, but they may still vary across populations with different disease spectra or different distributions of test results. A likelihood ratio calculated in a referral hospital population may not apply to a general practice population if the severity or stage of disease differs between these populations. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) notes that LR is a relative measure that ignores the absolute accuracy of tests, and two tests with different accuracy profiles may have the same LR. This limitation means that the clinician should consider sensitivity and specificity alongside likelihood ratios when selecting between tests. ### Ignoring Confidence Intervals Likelihood ratios are point estimates with associated uncertainty. The confidence intervals around likelihood ratios can be wide, particularly in studies with small sample sizes. The hyperadrenocorticism study reported an LR+ of 8.09 for the partial suppression pattern with a 95 percent confidence interval of 2 to 32.72. This wide confidence interval means that the true LR+ could be as low as 2, which would provide only weak evidence, or as high as 32.72, which would provide strong evidence. Clinicians should consider the confidence intervals when interpreting likelihood ratios, particularly when the point estimate is near a clinical decision threshold. A likelihood ratio with a wide confidence interval that spans a decision threshold provides less certain guidance than a likelihood ratio with a narrow confidence interval. ### Failing to Update Pretest Probability Sequentially The sequential application of likelihood ratios requires that the post-test probability from one test becomes the pretest probability for the next test. Clinicians who fail to update probabilities sequentially may underestimate or overestimate the cumulative diagnostic information from multiple tests. The FIP diagnostic process illustrates the importance of sequential updating. Each test result, whether positive or negative, shifts the probability of FIP. The clinician who accumulates test results without updating probabilities may reach a diagnostic conclusion that is inconsistent with the actual cumulative evidence. ### Confusing Likelihood Ratios with Predictive Values Predictive values and likelihood ratios answer different questions. Predictive values answer the question: given a positive or negative test result, what is the probability of disease? Likelihood ratios answer the question: how much does a positive or negative test result change the probability of disease? Predictive values depend on disease prevalence, while likelihood ratios do not. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) emphasizes that positive and negative predictive values combine the likelihoods of true and false results, but they also heavily depend on the prevalence in the tested populations and, therefore, cannot be generalized. Likelihood ratios, by contrast, are population independent and can be applied across different clinical settings. ## Limitations and Complementary Measures ### The Relative Nature of Likelihood Ratios The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) identifies a key limitation of likelihood ratios: as a relative measure, LR ignores the absolute accuracy of tests, and two tests with different accuracy profiles may have the same LR. This limitation can be mitigated by using complementary measures of accuracy, including diagnostic sensitivity and specificity, or ancillary selection criteria. Consider two tests for the same disease. Test A has 99 percent sensitivity and 90 percent specificity, giving an LR+ of 9.9. Test B has 50 percent sensitivity and 95 percent specificity, giving an LR+ of 10.0. Despite similar positive likelihood ratios, these tests behave very differently. Test A will detect nearly all diseased animals but will produce false positives in 10 percent of healthy animals. Test B will miss half of diseased animals but will produce false positives in only 5 percent of healthy animals. The choice between these tests depends on the clinical consequences of false negatives versus false positives. ### The Importance of Test Accuracy Context The [World Organisation for Animal Health](https://www.woah.org/en/what-we-do/animal-health-and-welfare) provides official animal-health and welfare context that emphasizes the importance of test accuracy in disease surveillance and control programs. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) illustrates the applications and benefits of LR using three assays certified by the World Organisation for Animal Health as serological tests for bovine tuberculosis. These assays have different accuracy profiles, and the choice between them depends on the surveillance objective. For disease eradication programs, a highly sensitive test may be preferred to minimize the risk of missing infected animals. For confirmatory testing, a highly specific test may be preferred to minimize the risk of false positives. The likelihood ratio provides a summary measure that balances these considerations, but the clinician must also consider the absolute accuracy of the test and the consequences of misclassification. ### The Challenge of Estimating Pretest Probability The application of likelihood ratios requires an estimate of pretest probability, which is inherently subjective. Different clinicians may estimate different pretest probabilities for the same patient, leading to different post-test probabilities and potentially different clinical decisions. The [Merck Veterinary Manual](https://www.merckvetmanual.com/) provides authoritative veterinary disease, husbandry, prevention, and diagnostic background that can inform pretest probability estimates. The [American Veterinary Medical Association](https://www.avma.org/resources-tools/pet-owners) provides general pet-owner education context that may be relevant when discussing diagnostic uncertainty with clients. ### The Assumption of Conditional Independence The sequential application of likelihood ratios assumes that test results are conditionally independent given disease status. This assumption may not hold when multiple tests measure related biological pathways or when the same underlying pathology affects multiple test results. For example, in FIP diagnosis, a low albumin-to-globulin ratio and elevated total protein may both reflect the same underlying inflammatory process. Applying likelihood ratios for both findings sequentially may overestimate the cumulative diagnostic information because the findings are not independent. The [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) emphasize the importance of understanding each diagnostic test's characteristics and building the index of suspicion brick by brick, which implicitly acknowledges the limitations of simple sequential probability updating. ## Welfare and Safety Context ### The Role of Diagnostic Accuracy in Animal Welfare Diagnostic accuracy has direct implications for animal welfare. False positive diagnoses can lead to unnecessary treatment, with associated stress, side effects, and costs. False negative diagnoses can delay appropriate treatment, allowing disease to progress and potentially causing unnecessary suffering. The [World Organisation for Animal Health](https://www.woah.org/en/what-we-do/animal-health-and-welfare) emphasizes the importance of animal health and welfare in its official guidance. The [American Veterinary Medical Association](https://www.avma.org/resources-tools/pet-owners) provides pet-owner education resources that emphasize the importance of preventive care and appropriate veterinary engagement. The [American Animal Hospital Association](https://www.aaha.org/resources) provides companion-animal preventive care and practice guidance that supports evidence-based diagnostic decision-making. ### The FIP Treatment Context The [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) note that FIP is fatal when untreated, making the ability to obtain a correct diagnosis critical. The guidelines also note that research has demonstrated efficacy of new antivirals in FIP treatment, but these products are not legally available in many countries at this time. This context highlights the importance of diagnostic accuracy in a disease where treatment decisions have life-or-death consequences. The guidelines emphasize that the clinician must consider the individual patient's history, signalment, and comprehensive physical examination findings when selecting diagnostic tests and sample types. This individualized approach recognizes that likelihood ratios provide quantitative guidance but cannot replace clinical judgment. ### The Cardiac Risk Stratification Context The [MINE score study](https://pubmed.ncbi.nlm.nih.gov/40865020) in dogs with preclinical MMVD demonstrates how risk stratification can inform monitoring and intervention decisions. Dogs classified as severe had a median time to cardiac event of 718 days, compared to 1216 days for moderate cases and 2604 days for mild cases. This information can guide the frequency of recheck examinations and the timing of intervention. The [World Small Animal Veterinary Association](https://wsava.org/global-guidelines) provides global companion-animal clinical-guideline context that may inform cardiac disease management recommendations. The [Cornell University College of Veterinary Medicine](https://www.vet.cornell.edu/) provides university veterinary education and animal-health context that may inform understanding of cardiac disease diagnosis and management. ## Professional Escalation Criteria ### When to Refer for Advanced Diagnostic Testing The clinician should consider referral for advanced diagnostic testing when the post-test probability remains uncertain after initial testing, when the clinical picture is atypical, or when the consequences of misdiagnosis are severe. The [Merck Veterinary Manual](https://www.merckvetmanual.com/) provides authoritative background on disease presentation and diagnostic approaches that can inform referral decisions. For FIP, the [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) emphasize that the approach to diagnosing FIP must be tailored to the specific presentation of the individual cat. When the diagnosis remains uncertain after initial testing, referral to a specialist or advanced diagnostic center may be appropriate. The guidelines note that FIP can be challenging to diagnose owing to the lack of pathognomonic clinical signs or laboratory changes, especially when no effusion is present. ### When to Seek Specialist Consultation Specialist consultation should be considered when the clinician lacks experience with a particular disease, when the diagnostic workup requires specialized equipment or expertise, or when the clinical situation is complex or unusual. The [Cornell University College of Veterinary Medicine](https://www.vet.cornell.edu/) provides university veterinary education and animal-health context that may inform understanding of when specialist consultation is appropriate. For cardiac disease, the [MINE score study](https://pubmed.ncbi.nlm.nih.gov/40865020) demonstrates the value of echocardiographic assessment in risk stratification. Clinicians without access to echocardiography or without experience in interpreting echocardiographic findings should consider referral to a veterinary cardiologist for dogs with suspected MMVD. ### When to Consider Euthanasia In diseases with grave prognoses and limited treatment options, the clinician may need to discuss euthanasia with the client. The FIP context illustrates this challenge, as the disease is fatal when untreated, and effective antiviral treatment is not legally available in many countries. The [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) emphasize the importance of obtaining a correct diagnosis, which may involve referral for advanced diagnostic testing. The decision to recommend euthanasia should be based on the best available diagnostic information, the prognosis, the animal's quality of life, and the client's resources and preferences. The [American Veterinary Medical Association](https://www.avma.org/resources-tools/pet-owners) provides pet-owner education resources that may inform discussions about end-of-life care. ## Records and Measurements ### Documenting Pretest Probability Estimates The medical record should document the pretest probability estimate and the reasoning behind it. This documentation should include the signalment, history, physical examination findings, and any population prevalence data that informed the estimate. The record should also note any uncertainty in the estimate. The [World Organisation for Animal Health](https://www.woah.org/en/what-we-do/animal-health-and-welfare) provides official animal-health and surveillance context that may inform prevalence estimates for specific diseases. The [American Veterinary Medical Association](https://www.avma.org/resources-tools/pet-owners) provides general pet-owner education context that may inform client communication about diagnostic uncertainty. ### Documenting Test Results and Likelihood Ratio Application The medical record should document the test performed, the result obtained, the likelihood ratio applied, and the post-test probability calculated. This documentation supports continuity of care and provides a basis for reviewing diagnostic decisions. The record should also note any limitations in the likelihood ratio application, such as wide confidence intervals or uncertainty about whether the likelihood ratio applies to the specific patient population. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) emphasizes the importance of considering complementary measures of accuracy alongside likelihood ratios. ### Tracking Diagnostic Outcomes The clinician should track diagnostic outcomes to evaluate the accuracy of likelihood ratio application in their practice. This tracking can identify situations where likelihood ratios derived from published studies do not perform as expected in the local population. The [World Small Animal Veterinary Association](https://wsava.org/global-guidelines) provides global clinical-guideline context that may inform quality improvement initiatives. The [American Animal Hospital Association](https://www.aaha.org/resources) provides practice guidance that may inform documentation and quality improvement standards. ## Frequently Asked Questions ### How do likelihood ratios differ from sensitivity and specificity? Sensitivity and specificity describe test performance in diseased and non-diseased populations separately. Likelihood ratios compare the likelihood of a test result being true versus false across both populations, providing a single measure that directly informs probability updating for individual patients. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) explains that diagnostic sensitivity and specificity report the accuracy of classification in infected and non-infected individuals separately and do not compare these likelihoods directly. ### Why are likelihood ratios considered population independent? Likelihood ratios do not depend on disease prevalence in the tested population, unlike predictive values. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) notes that positive and negative predictive values heavily depend on the prevalence in the tested populations and cannot be generalized, while LR balances the likelihoods of true versus false results and is population independent. This property allows likelihood ratios from published studies to be applied across different clinical settings. ### How do I calculate a positive likelihood ratio? The positive likelihood ratio is calculated as sensitivity divided by (1 minus specificity). This calculation answers the question of how much more likely a positive test result is in a diseased animal compared to a healthy animal. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) provides the conceptual framework for this calculation, emphasizing the comparison of likelihoods of true versus false results across infected and non-infected individuals. ### What does an infinite likelihood ratio mean? An infinite positive likelihood ratio occurs when specificity is 100 percent, meaning that false positives do not occur. The [2021 Veterinary Clinical Pathology study](https://pubmed.ncbi.nlm.nih.gov/33728722) reported an infinite positive likelihood ratio for the lack of suppression pattern on the LDDST to diagnose hyperadrenocorticism. This finding means that the lack of suppression pattern strongly supports a diagnosis of HAC, essentially confirming the diagnosis when present. ### How do I use a Fagan nomogram? The Fagan nomogram has three vertical scales: pretest probability on the left, likelihood ratio in the center, and post-test probability on the right. Place a straight edge from the pretest probability through the likelihood ratio value, and read the post-test probability where the line intersects the right scale. This graphical method avoids arithmetic errors in converting between probability and odds. ### Can likelihood ratios be used for prognostic assessment? Yes, likelihood ratios can be applied to prognostic questions by defining the outcome of interest and calculating the likelihood of that outcome given specific findings. The [2025 Journal of Veterinary Internal Medicine study](https://pubmed.ncbi.nlm.nih.gov/40865020) used the MINE score to stratify cardiac risk in dogs with preclinical MMVD, demonstrating how probability-based approaches can inform prognostic assessment and management decisions. ### What are the limitations of likelihood ratios? Likelihood ratios are relative measures that ignore the absolute accuracy of tests, and two tests with different accuracy profiles may have the same LR. The [2021 Revue Scientifique et Technique review](https://pubmed.ncbi.nlm.nih.gov/34140723) recommends using complementary measures of accuracy, including diagnostic sensitivity and specificity, or ancillary selection criteria. Likelihood ratios also require an estimate of pretest probability, which is inherently subjective. ### How do likelihood ratios apply to FIP diagnosis? The [2022 AAFP/EveryCat guidelines](https://pubmed.ncbi.nlm.nih.gov/36002137) emphasize that understanding each diagnostic test's sensitivity, specificity, predictive value, likelihood ratio, and diagnostic accuracy is important when building a case for FIP. The guidelines describe building the index of suspicion brick by brick, where each test result updates the probability of FIP. Given that the disease is fatal when untreated, the ability to obtain a correct diagnosis is critical. ## Related Veterinary Guides - [Likelihood Ratios in Veterinary Diagnostic Testing](/knowledge/veterinary-medicine/veterinary-epidemiology/likelihood-ratios-veterinary-diagnostic-testing) - [Diagnostic Test Evaluation: Sensitivity and Specificity in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-epidemiology/diagnostic-test-evaluation-sensitivity-specificity-veterinary-medicine) - [Diagnostic Test Accuracy Studies in Veterinary Medicine: Design and Reporting](/knowledge/veterinary-medicine/veterinary-research-methods/diagnostic-test-accuracy-studies-veterinary-design-reporting) - [Interpreting Diagnostic Test Accuracy: ROC Curves in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-epidemiology/interpreting-diagnostic-test-accuracy-roc-curves-veterinary-medicine) - [Meta-Analysis of Veterinary Diagnostic Test Accuracy](/knowledge/veterinary-medicine/veterinary-research-methods/meta-analysis-veterinary-diagnostic-test-accuracy) ## References and Further Reading - [Pet Care](https://www.avma.org/resources-tools/pet-owners). American Veterinary Medical Association. - [AAHA Guidelines](https://www.aaha.org/resources). American Animal Hospital Association. - [Global Guidelines](https://wsava.org/global-guidelines). World Small Animal Veterinary Association. - [Merck Veterinary Manual](https://www.merckvetmanual.com/). Merck Veterinary Manual. - [Cornell University College of Veterinary Medicine](https://www.vet.cornell.edu/). Cornell University. - [Animal Health and Welfare](https://www.woah.org/en/what-we-do/animal-health-and-welfare). World Organisation for Animal Health. - [2022 AAFP/EveryCat Feline Infectious Peritonitis Diagnosis Guidelines.](https://pubmed.ncbi.nlm.nih.gov/36002137). Journal of feline medicine and surgery, 2022. - [Diagnostic likelihood ratio - the next-generation of diagnostic test accuracy measurement.](https://pubmed.ncbi.nlm.nih.gov/34140723). Revue scientifique et technique (International Office of Epizootics), 2021. - [Patterns of the low-dose dexamethasone suppression test in canine hyperadrenocorticism revisited.](https://pubmed.ncbi.nlm.nih.gov/33728722). Veterinary clinical pathology, 2021. - [Risk Stratification Using Mitral INsufficiency Echocardiographic Score 2 in Dogs With Preclinical Mitral Valve Disease.](https://pubmed.ncbi.nlm.nih.gov/40865020). Journal of veterinary internal medicine, 2025. - [Pulmonary cystic echinococcosis.](https://pubmed.ncbi.nlm.nih.gov/20216420). Current opinion in pulmonary medicine, 2010. > This article is educational and is not a substitute for veterinary diagnosis or treatment. Contact a veterinarian for advice about an individual animal.