Clarifications needed on design of extrahepatic portosystemic shunts study

The paper “Dogs ≥ five years of age at the time of congenital extrahepatic portosystemic shunt diagnosis have better long-term outcomes with surgical attenuation than with medical management alone”¹ by 28 authors from 16 colleges of veterinary medicine purports to provide definitive evidence that surgical attenuation of portosystemic shunts in dogs diagnosed at or after 5 years of age extends life and lessens severity of illness. While providing a contribution to management of these patients, there appear to be design flaws that need clarification. In the Discussion section, the authors acknowledge limitations including the retrospective nature of the study, but these limitations may be understated. Our comments that follow presume that the predominant cause of death in either group was euthanasia. If not, the data on causes of death should have been provided.

The assignment of dogs to the medical group versus the surgical group appears to have been determined by the dog owners. Presumably, their decisions included counsel from their veterinarians. There could have been very different behavior characteristics between the 2 owner groups concerning the decision to euthanize. For example, the medical group could have been more likely to elect euthanasia, thereby leading to earlier euthanasia. To wit, if the owners who did elect surgical treatment were more pessimistic than those who opted for surgery, were more financially distressed, or received more pessimistic veterinary advice, they could understandably choose euthanasia sooner than those in the surgery group. This could skew the statistics.

The stage of illness at diagnosis may have varied between groups. The reasons for referral are not given or discussed. Were they referred for suspected shunts? Or were they referred for other reasons and then determined to have shunts? If shunts were suspected prior to referral, for how long? This information could have had an impact on actual “time zero” and by extension, survival time. “Time zero” was defined as time of diagnosis at the centers, which may not correlate with the time course of the disease. Dogs with more chronic illness might have been euthanized earlier than others, and these same dogs could have been more likely to land in the medically managed group. If an owner presented to a center as a last resort and did not deem surgical treatment a solution, they may have been more likely to elect euthanasia earlier than owners of dogs with less-severe illness.

The survival statistics are confusing. The cumulative probability of survival from time of diagnosis in the surgery group was 10.9 years compared to 3.4 years for the medically managed group. Age at time of diagnosis was 7 ± 1 years in the surgery group and 8 ± 2 years in the medically managed group. This would seem to indicate that the median life span of dogs in the surgery group was 17.9 (16.9 to 18.9) years versus 11.4 (9.4 to 13.4) years in the medical group. It seems highly unlikely that the surgery group consisting of 232 dogs could have lived so long beyond life expectancy.²

The study definition of medical management was that they were administered at least 1 medical treatment (lactulose, antimicrobial, antiseizure medication, or hepatic support diet). The term “required” was used to describe any one of these therapies that were given. Administration of medical therapy does not necessarily indicate its requirement and therefore may not reflect clinical status. Therefore, less medical treatment in the surgery/medicine group than in the medical group does not constitute firm evidence that the clinical illness was less severe in the surgery/medicine group.

William Rogers, DVM, DipACVIM

Cheryl L. Harris, DVM, DipACVIM (Oncology)

The authors respond:

We would like to thank Drs. Rogers and Harris for their interest in this study.

Regarding cause of death for dogs in the study, 50% of dogs were euthanized (44.8% of the surgical attenuation group and 54.9% of the medical management group), 16% died naturally, and the remainder did not have a cause of death listed in the medical records. Of those that were euthanized, 48.9% and 40% were euthanized for reasons related to their extrahepatic portosystemic shunt (EHPSS) in the medical management and surgical attenuation groups, respectively. Only data regarding whether or not death was related to EHPSS, which we felt was most relevant, were provided in the manuscript, as a higher percentage of dogs in the medical management group had a cause of death or reason for euthanasia that was related to signs attributable to their EHPSS diagnosis. We hope these additional data are helpful.

We agree that having the treatment determined by clients does add an acknowledged confounding variable, but this is unavoidable in a clinical study such as this. As we discussed, it is impossible in a retrospective study to know what factors directed an owner to choose surgical treatment over medical management and whether those choices may have also led to a difference in timing of end-of-life decisions. Finances, opinions of the primary care or other veterinarian, or influence of popular press certainly play a role in client decision-making in many veterinary decisions, and this limitation is inherent to all retrospective studies.

The main statistically significant differences between our groups were that dogs in the medical management group were slightly older (8 years vs 7 years) than dogs in the surgical attenuation group, dogs in the medical management group had a higher BUN (8 mg/dL vs 7 mg/dL) than dogs in the surgical attenuation group, and dogs in the surgical attenuation group were more likely to have clinical signs attributed to EHPSS than dogs in the medical management group (94.4% vs 84.9%). These differences between groups were either very small or clinically insignificant, indicating that the groups were balanced, despite owner selection of the treatment group. Further, when considering more optimal study designs in light of the results of this study and others,¹ there are ethical concerns to prospectively limiting optimal care to dogs with EHPSS, suggesting this study is impossible in any prospective, randomized design.

Data on reason for referral or duration of clinical signs were not collected. We agree that the reported duration of clinical signs could be an interesting variable to evaluate; however, our clinical impression is that duration of signs does not correlate with severity of clinical signs at presentation. There are also no data to support that duration of clinical signs impacts the client’s decision to pursue surgical attenuation over medical management. We disagree that the start of first-noticed clinical signs or first clinical suspicion of a shunt would be a more accurate “time zero” for survival time. Using that logic, as this is a congenital disease, it would be most accurate to call “time zero” as the date of birth. Date of definitive EHPSS diagnosis via imaging or surgical visualization allowed for a consistent time point from which to evaluate all cases.

The statistical evaluation of survival time was performed by a biostatistician with a wealth of experience in evaluation and interpretation of clinical data. While we mentioned dogs being censored from analysis in the manuscript, which dogs were censored from analysis was not described in detail, which may be adding to your confusion. Dogs were censored from survival analysis for standard reasons including if they were still alive at the end of the study, were lost to follow-up, or died of causes unrelated to EHPSS. Censoring deaths unrelated to EHPSS in older dogs caused survival times that were higher than life span. The median (interquartile [25th to 75th percentile] range) age at death of dogs that died from unrelated causes (n = 55) was 13 (11 to 14) years. If one calculated uncensored overall survival time for comparison, the median (95% CI) age of dogs in the surgical group was 6.7 (5.8 to 7.6) years compared to 2.3 (1.5 to 3.4) years in the medical group. This difference is statistically significant (P < .001). Data lost to follow-up, common in any study over long periods of time, can also cause an overestimation of survival times; however, this limitation would apply equally to both groups.

Clinical signs attributed to having an EHPSS, including signs of hepatic encephalopathy, can be ameliorated with a single form of medical management, especially in a chronic situation. Success of medical management is dependent on the severity of the clinical signs and each individual dog’s response to therapy. While multimodal therapy is frequently required acutely, our goal is to ultimately decrease to the lowest effective medical therapy regimen, as has been previously reported.¹ In our practices, dogs with EHPSS are often chronically managed with only lactulose or a hepatic support diet. Additionally, a recent study² in dogs with portosystemic shunts found that hepatic support diet alone resulted in a similar survival time to hepatic support diet and lactulose given concurrently, with dogs overall having an improved quality of life after beginning medical management.

The authors agree that the word “required” here in reference to continued medical therapies may have been better stated as “administered.” While we would like to believe that the need for medical therapies is reassessed and dosages lowered as needed in patients with EHPSS as stated above, unfortunately, this may not always be the case. Again, this was only 1 marker used in addition to the data presented on continued presence of clinical signs in each group to show quality of life of the animals in each group.

Our goal for publication of this information was to provide as much information as possible to guide decision-making for older dogs diagnosed with EHPSS to allow these patients to have the best long-term quality of life possible. Every dog is different, and as with any clinical decision-making, the veterinarian must use available evidence and their best judgment regarding whether surgical or medical management will be best for each individual animal.

Mandy L. Wallace, DVM, MS, DACVS-SA

University of Georgia

Janet A. Grimes, DVM, MS, DACVS-SA

University of Georgia

Cassie N. Lux, DVM, DACVS-SA

University of Tennessee

Vanna M. Dickerson, DVM, MS, DACVS-SA

Texas A&M University

Kenneth A. Carroll, BVSC

North Carolina State University

Valery F. Scharf, DVM, MS, DACVS-SA

North Carolina State University

Raymond K. Kudej, DVM, PhD, DACVS

Tufts University

Aki Otomo, DVM

Louisiana State University

Ameet Singh, DVM, DVSc, DACVS

University of Guelph

Annellie Miller, DVM, MPH

Colorado State University

Penny J. Regier, DVM, MS, DACVS-SA

University of Florida

Chiara Curcillo, VMD

University of Pennsylvania

David E. Holt, BVSc, DACVS

University of Pennsylvania

Jessica A. Ogden, DVM, DACVS-SA

Purdue University

Shiori Arai, DVM, PhD, DACVS-SA

University of Minnesota

David A. Upchurch, DVM, MS, DACVS-SA

Kansas State University

James Howard, DVM, MS, DACVS-SA

The Ohio State University

Robert J. Hardie, DVM, DACVS, DECVS

University of Wisconsin

Milan Milovancev, DVM, DACVS-SA

Summit Veterinary Referral Center

Brad M. Matz, DVM, MS, DACVS-SA

Auburn University

Chad W. Schmiedt, DVM, DACVS

University of Georgia