Bishop et al assert that we made errors in diagnostic criteria, construction of sample populations, and genotype data analyses that prohibited us from confirming the results of their study. We appreciate the opportunity to clarify the misunderstanding of genetic study design and the misinterpretation of our own study that are evident in their letter.
A genetic association study refers to the association between a genetic variant and a phenotype of interest. In our study, we repeatedly stress that our phenotype of interest was clinical pancreatitis and not subclinical pancreatitis. Several of the Bishop et al critiques relate to the inability of our study to test for an association with subclinical disease. As discussed in our article, we cannot rule out a relationship between the variant and subclinical pancreatic changes. We recognize that this could be a scientifically interesting relationship, but it is irrelevant to the goal of determining if the SPINK1 genotype can predict dogs that will develop clinical pancreatitis.
Bishop et al mistakenly state that control groups in genetic studies require the same diagnostic tests as the disease population. In fact, diagnostic test requirements can be and often are completely different for the case and control groups in genetic association studies, and it can even be inappropriate to use the same tests. For example, a cPL concentration can assist in the diagnosis of a pancreatitis case, but it has minimal benefit for phenotyping controls. A normal concentration provides a picture at a single moment in time, and its predictive value is only applicable to that moment. It neither demonstrates that a dog has been free of pancreatitis in the past nor does it determine whether a dog will develop pancreatitis in the future.
Bishop et al also criticize the diagnoses of our pancreatitis case group. First, they question the diagnoses of the 5 dogs with positive SNAP cPL tests but no SPEC cPL concentration. Although we agree that a positive SNAP cPL test alone does not provide a definitive diagnosis, this test result was only one of the criteria that led to the phenotypic assignment for this subset of cases. As described in the materials and methods, these dogs had compatible clinical signs for which other causes had been excluded as determined by a review of abdominal radiographs, serum biochemistry results, and at least 1 year of follow-up. Furthermore, a recent study estimated the specificity of the SNAP cPL to be 71-78%. Thus, most of the dogs diagnosed in this fashion truly are affected by pancreatitis. Secondly, Bishop et al question the cases where ultrasound was used in the diagnosis of pancreatitis by quoting a study where a small subset of dogs with ultrasonographic evidence of pancreatitis had normal or hyperplastic pancreata on laparoscopic biopsy. However, given that pancreatitis can be a focal disease, the absence of identification in laparoscopy biopsy samples is not sufficient to rule out its presence or comment on the specificity of ultrasound for the diagnosis of pancreatitis.
The major limitation of the previous study by Bishop et al was a small and flawed control group. Briefly, the 25 control Miniature Schnauzers in that study were too young, had an inappropriately short clinical history (3 months free of clinical signs), and, as discussed above, the normal cPL concentrations provided little to no phenotyping benefit. In our study, the 60 control dogs were at least 9 years old and had never once developed gastrointestinal signs that required veterinary care. For the purpose of phenotyping clinical pancreatitis in a genetic association study, this is an excellent control group.
Perhaps most importantly, Bishop et al misinterpret the statistical analysis of our genotype data. They state that from Table 1 of our study “it is clear that the diseased group had a higher percentage of homozygous dogs than heterozygous dogs.” In reality, both the diseased and control groups had more homozygous than heterozygous dogs because the c.74C variant allele was the most common allele in both populations. The proportion of dogs with each genotype was not statistically different between the groups and results in the conclusion of no association between the c.74C variant and the clinical pancreatitis phenotype. Few to no dogs in either group were homozygous for the reference allele. This is not, as they suggest, a surprising finding. In reality, it is follows expectations based on the Hardy-Weinberg equilibrium.
We hope that interested readers will come away from this discussion with the knowledge that the prevalence of the SPINK1 variant was high in our population of 60 Miniature Schnauzer dogs that never once had clinically relevant pancreatitis over at least a 9-year time period and was no different from a group of 17 Miniature Schnauzers diagnosed with pancreatitis. In sum, the data from our population of dogs did not demonstrate any relationship between the variant and clinical pancreatitis.