Page 105 - Withrow and MacEwen's Small Animal Clinical Oncology, 6th Edition

84 PART I The Biology and Pathogenesis of Cancer

proportional statistics from a necropsy database and concludes that would die before 10 years of age, whereas the values for golden
golden retrievers have an increased “risk” of tumors similar to that retrievers are somewhat different (30%, 50%). Of course, there may
22
be true differences between the two study populations, and/or the
for Boxers. However, only the proportion of dead dogs that had
VetBooks.ir cancer are available in that study, and these data cannot be used to differences may be influenced by referral bias and the high propor-
tion of unclassifiable deaths in the VMDB study.
estimate risk. Although proportional measures, such as those pre-
sented in an article by Fleming et al, have some usefulness for To further illustrate this example, using just the breeds in Fig.
23
describing patterns within a breed, they are very risky to use for 4.1 and data from the Swedish insurance database, if one ranked
comparison across breeds in which population-based measures are the breeds based on actual numbers of dogs that died because of
unavailable and the degree of referral bias is unknown. In addition, tumors (e.g., perhaps how an oncology clinician would perceive
in those data, 40% of deaths could not be classified pathologically the “risk” based on dogs that present to a specialty clinic), golden
and the unclassified proportion showed extreme variation across retrievers would be number one because they are among the more
breed (e.g., 16%–60%). Fig. 4.1 shows a comparison between pro- numerous breeds in this population. Likewise, if one ranked the
portional mortality ratios and true mortality rate using a subset of breeds by the proportion of dead dogs that had tumors (e.g., simi-
data from Bonnett et al, a study with information on the population lar to what would be reported in analysis of postmortem data),
at risk and data from the Veterinary Medical Database (VMDB) the top three would be BMDs, Boxers, and golden retrievers. So,
study. 23,24 For golden retrievers, 30% of deaths (before 10 years of in these examples, as has been frequently reported in the United
age in the Swedish insurance population) were a result of cancer. States, based on proportional statistics, golden retrievers would be
24
For Leonbergers and Boxers, the proportional mortality was 28% labeled as one of the highest risk breeds. However, in looking at
and 37%, respectively. Proportional values for these three breeds the true incidence based on these Swedish data, they do not have
24
may be similar, but, in fact, Leonbergers and Boxers have a risk for an increased risk (before 10 years of age) compared with all breeds.
death resulting from cancer (before 10 years of age) that is almost There is likely considerable misunderstanding of the occurrence
four times as high as that for golden retrievers (approximately 200 of cancer in dogs in the United States because of the lack of
24
deaths per 10,000 years-at-risk versus 55 [p > 0.05]). Irish wolf- accurate incidence data and confusion about the interpretation
hounds and Bernese mountain dogs (BMDs) have an equal risk of proportional statistics. Of course, where a breed is very com-
(approximately 300 deaths resulting from tumors per 10,000 dog- mon, such as the golden retriever, and given that a considerable
years at risk [DYAR]), but tumors account for more than 40% of proportion of them die of cancer, that will represent an important
deaths in BMDs and only 22% in Irish wolfhounds. Note that population burden of disease, even if they are not truly the “high-
24
these are deaths before 10 years of age. Comparing the proportional est risk” breed. This is why golden retrievers were selected for the
mortality values between the two studies, values for BMDs are very largest prospective cohort study conducted in a dog population,
similar (42%, 45%), perhaps because almost all dogs of this breed the Golden Retriever Lifetime Study (GRLS), to investigate the

Yearly cancer mortality rate (deaths due to cancer per 10,000 dog-years-at-risk in Swedish dogs ≤ 10 years of age).*
Proportional cancer mortality (%) (deaths due to cancer ÷ all deaths in Swedish dogs ≤ 10 years of age).*
Relative Frequency (%) (deaths due to cancer ÷ eligible deaths in dogs from VMDB). † 306
296

197 203
168

119
105

45 34 50 55 50 37 44 32 41 45
21 18 27 30 14 20 13 23 23 26 28 na 22
Irish Wolfhound 6,7
Labrador Retriever 1 All Breeds 1 Newfoundland 3 Great Dane 3,4 Doberman 4,5 Leonberger 5,6 Boxer 5,6 Bernese Mountain Dog 7
Golden Retriever 1

• Fig. 4.1 Comparison of true mortality rate (blue column) and proportional mortality (red column) for
selected dog breeds. True mortality is reported as the total number of deaths resulting from cancer per
10,000 dog-years-at-risk (e.g., the cancer mortality in Labrador retrievers was 45 deaths per 10,000
dogs per year). The proportional mortality is reported as the percentage of deaths resulting from cancer
compared with all deaths reported in that breed (e.g., 21% of deaths in Labrador retrievers were caused
by cancer). The 95% confidence intervals for mortality rates overlap for breeds with the same number
(e.g., mortality risk in Labrador retrievers and golden retrievers was not different from that for all breeds
combined, but was different from the other breeds listed in the table). 24,25