Dog Longevity 

Created by Dr. Kelly M. Cassidy, 2007

Surveys - MHO (My Humble Opinion)

Or a few modifications I'd like to see in health and longevity sampling, analyses, and reports

Last updated 07/04/2007

After reading dozens of dog health and longevity survey reports, it is impossible not to develop opinions about what makes a good sampling design, questionnaire and survey report.  Most of the single-breed surveys are conducted by breed clubs, but they vary radically in quality.

Sampling design

If you are trying to measure an attribute of a population (such as the attribute "longevity" in the population of a certain dog breed), obtaining a representative sample is as important as sample size.  If you have a large sample size, but most of the sample comes from an unusual type of owner, a particular geographic area, a single blood line in a breed with multiple blood lines, etc., the results may not be a good representation of dogs from a different area, blood line, or owner type.

Most breed surveys are done by breed clubs and often, about half of the responses are by breed club members.  Most also had an online survey form.  Several solicited responses from email lists for the breed.  A very small number contacted a kennel club, such as the AKC, for a list of breed registrations and mailed forms to a random subsample.  The last sampling technique probably produces the most representative sample of the common techniques used, but response rates are typically lower than for the more motivated club members and email list members.

Whatever the sampling strategy, if the sample sizes are large enough and different groups are sampled, results for the most important basic data should be analyzed for all groups together and separately.  For example, longevity results should be reported for the entire sample, but also broken down by whether the respondent was or wasn't a breed club member, did or didn't belong to a breed email list, was or wasn't a breeder, etc. 

A potentially useful bit of information for analyses for purebred dog surveys is the registered name and (if there is no public pedigree database for the breed) the pedigree of the dog.  More respondents will be likely to give such information if the group conducting the survey is impartial and not associated with the breed being surveyed and if the respondent can request the information remain confidential.  Longevity and morbidity data for large numbers of dogs with known pedigrees can allow investigation into questions such as: Do dogs with conformation titles have significantly different mortality or morbidity than dogs without? What about dogs with performance titles? What about dogs with ancestors with conformation or performance titles? If longevity of parents is known, is parental longevity correlated with offspring longevity?  If the parents passed standard pre-breeding health screening, are pups likely to live longer or be healthier? Etc. 

Reporting Longevity Data

I was primarily interested in breed longevities obtained by these surveys.  You might think longevity is an unambiguous number easy to quantify, report, and compare among breeds, but reported longevity is not as simple as it appears::

	Mean or median? Some studies reported median longevity; others report mean (=average).  The mean is often considered less appropriate than the median because dog ages at death are usually not normally distributed.  Normally distributed data have a "bell-curve" shape.  For example, if the mean is 10, and the ages at death are normally distributed, then there would be about the same numbers of deaths between 9 and 10 as between 10 and 11.  For most dog longevity data, their is a gradually increasing mortality rate up to a certain age, then a steep increase.  The median is more often used on data that might be skewed to one side or the other.  The median longevity is the age at which about half the dogs in the population will have died and half will still be alive.  So, if the median is 10 yrs, about half the dogs will be dead before the age of 10 and half will live longer than 10 years.
	Measure of dispersion? Few studies reported any measure of dispersion.  A measure of dispersion indicates the "spread" of the data.  The standard deviation is a measure of dispersion of the mean.  For example, breed A might have a mean longevity of 10.5 ± 2.4 yrs, where 2.4 yrs is one standard deviation.  For normally distributed data, about 67% of the data points will be within one standard deviation.  For this example, 67% of dogs would be expected to die between 8.1 and 12.5 yrs. About 95% of dogs will die between two standard deviations of the mean (5.7 to 14.9 yrs).  If median is calculated, it is common to indicate dispersion in the form of quartiles.  The age by which 25% of dogs are dead is the first quartile, 25% to 50% is the second quartile (= the median), 50% to 75% is the third and 75-100% is the fourth.
	Starting age? Some studies report longevity from birth, others from about two or three months (when most people acquire their puppy).  The starting age can make a big difference because the first few days of a puppy's life are a period of high mortality for all breeds.  In some breeds were genetic problems are common or undesirable phenotypes are common, pre-weaning mortality can be even higher. For example, in breeds where deafness is common at birth (e.g., Dalmatians) puppies are tested for deafness before placement and euthanized if they are bilaterally or (sometimes) unilaterally deaf. An example of a breed in which puppies are often born with an undesirable phenotype is the Rhodesian Ridgeback. Puppies born without a "ridge" are commonly euthanized.
	Include deaths due to accidents or temperament euthanasia? Some studies exclude deaths due to accidents or euthanasia for temperament problems (usually aggression) in longevity calculations.  Since these causes of death are, in most breeds, skewed towards younger dogs, their exclusion usually increases reported longevity.  In most studies, deaths due to these causes among the surveyed population (which typically includes the most responsible of dog owners) is less than 5%.  In a few breeds, deaths due to these causes can approach 30% and will have a significant effect on longevity calculations.
	Statistical analyses? In many studies, there are comparisons of longevities between two samples, e.g., between males and females, dogs obtained from different sources, etc.  Few of these comparisons are subjected to statistical analyses.  For example, if a study shows that males live an average of 11.2 ± 3.4 yrs and females live an average of 11.5 ± 3.8 yrs, it is unlikely that the difference of 0.3 yrs is "statistically significant" because the standard deviation is comparatively large.   Statistical significance means that it is unlikely that the difference between two measures is due to chance.
	Leaving important data out of the report!!! This aspect of reports was one of the most frustrating I encountered.  It was surprisingly common for report writers to leave out sample size, i.e., the number of dogs used to calculate longevities.  In some cases, such as the Petit Basset Griffon Vendeen study, I was able to calculate the probable sample size from other numbers provided in the report.  In others cases, there was not enough information and attempts to get the sample size from club members associated with the study were unsuccessful.   Several health surveys did not include any longevity calculations.  In most cases, that was apparently because the sample size of deceased dogs was too low.  In at least one case (the health surveys from 1996 and 2001 from the Rhodesian Ridgeback Club of the United States), there was extensive information on the causes of mortality for several hundred dogs, but no mean or median longevity provided in the report.

For report writers, I would suggest the following:

Sample size.  ALWAYS REPORT SAMPLE SIZE FOR ANY REPORTED PARAMETER!!!  Pardon the shouting, but this is an extremely important piece of information.

Median, mean, measures of dispersion, accidents, euthanasia, etc. Since there are so many different ways to calculate longevity and measures of dispersion, I suggest that the report include multiple calculations.  

Some reports calculate mortality from birth, some from when most owners acquire their puppy.  Some reports are not clear about which method they used.  I suggest treating the two separately.  For most people, the important number  is the expected lifespan of their dog from the time they acquire it at about 2-6 months. 

IMAGINARY LONGEVITY SURVEY REPORT TABLE

Breed longevity, calculated from the time most owners obtain a puppy (about 2-6 months)

Note that it is clear that puppy mortality from birth to the time the puppy goes to a new home is treated separately

Puppy Mortality

There are many good examples (such as the Purdue studies) of reporting on puppy mortality (birth to when the pups go their new home), so I will not provide one.  Puppy mortality reports should  include, at a minimum, the number of puppies per litter, the number stillborn, and the number that died or were euthanized before going to their new homes (or up to 2 months if the breeder keeps them).  It's also nice to see the average age of the mother at first, second, third, etc. litter, the number of times a female is bred,  litter size as a function of age and/or litter number.  

I would like to see more information about sires, such as asking breeders why a particular sire was chosen, his age at breeding, etc.

Measuring and Reporting Morbidity

Good morbidity (nonfatal illness) information is a critical element of longevity surveys.   As vet care improves, longevity should increase. So, if a breed's longevity increases over time, how can you answer the question: Is breed health improving or is vet care improving?  

As an example, bloat and torsion (GDV) is rampant in many breeds. The risk of GDV rises in older dogs, but it often strikes young dogs.  Improved veterinary care is saving dogs that would otherwise have died young, which could cause a substantial increase in longevity of a breed.  Saving a dog that has torsioned can cost several thousand dollars. Most owners would prefer that longevity be improved by breeding healthier dogs instead of through expensive vet care.

So, it is important to determine: What would be a breed's longevity in the absence of non-routine vet care and How healthy is the average breed member at age 1, age 2, age 3, etc.?  

A number of health survey reports include statistics on morbidity. Most break out health issues into categories (skin, respiratory, skeletal, etc.)  and discuss the incidence of each in the survey population.  

Frequently, there will be (paraphrasing from multiple reports) a few summary sentences like this:   

"The average age of the survey population was 4.5 years. Each dog had an average of 2.4 health issues. Owners reported that 75% of dogs were in good health." 

This type of information is virtually worthless.   Were the "2.4 health issues/per dog" minor ear infections or diseases like lymphoma?   The information in the last sentence is equally worthless because dogs of all ages are grouped together.   Most young dogs are healthy. Rates of morbidity will increase with age.  Breeds in which sudden early death is common may have a higher percentage of healthy live dogs than a long-lived breed with nonfatal morbidity in the last years of its life. 

More useful reporting of morbidity

Morbidity statistics need to include some measure of the severity of problems and the relationship between morbidity and age.

Age at death is an easier parameter to quantify than morbidity because a dog is either dead or alive.  If  a dog is sick, how sick is he?  Some ailments are always serious, e.g., bloat, lymphoma, or blindness. But what about diarrhea? Allergies? Loss of hearing in only one ear?  A single bout of diarrhea in a puppy is usually minor.  Chronic, severe diarrhea can kill a dog or shorten its life.  

I suggest a simple classification of the severity of ailments, as determined by the owner:  Here is an example (which, no doubt, could be improved upon):

Mild morbidity - Dog would likely have survived with or without vet care; vet care was less than $200; and condition did not substantially degrade the dog's future quality of life.  Examples: minor ear infection, kennel cough, a bout of diarrhea, a mild heart murmur.

Moderate morbidity - Dog might have survived (or did survive) without vet care. Treatment less than $1000, if treated. Condition adversely affects quality of life, but dog is able to have a good quality of life. Condition does not require life-long medication. Examples: Periodic flare-ups of moderate skin allergies, deaf in one ear, mild hip dysplasia that may make the dog unable to complete an agility course, but allows the dog a moderate exercise regimen.. 

Major morbidity - Dog would have died without extensive veterinary intervention OR condition severely affects dog's quality of life OR condition requires lifelong medication. Examples: bloat, terminal cancer, blind or nearly blind in both eyes, crippling dysplasia that requires surgery or greatly reduces quality of life.

For each dog in the survey, the owner should be asked to list all conditions the dog has or had, the age of onset, and whether the condition would be classed as minor, moderate, or major.

Example:

Dog: Snuffy, age 8 yrs, 3 mo at time of survey, currently alive.

Owner is asked to rank Snuffy's current health status as either very poor, poor, fair, good, or excellent.

Owner chooses excellent because Snuffy is an active, energetic dog.

Owner is also asked to list all health issues Snuffy has had to date and rank their severity:

In the actual survey, the survey form would have included a long list of conditions for the owner to go through and rank (if applicable to their dog).

In Snuffy's case, he is alive and healthy at the time of the survey.  If Snuffy is a Great Dane, he has exceeded the median average already by nearly 1.5 yrs.  

Yet, the morbidity information reveal that Snuffy, without veterinary intervention, would have been dead at age 7.5.  

With sufficient morbidity and mortality information for a breed, you can construct a table like this:

Fictitious example of summary morbidity/mortality information for breed X, a short-lived breed

In this fictitious case, the median age at death (the age at which 50% of dogs have died) is 7 years. However, at age 7 yrs, 65% of dogs still living have experienced at least one major health problem that would have killed them without extraordinary veterinary intervention, caused a serious decline in the dog's quality of life, or which required life-long medication to keep them alive. 

Thus, by age 7 yrs, only 35% of 50% (the live dogs) of the entire group of dogs (17.5%) is alive and has never had a major health issue.

If the survey is repeated after 10 or 20 yrs, the median age at death might have increased due to better vet treatments.  A comparison of morbidity data, however, will reveal whether the overall health of the breed has actually improved or if declining breed health is being masked by veterinary advances.  Major health issues will not (generally) be prevented by vet med, but vet med can increase the dog's survival rate.