Battle of Theories

I've read and thought about these two quite a bit.  (This article only examines numbers for Swissies.  These numbers do not apply to other breeds.)  I've done some number crunching. For once I'm actually putting what I learned in school to good use.  Looked at it from a scientific view as if OFA and PennHip were two competing theories. And like all other theories, they aren't perfect. They tend to make broad suggestions about changing the world. 

OFA has his advantages and disadvantages. Advantages being that you get a grade (excellent, good, fair, etc...), and dogs do not get rated unless x-rays are done after 2 years of age. This is not to say that hips can't fall apart after 2 years. Disadvantages are that it is much more subjective, ratings on the same dog can change. There is also that option of keeping your dog awake which affects accuracy of the x-ray. 

PennHip on the other hand is a statistic. It tells you "how loose" the hips are, which are correlated with hip problems. There is no chart on the correlation between PennHip and DJD for Swissies, but there is one for Rotties, which are similar in size/structure. If you look at the distraction indexes that Swissies have gotten (mean at .55 and looking at ranges on the health database), most have anywhere between 5% and 40% chance of developing DJD. Those odds are better than raw chance (50%). Surprise! 

However, I'm guessing from the available DI and percentile information, the distribution of Swissy distraction indexes is leptokurtic, peaks high and narrow. In short, it means that dogs who are slightly tighter than average will be in the high percentiles. Dogs who are slightly looser than average will be in the low percentiles. PennHip never implies that the issue is corrected nor do I known if there are enough Swissies that have done PennHip to fix the issue. 

But if you look at the mean PennHip score, it is right at 20% risk of developing DJD, which is the same statistic that OFA gets for hip dysplasia (19% of Swissies have abnormal hips). That should make you feel better in terms of number crunching. 

As far as correlations between PennHip and OFA reports, it seems pretty low. From my point of view, PennHip doesn't have much predictive power unless you have a dog in the top 10-20% in Swissies ONLY.  Pennhip scores in the 0-80% tile don't mean much of anything.  Dogs in the 20% have OFAed as Good, and Dogs in the 60% have failed OFA.  This is even true of other breeds.  I've heard of border collies who have OFAed in the 80% but still failed OFA.  If PennHip was really going to improve the frequency of hip problems, for Swissies, they should advocate breeding ONLY dogs in the top 20%.  Otherwise, predictive power is loss and the numbers are meaningless.  However, this is somewhat unrealistic, so OFA is another option for health clearances.   

It boils down to doing OFA (that is highly subjective and does not provide any predictive value) or doing PennHip (provides you with some sort of an idea, but since correlations are pretty low, it is more unreliable than OFA's subjectiveness). Layman's terms - OFA tells you whether or not your dog has dysplasia, but not the likelihood of your dog developing it. PennHip tells you the chances your dog might develop hip problems, but not an absolute yes or no. And keep in mind there are dogs who have tight hips according to PennHip and moderate hip dysplasia. And there are dogs who have excellent OFA rated hips, but loose according to PennHip. 

I wouldn't say that one is better than the other. They are both very different in methods and interpretation. In a perfect world, since Pennhip uses an OFA view, you could also get an OFA rating at the same time. Then it would only help us understand how to evaluate breeding stock better. Then again, it is probably like asking Biologists and Sociologists to play nice.
 

New: Post I made on Show-Dog L - OFA vs. PennHip article below.

A while back, there was debating on the merits of OFA and PennHip. I
decided to do an investigation on the results, meaning, and generalizability
of both methods in regards to my breed.

I'm formally trained in statistics specializing in hierarchical multi-group
and nested modeling. Playing with numbers is something I do on a regular
bases, and yes, I love it.

I analyzed the methods and results of both methods and wrote up a short
article based mostly on statistics. Full article in on my website. What I
found in my breed (using a similarly built and sized breed which PennHip has
a large amount of data on a reference group) regarding PennHip was that
Distraction indices in my breed were fairly narrow (little variation).
Also, the majority of the dogs PennHipped had passed OFA ratings.

The majority of dogs in our sample for my breed had a distraction index that
would be correlated to no DJD. In statistics, little variation is a very
very bad problem that could be due to sampling error (self-selection in this
case) as well as a variety of other things. Dogs (using the reference breed
of similar size and build) had the 50%-tile mark had equal chance of dogs
who were OFAed of developing HD - which is 20% of the breed. This
interpretation is good for both PennHip and OFA because it shows that both
methods are reliable measures. This is bad however for people who assume
that all dogs PennHipped over the 50%-tile mark has no chance of developing
HD.

What did this mean in terms of statistics? To me, it means that PennHip in
my breed basically doesn't tell you anything unless the dog was at 80%-tile
or higher. At 80%-tile, the reference breed had little to no chance of
developing HD.

OFA also has inherent problems from age of HD development (if after age 2),
objectivity, quality of x-rays, etc...

Keep in mind that OFA and PennHip is based on research, to which neither
side will admit flaws in their methodology or setbacks in their analyses.
Being in biobehavioral research, I know that all too well. Scientists will
defend their theories until death, and sometimes even from beyond the grave.

Anyways, my interpretations only generalize to my breed. If you would like
for me to examine PennHip statistics for your breed, I'd be more than happy
to do so.