Skip to main content

Dane County Board

The Board of Dane county, which includes the city of Madison, WI, location of the University of WI and the Wisconsin National Primate Research Center, is exploring whether research using monkeys is appropriate. (For details, see an article in the Wisconsin State Journal.) I applaud this effort and trust the Board will examine the scientific issues surrounding using monkeys in research as well as the ethical issues.

As I have stated numerous times in this blog and elsewhere, nonhuman animals including monkeys and chimpanzees cannot be used to predict human response to drugs and diseases like HIV. Occasionally an animal, such as a species of monkey, will respond to a disease like humans respond but 1) this can only be ascertained in retrospect and 2) in order for a modality such as using monkeys to be considered predictive in science it must have a track record of getting the right answer. One cannot cite the outcome from a single experiment in monkeys that eventually was shown to correlate with humans and conclude that experiments in monkeys per se are predictive for humans. Hopefully the following will shed some light on this.

Six different horses run in each of six races for a total of 36 possible winners in 6 races (see table). Obviously only one horse will win each race so there will be 6 winners and 30 losers. Also assume there are people outside the race track selling 6 different “tip sheets,” each of which purport to predict the winners of the 6 races. Let’s examine the table and see how the tip sheets did.


Race    Horse name                            

1          AppleBertha  Charlie   Devil  Everready   Fastest       

2          GoBoy  Hell’sFury   Ichabod   Jumper   Kaloza   Loper   

3          MinuteMan Noshow  OutOfHere Prancer  Quest  Racer  

4          Stoned  TurnItOn  Unbridled  VictoryWinIt    X-Bet     

5          Yazoo  Zeke     Accelerate     BreakAway    CloseTheGapDiablo           

6          Explode     Frank     Gertrude Henry     Ivan   Jay      

Tip Sheets       Winner’sBestBet’sHappyday’sGold’s Trifecta’sA+++

1                      Apple              Charlie             Bertha              Bertha  Devil             Apple

2                      Kaloza             Hell’sFury        Kaloz               Loper   Loper             GoBoy

3                      Quest              Racer               OutOfHere      Prnacer Quest            Quest

4                      WinIt               WinIt               Stoned             Victory WinIt              WinIt

5                      Diablo              Zeke                Diablo             Diablo CloseTheGap   Zeke

6                      Ivan                 Ivan                 Ivan                 Jay       Jay              Ivan


1 Apple

2 Hell’sFury

3 OutOfHere

4 Victory

5 CloseTheGap

6 Henry

In race #1, two tip sheets predicted (I am using the word predict the way nonscientists use it for this example) the winner. In race #6 no tip sheets predicted the winner and in all the other races 1 tip sheet predicted the winner. Collectively the tip sheets predicted 6 winners in 6 races. If one only examines that particular aspect of the data 6 out of 6 looks pretty good. But in fact the tip sheets, collectively, predicted only 6 out of 36 possible winners. That is a success rate of around 17%. (There is a more scientific way to measure and express this, see below.)

If we examine the tip sheets individually we see about the same rate of success for each.

Each predicted 1 winner but no more and there is no correlation between how many tip sheets predicted a certain horse and whether that horse won. In other words the tip sheets have no useful function other than to get the potential gambler to part with his money.

Results from studies comparing human outcomes with animal outcomes reveal a similar degree of predictability. (See Animal Models in Light of Evolutionfor more on the above). In medical science one cannot claim a modality is predictive (in the scientific sense of the word) when that modality only gets the right answer a small percentage of the time. In medical science, we use statistical measures like positive and negative predictive value, sensitivity, and specificity when discussing this topic. This is how we judge whether a test or modality is predictive for humans and therefore whether we will use it. When the vested interest groups point to a single correct correlation or even a group of correct correlations, this is meaningless without knowing the above statistics. When the above statistics are known they reveal that animal models, including our closest cousins—monkeys and chimpanzees—are simply not predictive for drug and disease response in humans. Study them all you want, you are not going to make new drugs safer or predict how a disease is going to affect humans.

The downsides of pretending that animals are predictive include losing good drugs, increasing the cost of drugs in general, misleading researchers which results in harm to humans, and wasting resources that could have been spent on human-based research. These are not insignificant costs to the continued use of animals like monkeys.

See Animal Models in Light of Evolutionfor a more scientific explanation of the above.


Popular Video