A Failure of Extreme, but Measurable, Proportions
by David Lee Winston Miller
I have absolutely no doubt that the vast majority of people would find the LD50 ("Lethal Dose, 50%") test to be inhumane if they witnessed it:
The basic idea (and practice) of the test (and similar lethal dose tests) is to take healthy animals (usually mice or rats but sometimes dogs, monkeys or other animals) and force feed them enough poison to kill (usually slowly) approximately 50% of them. (Variations include starving the individual before testing, injecting the tested substance, or coating the animal's skin with the tested chemical.)
It is my position that animals have intrinsic worth that deems them ethically unsuitable for this test. (See my web article, Animal Properties.) This position is defensible even if the LD50 were, in fact, a useful and accurate test.
However, this page is not primarily about the ethics of the test. Instead, the accuracy of the test is examined. With respect to accuracy, be prepared to be shocked. Most of the information on this page comes from a scientific review, Significance of the LD50-Test for the Toxicological Evaluation of Chemical Substances1 by G. Zbinden and M. Flury-Roversi. A reading of this paper should cause anyone to rethink their position on this crude test, which dates back to the twenties.
Lets look at the question of variability in LD50 determinations from laboratory to laboratory using the same animal species:
The question of variability (and thus accuracy) has been examined in a large study2 organized by the Commission of the European Communities. The study involved 65 toxicolological laboratories testing five substances in male rats. It is stunning to see the kind of variation that was produced. In many of the sciences, researchers are routinely required to produce work that reduces inaccuracies to less than 1%. And it would be shocking to learn that same-species LD50s (the dose level that kills 1/2 of the test subjects) varied 20 to 40%. However, the study revealed maximum variations much greater: 266 to 1089 percent! (Notes: Depending on the chemical tested. Variation was from smallest obtained LD50 value to the largest. Although not immediately obvious, the 266% to 1089% figures correspond mathematically to the reported max/min ratios of 3.66 to 11.89 depending on the chemical tested. Max/min ratios are determined by dividing the maximum determined value by the minimum determined value. In contrast, the variation was determined here by dividing the difference of the two values by the minimum value. Of course one could argue that the high value, or an intermediate one, might be the "correct" value and the low one an erroneous value. From that perspective, or any other perspective, the error would still be huge.) Lest you think someone left out a decimal point, the inaccuracy of the LD50 is often, indeed, on an order of magnitude.
Keep in mind that these results were obtained from restricting
the tested individuals to the same species and sex. The experiments
were conducted all over again using one hundred laboratories (approx.
80 reporting) and requiring strict changes in procedures
such as restricting the age, weight and feeding conditions of the animals.
Listed below are the results of the "first" and "second"
studies. As you can see, the accuracy of the second study was still dismaying:
Maximum variation ranged from 144% to 738% (corresponding to max/min
ratios of 2.44 to 8.38)--a slight refinement of gross
Variation (by ratio)
between laboratories for LD50's
same species, same sex (male rats)
|Substance||1st or 2nd Study||LD50
|PCP (lipid soluble)||1st Study
| 44- 523
|Na salicylate (water soluble)||1st Study
|Aniline (lipid soluble)||1st Study
|Acetanilide (insoluble)||1st Study
|Cadmium chloride (water soluble)||1st Study
| 70- 513
Adapted from the Zbinden and Flury-Roversi review. *The document reported the PCP LD50 values of the second study as 74-2,328 mg/kg resulting in a ratio of 8.38. The 2,328 figure appears to be a typo. (If the 2,328 value were correct, the inaccuracy would be even worse.)
In other words, even under the strictest controls, using only one species and restricting the animals to only one sex, the results still varied by as much as 8.38 fold!
Note that it might have been better to present the data in a more statistical manner. However, the laboratories were free to use different strains of rats and the extreme variations reported are in line with extreme variations known to exist between different strains of species. For example, one study demonstrated a LD50 for thiourea of 4 mg/kg for one strain of rat and 1,830 mg/kg for another--A stellar, 458 fold difference in the same species..
Clearly the results are not acceptable for human health policy. The variation between non-human animals and humans logically can be predicted to be (and, in fact, is) routinely astronomical. For example, lethal dose levels in humans compared to various animals are 4-100 fold for iodoform, 600-1,000 fold for atropin, and 500-2,000 fold for pilocarpine as revealed in the Zbinden and Flury-Roversi paper.
But what if we restrict testing only to similar species? The results reviewed by Zbinden and Flury-Roversi still boggle the mind. For instance, the oral LD50 for 5-(N-piperidino)10,11-dihydro-5H(a.d.)cycloheptene is 1,160 mg/kg for the male mouse, but only 6.6mg/kg for the male rat. This is not a mere 2 fold difference or just an order of magnitude difference, but rather a 176 fold difference! It should be noted that wild variation between species is the rule, not the exception. This leads to the absurd question of applying LD50 data: Is a human more like a rat or a mouse? Which wild number do we use for humans?
Obviously, at least one of these animals is a poor model for human reaction to this substance. But which one? Defenders of such tests will sometimes reply that they know which animal is the best model because of survival data from accidental human poisonings (emergency-room data). But what kind of answer is that? If you don't know the answer until a number of people are poisoned, what good is the test? Using this logic, emergency-room data could also be used to validate enough spins of a roulette wheel. Obviously, the emergency-room data has validity; a roulette wheel does not.
This leads me to my basic position:
The idea of the LD50 (and many similar tests) is that we can accurately extrapolate from non-human species to the human species. But the fact is, we cannot even make crude extrapolations between similar species. We cannot even make reliable extrapolations between different strains of the same species.
All of this would be absurdly humorous if it were not so profoundly tragic. So frequently we read of people harmed or even killed by chemicals and drugs previously thought to be safe. Animal tests, as a general rule, have failed to assure human safety because of the difficulty of cross-species extrapolation. The LD50 stands out as one of the greatest--and cruelest--failures. And every day, countless animals from rats to cats and dogs slowly die in unmentionable suffering to produce meaningless--if not, dangerous--data that serves only for protection in litigation.
It is the contention of many that litigation is the reason that misleading tests survive despite the scientific evidence against such tests. Hopefully, as the public begins to become more educated on such matters, juries will be more open to expert testimony that challenges the validity of tests like the LD50. Indeed, such tests may become liabilities if the methods used to assure human safety are known to be virtual roulette wheels. The public will demand better methods.
The LD50 (and most tests that rely on cross-species extrapolation, such as the Draize test) can no longer be scientifically supported. Even before the onset of a wealth of alternatives in the last two decades, we should have already asked ourselves if a pharmacologist's educated guess--or even the flip of a coin!--would have served us better than this absurdly inaccurate test. Now, we don't have to answer that question. Although it is sometimes argued that no single alternative can replace these antiquated procedures, it is undeniable that for every individual testing need there exists a humane test--or at least, a battery of humane tests and/or methods--that is more scientifically justified and a better tool for human risk analysis.
If we cannot eliminate one of the most inhumane and scientifically
absurd procedures, what can we eliminate?
"It is the opinion of the NSMR that the routine use of the quantitative LD50 test is not now scientifically justified . . ."
--National Society of Medical Research3
"There is really little scientific justification for the test because reproducibility is not good, it can even vary from day to day, and the results are dependent on the animal strain used . . ."
--P.S. Rogers, Managing Director,
Hazelton Laboratories Europe Ltd.4
". . . it is simply not possible with all the animals in the world to go through new chemicals in the blind way that we have at the present time, and reach credible conclusions about the hazards to human health . . ."
--Joshua Lederberg, Ph.D
Nobel Prize winning geneticist
and President of Rockefeller University5
Note: To explore the philosophical side of this and similar questions, please see my web article, Animal Properties.
1. Significance of the LD50-Test for the Toxicological Evaluation of Chemical Substances by G. Zbinden and M. Flury-Roversi of the Institute of Toxicology, Swiss Federal Institute of Technology and the University of Zurich, published in the Archives of Toxicology (1981) 47:77-99. It should be noted that Zbinden and Flury-Roversi definitely do not take my abolitionist position. However, they are very critical of many LD50 practices and state that "In reality the LD50 is far from being a biological constant. It can vary markedly from one animal species to the other, and within one species of laboratory animals the numerical value of the LD50 determined experimentally is influenced by a large number of factors." I believe that the information revealed in their review supports my position.
2. This large study was organized in 1977 by the Commission of the European Communities. The Zbinden & Flury-Roversi review contained the following reference about the study: "Hunter W J, Lingk W, Recht P (not dated) An intercomparison study conducted by the Commission of the European Communities on the determination of the single administration toxicity in rats Communicated by the Health and Safety Directorate. Commission of the European Communities (unpublished document)."
3. From an NSMR policy statement dated 1982. (As quoted by Friends of Animals in a brochure, "The Case Against the LD*50 Test").
4. Letter to Royal Society for the Prevention of Cruelty to Animals, June 16, 1977. (As quoted in the before mentioned Friends of Animals brochure.)
5. Speech given at the World Environment Center, 1981.
(As quoted by Friends of Animals above.)
Copyright 1996, David Lee Winston Miller. This document may be reproduced provided that it is duplicated in its entirety including all links, credits and copyright notices. This page has been accessed times since 11-06-96.