The late Karl Popper, the premier philosopher of science, [3] said that science consists of two activities. The first is the formulation of a hypothesis that is a tentative description of the way a part of the universe operates. The second activity is testing the hypothesis. Without testing, we will have no idea of whether the hypothesis is or is not a good description of reality. Importantly, hypothesis testing cannot depend on picking and choosing data because picking and choosing leads to ignoring information that might favor the null hypothesis. Popper and other scientists have emphasized the importance of disproving the null hypothesis, which is a shorthand way to refer to the idea that the hypothesis that is being considered is wrong and evidence must be produced to support it. [4]
IOM considered a possible link between Agent Orange and spina bifida in both its 1993 [5] and its 1996 report. In 1993, it suggested no link, and, as we all know, in 1996, it concluded that there was “limited/suggestive evidence for an association between exposure to herbicides used in Vietnam and spina bifida in offspring” (IOM 1996 at p. 1–8). The reason for the changed conclusion was information in the Ranch Hand study (IOM 1996 at p. 9–8).
The Ranch Hand study compared birth outcomes of the children born to Air Force personnel with known exposures to herbicides with the outcomes for children born to a Comparison group that had no herbicide exposure. What would we expect if exposures to herbicides had absolutely no impact on birth defects and we compared the kinds and numbers of birth defects in the children of a group of herbicide‐exposed men to the birth defects in a group of non‐herbicide‐exposed men? Would we expect exactly the same kinds of birth defects? Or exactly the same numbers? Of course not, based on chance, we would expect that some birth defects would be more common in the exposed group and that some would be more common in the non‐exposed group.
That is exactly what we find when we look at the results of the Ranch Hand study. There is no statistically significant difference between the frequency of birth defects in the Comparison and Ranch Hand populations. The ratio of birth defects/child was 0.21 in the Comparisons and 0.22 in the Ranch Hands. When comparisons were made of the frequency of birth defects in different organ systems–nervous system, genital system, digestive system, etc.–birth defects in four organ systems were more common in the children of the Comparisons and birth defects in the other six organ systems were more common in the children of the Ranch Hands.
Two specific anomalies and two developmental disabilities were common enough in the children of Ranch Hands for statistical analysis on the bases of the fathers’ herbicide exposures. The frequencies of the two specific anomalies–major birth defects and multiple birth defects–were comparable in the children of the Comparisons and Ranch Hands. Six percent, of the Comparisons’ children and eight percent of the Ranch Hands’ children had major birth defects, and 0.4 percent of the Comparisons’ children and 0.6 percent of the Ranch Hands’ children had multiple birth defects. [6] The small numbers of these defects makes the differences in percentage not statistically different from each other.
To examine the possible effect of increasing herbicide exposures, the Ranch Hands were divided into three groups–(1) those with no evidence of exposure above background levels, (2) those with low exposures, and (3) those with high exposures. If herbicides cause birth defects, we would expect the frequency of the birth defects to increase along with exposures. That is not what was found. Instead, major birth defects and multiple birth defects were most common in the children of Ranch Hands with low exposures. There is no way to explain those results as being related to herbicide exposures. If increasing exposures caused those birth defects, the defects would be more common in the children of the “high” exposure group of Ranch Hands. The most reasonable explanation is that herbicides had no effect on these birth defects and that the distribution of birth defects in children of men with background, low, and high exposures was a matter of chance.
Examination of the two developmental disabilities that were sufficiently common for statistical analysis supports the conclusion that there is no relationship between herbicide exposures and frequency of birth defects. [7] Delayed development was more common in the children of the “low” exposure group than in the children of fathers in either the “background” or “high” exposure group. Moreover, delayed development was more common in the children of the “background” exposure group than in those of the “high” exposure group. Hyperkinetic syndrome was most common in the children of the background exposure group.
Rather than dwelling on the differences in the frequency of these anomalies and developmental disabilities in the children of men with different exposure histories, it makes more sense to say that there is no relationship between exposure and these anomalies and disabilities. Indeed, that is how IOM interpreted those data.
IOM interpreted the spina bifida data differently. The Ranch Hand study reported the numbers of 14 specific anomalies or developmental disabilities identified in the children of Comparisons or Ranch Hands or both for which “Counts and rates…[were] too sparse to analyze…” (Wolfe et al. 1995, p. 20).
Two neural tube anomalies–one case of anencephaly and three cases of spina bifida–occurred only in the children of Ranch Hands. Those two conditions account for four of the five cases of nervous system anomalies reported in Ranch Hand children. There were three nervous system anomaly cases in the Comparisons’ children. In addition, two other birth defects–one case of polydactyly and one case of reduced limb deformity–occurred only in the Ranch Hands’ children.
In contrast, cleft palate and cleft lip/palate occurred only in the Comparison children. In the past, these conditions had often been suggested as possibly associated with dioxin because exposure of pregnant mice to dioxin increased the frequency of cleft palate in their offspring. [8] There is, however, no evidence for exposure of male breeding mice to Agent Orange having any effect on their offspring. [9] The Ranch Hand study is consistent with the conclusion that parental exposure does not cause cleft palate. In addition to cleft palates that occurred only in Comparisons’ children, the single case of hydrocephalus that was seen in the study occurred in the child of a Comparisons.
Given the distribution of birth defects between Comparisons and Ranch Hands, there is as much logic in suggesting that exposures to herbicides protects against cleft palate and hydrocephalus as there was in IOM’s concluding that the results about spina bifida are suggestive of an association.
Neither the authors of the Ranch Hand study, the Department of Health and Human Services committee that reviewed the study before its publication, [10] the reviewers and editors of the journal Epidemiology that published the Air Force study, or a scientist who commented on the Air Force study for Epidemiology [11] concluded that there was “limited/suggestive” evidence for a connection between herbicide exposure and spina bifida. Indeed, none of them mentioned any evidence about that possible association at all.
Why did IOM reach its conclusion that the evidence was “limited/suggestive”? In my opinion, the reason is obvious. IOM assigns that classification when
Evidence is suggestive of an association between herbicides and the outcome but is limited because chance, bias, and confounding could not be ruled out with confidence. For example, at least one high‐quality study shows a positive association, but the results of other studies are inconsistent. [12]
That criterion flies in the face of Popper’s and other scientists’ insistence that all the data be considered and weighed together. In fact, it literally throws out any consideration of the null hypothesis because it lets the analysts focus on a single, isolated finding as the proof of their case. This is bad science. It was not, as is sometimes suggested, forced on IOM by Congress. IOM set its criterion of “one good study” for itself.
There is no known biological mechanism by which parental exposure to herbicides can cause spina bifida in the veterans’ children. [13] Neither 2,4-D nor 2,4,5-T, the principle components of Agent Orange, [14] nor dioxin [15] is a mutagen, and none can affect the DNA in a sperm cell. Both 2,4-D and 2,4,5-T are eliminated from the body in a period of weeks, so that those chemicals would not have been present in the fathers’ bodies when children were conceived after Vietnam service. Dioxin, on the other hand, is very persistent in the body. Since dioxin cannot affect DNA, the only method by which it could cause birth defects is by being transferred from the father to the mother through semen. The few molecules of dioxin that would be transferred would be added to the trillions more molecules in the mother’s body. How could that tiny addition have an effect? It could not.
IOM states (IOM 1996 p. 9–18) that “Laboratory studies of the potential developmental toxicity…of TCDD as a result of exposure to adult male animals are too limited to permit conclusions.” This may be a correct summation, but it ignores the study by Lamb et al. that showed that feeding Agent Orange to male mice caused no birth defects even when it caused toxic effects in the male mice.
The IOM also brushed aside the study of 15,291 births to residents of Seveso, Italy. Seveso was the scene of a chemical plant accident in 1976 that spread dioxin and other chemicals over an area with a population of 37,000. In 1988, a group of Italian physicians published a paper that compared the frequency and kinds of birth defects in children born in the Seveso area in the six years after the accident to those in children born in the surrounding, uncontaminated area during the same period. [16]
The contaminated area of Seveso is divided into three zones, A being most contaminated, B less contaminated, and R still less contaminated. The noncontaminated comparison area is called nonABR.
I am going to use the IOM interpretation of the Air Force study as the basis of a hypothesis and use the Seveso results to test it. This exercise fits into Popper’s description of how science is done, and it is nothing that IOM could not have done. It is something that IOM should have done.
The Air Force reported 5 nervous system anomalies in the Ranch Hands and 3 in the Comparisons. If there is a relationship between dioxin exposure and nervous system anomalies, those birth defects would be expected to be higher in the children born to the exposed parents at Seveso. There were 2 such defects among the 2900 children born to the exposed parents and 22 in the 12391 children born to the unexposed parents. The frequency of such birth defects is 0.07 percent in the exposed group and 0.17 percent in the unexposed group, [17] or 2.5-times higher in the unexposed group. The Seveso data provide no support for the idea that dioxin exposure causes central nervous system defects.
Some dioxin exposures were higher at Seveso than those experienced by the Ranch Hands (see table 1), more people were exposed at Seveso, and, very importantly, both men and women were exposed. Table 2 presents the results of calculating the expected number of cases of spina bifida in the Seveso population if IOM’s conclusion about the relationship between dioxin and spina bifida is correct. (Blood samples were obtained from many Seveso residents, and as they are assayed for dioxin, information about exposures there will become more certain.)
The absence of spina bifida from the residents of Seveso Zones A and B and its absence or near absence from Zone R is a strong argument against the conclusion reached by IOM. In contrast to the male‐only exposure of the Ranch Hands, both sexes were exposed at Seveso, and couples composed of highly exposed men and women have had children. [18] The animal experiments indicate that sufficiently high exposures of pregnant mice and rats can cause birth defects. If that is the case in humans, even the exposures of the women at Seveso were not sufficiently high to increase the rates of neural tube defects or overall birth defects (Mastroiacovo et al. 1988). The animal experiments provide no evidence for dioxin causing birth defects when administered to the male parent, and there is no biological theory to explain how dioxin could have that effect. The absence of excess birth defects in the Seveso population conforms with predictions from the animal experiments and with theory. The Seveso study contradicts the IOM’s conclusion.
Conclusion
No biological mechanism is known that would explain how dioxin exposure of men could cause birth defects in their children; a well‐done animal experiment demonstrated that exposing male mice to herbicides did not increase birth defects among their offspring; none of the other experiments that IOM cites as supporting its conclusions about a “limited/suggestive” association between dioxin and spina bifida has any verifiable information about exposure and there is no evidence that men classified on the basis of records as likely exposed were actually exposed, [19] the Ranch Hand study is best interpreted as showing no connection between paternal dioxin exposure and birth defects, and the absence of spina bifida from the Seveso population with higher exposures and many more births than in the Ranch Hand study directly contradicts IOM’s conclusion. The IOM committee misled itself by deciding to use the criterion that one good positive study would provide sufficient evidence for a limited/suggestive association. That decision when combined with picking through the available data and selecting one finding that supports their conclusion is not scientific. It led IOM to a decision that has no support.
Table 1
Dioxin Levels in the Ranch Hands and the Seveso Population Ranch Hands, calculated initial dioxin concentrations.b
Parental Concentrations (ppt)______________
Classification Na mean 75%tile maximum
background 283 — — <10
low exposure 241 60 — 109
high exposure 268 294 — 2020
Seveso population, measured concentrations in samples taken soon after exposure
Parental Concentrations (ppt)_______________
Classification Na mean 75%tile maximum
Zone A 198c ca. 500d ca. 2000d 56,000e
Zone B 435f ca. 125d — —
Zone R 2439f ca. 60g — —
__________
a. Number of children born to parents in the indicated classification.
b. Joel Michalek, principal investigator, Air Force Ranch Hand study, email Aug. 28, 1996.
c. Number of births through December 1994 reported in Mocarelli, P., P. Brambilla, P.M. Gerthoux, et al. 1996. The Lancet 348: No birth defects have been reported among the children born to parents who lived in Zone A, although there was an excess of female births during the first seven years after exposure. In contrast, less than 50 percent of the children born to Ranch Hands were female (Joel Michalek, email, September 10, 1996.)
d. Approximations supplied by Larry Needham, Centers for Disease Control, telephone conversation, Sept. 9, 1996.
e. Mocarelli, P., D.G. Patterson, A. Marocchi, and L.L. Needham. 1990. Chemosphere 20:967–974.
f. Mastroiacovo, P., A. Spagnolo, M. Ernesto, et al. 1988. Journal of the American Medical Association 259:1668–1672.
g. There are no published values for dioxin concentrations in residents of Zone R. I have approximated the concentration at 1/2 the Zone B concentration because: The dioxin level in the soil of Zone R is 1/3 the level in Zone B and animal mortality and the prevalence of chloracne (a skin disease that is indicative of dioxin exposure) are essentially the same in Zones B and R (Mastroiacovo et al. 1988. Journal of the American Medical Association 259:1668–1672.)
Table 2
Expected Numbers of Cases of Spina Bifida at Seveso If the IOM’s Conclusion is Correct Compared to the Reported Numbers
Area at Seveso Calculated Number Observed Number
of spina bifida casesa of spina bifida casesb
Zone A 2.4 to 6.5 0
Zone A (75%tile)c >2.4 to >6.5 0
Zone B 1.3 to 3.6 0
Zone R 3.5 to 9.7 0
___________
a. 1 of 241 children fathered by Ranch Hands with low exposures (60 ppt dioxin) had spina bifida, and 2 of 268 children fathered by Ranch Hands with high exposures (294 ppt dioxin) had spina bifida. The expected frequency of cases of spina bifida in the Seveso children based on those relationships are calculated as:
frequency spina bifidarh = predicted frequency spina bifidas
dioxin concentrationrh dioxin concentrations.
For instance, the frequency of spina bifida in the low exposure Ranch Hands is 1/241=0.4%, and their dioxin concentration is 60 ppt. The dioxin concentration of residents of Zone A is 500 ppt, and the relationship becomes
0.4% spina bifida = X% spina bifida, and X = 3.3%.
60 ppt 500 ppt
For the high exposure Ranch Hands the relationship is
0.7% spina bifida = X% spina bifida, and X = 1.2%.
294 ppt 500 ppt
Multiplying a frequency times the number of births yields an estimate of the expected number of births with spina bifida. For instance 198 births in Zone A multiplied times the expected frequency of 3.3% is 6.5, and multiplied by the expected frequency of 1.2% is 2.4.
b. No birth defect was reported in Zone A. One neural tube defect was reported in Zone B; from other information in Mastroiacovo et al. (1988), it appears that that birth defect was a brain tumor and not spina bifida. One neural tube defect was reported in Zone R, and it is impossible to tell from the paper if that defect was a spina bifida. At my request, a scientist at the Centers for Disease Control contacted Dr. Mastroiacovo and asked him if any of the neural tube defects at Seveso was spina bifida. The scientist emailed on September 16, “I e‐mailed Dr. Mastroiacovo but unfortunately all the records are in storage and he is leaving today for a meeting. I’m sure he would be happy to try to get the information for you at a later time, but he won’t be able to access them before the hearing.” (I’m surprised that this important information is not readily available.) However, an Italian scientist visiting Dr. L. Needham at CDC stated that there was no spina bifida in zones A, B, or R (telephone conversation, September 16, 1996).
c. The 75th percentile measurement in Zone A is about 2000. These calculations were based on an assumed average exposure of 2000 ppt, which is too low, and the assumption that the 25 percent of people with high exposures were parents of 25 percent of the children in Zone A.
Notes
[1] Wolfe, W.H., J.E. Michalek, J.C. Miner, et al. 1995. Epidemiology 6:17–22.
[2] Institute of Medicine. 1996. Veterans and Agent Orange: Update 1996. National Academy Press: Washington, DC.
[3] Bondi, H. 1992. The philosopher for science. Nature 358:363.
[4] Popper, K., in Miller, D. (Ed). 1985. Popper Selections. Princeton University Press: Princeton, NJ. pp. 133–151.
[5] Institute of Medicine. 1993. Veterans and Agent Orange. National Academy Press: Washington, DC.
[6] There were 56 major birth defects in the children of the Comparisons; 59 in the Ranch Hands and 4 cases of multiple birth defects in the children of both the Comparisons and Ranch Hands.
[7] There were 71 cases of delayed development in the children of both the Comparisons and Ranch Hands and 32 cases of hyperkinetic syndrome in the children of Comparisons and 30 cases in the children of Ranch Hands.
[8] Courtney, K.D. and J.A. Moore. 1971. Toxicology and Applied Pharmacology 20:396–403.
[9] Lamb, J.C., J.A Moore, and T.A. Marks. 1980. Evaluation of 2,4-dichlorophenoxyacetic acid (2,4-D) and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) toxicity in C57BL-6 mice: Reproduction and fertility in treated male mice and evaluation of congenital malformations in their offspring. National Toxicology Program: Research Triangle Park, NC. [NTP-80–44].
[10] I chaired that committee when it reviewed the Air Force birth defects study.
[11] Lindbohm, M-L. 1995. Epidemiology 6:4–6.
[12] Institute of Medicine. 1993. Veterans and Agent Orange. National Academy Press: Washington, DC. At p. 6.
[13] IOM soft‐pedals this idea, but it is clear from its discussion in the 1993 report at pp. 593–595 that any connection between paternal exposures to nonmutagenic agents and effects in offspring is speculative.
[14] Mortelmans, K., S. Haworth, W. Speck, et al. 1984. Toxicology and Applied Pharmacology 75:137–146.
[15] Environmental Protection Agency. 1994. Health Assessment Document for 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds. USEPA: Washington, DC. [EPA/600/BP-92–001a.] Volume I of III at pp. 6–12 to 6–14.
[16] Mastroiacovo, P., A. Spagnolo, M. Ernesto, et al. 1988. Journal of the American Medical Association 259:1668–1672.
[17] I included “neural tube defect,” “microencephaly,” and “other CNS defects” in this tabulation. Two other categories, “eye defects” and “ear defects” that involve the nervous system are also tabulated by Mastroiacovo et al. Adding those in does not affect the conclusion about no relationship between dioxin exposure and nervous system effects, but it would increase the number of birth defects from 2 to 5 among the exposed group and from 22 to 39 among the unexposed group. The corresponding frequencies are 0.017 among the exposed group and 0.031 among unexposed group.
[18] Mocarell, P., P. Brambilla, P.M. Gerthoux, et al. Change in sex ratio with exposure to dioxin. The Lancet 348: .
[19] Centers for Disease Control. 1988. Journal of the American Medical Association 260:1249–1254.