Since we were given the CDC birth defects protocol, Birth Defects and Military Service in Vietnam, some time ago, we solicited comments from several interested acientists. Their comments about the protocol have been incorporated into this letter. We trust that we are not too late and that these comments will be of assistance in the development of a scientifically valid protocol and study. The comments are divided according to the following sub-headings: (a) lack of exposure stratification, (b) power and sample size, (c) exclusion of "minor" defects, (d) other considerations, and (e) specific recommendations.

The draft design fails to address in any detail how veterans' exposure categories will be defined or analyzed. Section 4.9, p. 16 indicates that an attempt will be made to "segregate veterans into categories reflecting probability of exposure", by comparing DOD information with unit(s) of service reported by veterans and their wives during the phone interviews. It is crucial that this effort be undertaken. The draft design would devote a great deal of interview time and data analysis to identifying potentially confounding risk factors such as parental drug use, occupational exposures, and familial histories. At the very least, a comparable effort should be made to assess exposure probabilities with the same exactitude.

If exposure categories are not delineated (based on unit and spray records, MOS, combat experience) and the analysis proceeds to use Vietnam service per se as the only exposure

criteria, serious misclassification errors and diminution of study power could result. This is alluded to, but not addressed, on p. 8, section 4.3. For example, if 1/4 (one-quarter)

of all Vietnam veterans in the study population received toxic doses of Agent Orange that doubled their risk of fathering children with major birth defects (the suggested end-point proposed for this study), an analysis without exposure stratification would dilute that doubling to a 25% increased tisk in the overall veteran population. Moreover, the power of the study in this event drops to 70% (i.e., the probability of failing to find a statistically significant difference when in fact there is one, is 30%), instead of the 99% power to detect a doubled risk of all defects shown in Figure 1. Given the current design and analysis, risk among that hypothetical target 1/4 of the veteran population would have to be triple that of the general populace in order to be detected with good sensitivity - and it would be reported as a 50% increase in defects, rather than a 300% increase.

Any epidemiologic study of this population will suffer from reduced power due to misclassification bias, since any Agent Orange exposure index will be necessarily imprecise, but lumping all veterans into a single "exposed" category creates more bias and a greater power reduction than would a reasonably-wrought exposure index.

b. Power and Sample Size

The power computations (p. 26) seem to have been computed on the basis of a 5% background prevalence rate for all defects (in live and stillbirths). However, the study includes among cases only "major" defects in live births, which have about a 2-3% background rate (p. 7). These power computations do not apply to the study described.

c.

No Justification Offered for Exclusion of "Minor" Defects

This study sets out to determine whether Vietnam veterans are at increased risk of siring babies with birth defects (p. 1). Yet it is designed only to determine if Vietnam veterans are at increased risk of siring babies with "major" birth defects. That is, it should have set out and been designed to determine whether Vietnam veterans exposed to Agent Orange experienced any reproductive dysfunction above background rates.

No scientific justification is offered for limiting cases to the diagnostic rubrics noted on p. 7. Perhaps this is because the protocol never defines what hypothesis (es) are being tested.

Little clinical, laboratory, or epidemiological research has been done to date investigating the mechanisms or outcomes of reproductive toxicity in males; particularly little is known about what reproductive dysfunctions might ensue in the decade following chronic exposures to lipid-soluble compounds. that accumulate in the body's fat depots. To assume that only certain categories of congenital anomalies will ensue from chronic reproductive roxicity in males is unfounded, based on our current limited knowledge. This is the standard approach used to determine whether a teratogen to which women were exposed during pregnancy has increased the frequency of a "marker" defect. Where male reproductive toxicity is suspec

ted, a wide spectrum of effects, including infertility, spontaneous abortion, stillbirth, low birth weight, congenital and developmental abnormalities, neonatal and infant mortality, and childhood cancer, should be investigated. Otherwise, we are simply conducting research that gives the "right" answer to the wrong question.

The protocol's only suggestion of a plausible biological mechanism (p. 14, section 4.8) is curious. It proposes a search for a "fresh dominant mutation"; presumably this refers to a perceptibly increased frequency of single-gene dominant disorders caused by an increased mutation rate. First, if one were to conduct this search, it would be senseless to exclude single-gene dominant disorders such as polydactyly and mental retardation from the cases, as the protocol now requires: Secondly, according to current theories, (based on follow-up studies of A-bomb survivors) the sample size of 280,000 live births (and at most 25,000 live births to exposed individuals) is insufficient to detect an increased mutation rate if it existed.

· d. Other Considerations

A major question with the study is whether fathers of out-of-wedlock (OOW) births are to be included in the phone survey. If not, this could introduce serious bias: black OOW births probably exceed white OOW births in the metropolitan area covered; there is reason to believe that blacks were at greater risk of Agent Orange exposure since they had a greater chance of combat experience: Neglecting OOW births might mean missing much of the target population.

Another question is the validity of this case-control design in the event that infertility or abortion has reduced the number of offspring born to the exposed cohort.

If the study is conducted as described in the draft protocol it is virtually guaranteed to discover no serviceassociated risk. Because of the study's design limitations, that negative finding will not be interpretable, i.e., it will not be "of substantial value in easing the concerns of a great many veterans" (p. 16). To accomplish that goal, innovative research programs that make a serious effort to stratify exposure cohorts, investigate the full reproductive histories of those cohorts, and perform systematic karyotyping of veteran's offspring would have. to be conducted. Costs of these programs need not "far exceed the cost of the present study" (p. 9) since power would be appreciably enhanced by more accurate delineation of exposure and outcome variables, based on plausible mechanisms of male reproductive toxicity.

Since this study is being conducted, the following are some changes and additions that could be incorporated into the present design to improve its interpretability:

1.

Include questions in the father's phone interview to shed light on Agent Orange exposure, e.g. Were you in combat? Did you ever have a skin rash following field work? Did you ever drink the surface water? Did you eat local fish or shellfish? local food? (Other questions might be suggested by veterans.)

2. Expand the case population to include all recorded birth defects except those clearly associated with maternal factors, e.g. congenital rubella. This means expanding beyond ICDA 740-759, to include congenital neoplasms, cretinism, disorders, dento-facial anomalies; this may not be possible if only 74-759 are included in the MACDSP.

births.

3. Expand the case population to include all still

4. Retrieve reproductive histories and birth weights for cases and controls from hospital and state vital records and assess risks of abortion and small-for-date births.

them.

5. Retrieve infant and childhood mortality data ¡ for these populations from state vital records and snalyze the current design neglects survival, childhood cancer of veterans' offspring not chosen as index births. Include linkage to state specialized registries also, if possible, to trace developmental abnormalities not noted at birth.

6. Follow up fathers of out-of-wedlock births.

7. Clarify the hypotheses to be tested and recalculate study power assuming various exposure rates in the veteran population.

8. Establish a background rate for adverse reproductive outcomes based on Vietnam-Era veterans who did not serve in Vietnam; this would address the problem of selection of a healthy subset of the population, i.e., "health veteran effect.

9. Analyze temporal patterns and secular trends with respect to time lapsed since Vietnam service, e.g., years until first birth, taking marriage date into account. This could suggest early infertility problems.

I also would like to know the stage of peer review, assuming it has not been completed, and the identity of persons who served on the review panel.

If you have any questions about these comments, please feel free to contact us. Again, I apologize for the delay, but trust that the comments will be useful.

Sincerely,

Lewis Melford

Lewis M. Milford, Esq.

CC:

Les Platt.

John Moore

Patricia A. Honchar

LMM:bas