Snippets From the Past: Imaginative Designs—Separating Hereditary From Environmental Effects in pre-DNA Times

Abstract

In an attempt to reproduce the results of an inconclusive 1927 report by the British Medical Research Council on the hereditary versus social origins of rheumatic fever, Read, Ciocco, and Taussig, from Johns Hopkins University, with the support of Frost, conducted a case-control study in 1935 and 1936. Their study, which appeared in the American Journal of Hygiene in 1938, was outstanding for its clear and tidy rational for separating hereditary from environmental causes. The authors compared the prevalence of rheumatic fever among the relatives of 33 children admitted for “incident” rheumatic fever and 33 control children admitted in a tuberculosis clinic for reasons other than rheumatic fever. Both rheumatic fever (cases) and tuberculosis (controls) were diseases of the poor. All family members of both cases and controls, including uncles, aunts, and grandparents, were eligible for interview and physical examination. The results were compatible with the presence of an “inherited predisposition” to rheumatic fever because the disease was more prevalent among the uncles, aunts, and grandparents of case patients than among those of control patients. Methodologically, the paper by Read, Ciocco, and Taussig is an important but almost completely forgotten milestone in the evolution of case-control studies and of genetic epidemiology.

During the early part of the 20th century, chorea, polyarthritis, and carditis, which were manifestations of rheumatic fever, were first suspected and later proved to be secondary to a throat infection with a group A β-hemolytic streptococcus (1). Yale epidemiologist John R. Paul, for example, evolved from considering the causative agent of rheumatic fever elusive in 1930 to attributing it to a streptococcal infection in 1947 (1, 2).

The infectious theory had previously competed with the belief that rheumatic fever had some hereditary antecedents because the disease occurred more frequently among the parents and siblings of rheumatic patients than in the general population (3). However, separating the influence of heredity from that of the environment was tricky then (as it is now). Was the familial clustering attributable to heredity or to relatives sharing the same household environment in close personal contact? (4) Moreover, in ecologic studies, rheumatic fever was a disease of the poor, which was compatible with it having some environmental causes.

In 1927, a 108-page report from the British Medical Research Council compared the familial histories of rheumatic fever and social conditions of 721 families with at least 1 child who was or had been under treatment for rheumatism with the histories of 200 families each containing a nonrheumatic child who was under hospital treatment (5). Eligible families were identified from hospital records. One hypothesis was that if rheumatic fever were hereditary, uncles, aunts, and grandparents of patients would be at a higher risk of developing rheumatic fever even though they usually lived in separate households. Otherwise, if rheumatic fever had a solely environmental cause, the excess risk would be restricted to parents and siblings. Indeed, rheumatic fever was more prevalent among the close relatives of the cases than among those of controls, and no difference was found for the grandparents, aunts, and uncles. Even though the study failed to provide evidence for a hereditary factor, it was interpreted with caution because the information on grandparents, aunts, and uncles provided by the parents seemed inaccurate.

The negative results of this huge-for-the-epoch British case-control study might have prompted researchers in Baltimore, Maryland, to reproduce its family component while improving upon the quality of the family history data. The authors were Frances E. M. Read and Helen B. Taussig from the Cardiac Clinic of the Harriet Lane Home (Department of Pediatrics) of the Johns Hopkins Hospital and Antonio Ciocco from the Department of Epidemiology of the Johns Hopkins School of Hygiene and Public Health. Wade Hampton Frost, Chair of the Department of Epidemiology, was not an author but was acknowledged not only to have “guided” the study but also to have spent many hours analyzing the data. Their paper was published by the American Journal of Hygiene in 1938 (6).

THE CASE-CONTROL STUDY

Between September 1, 1935, and March 30, 1936, 33 white children younger than 15 years of age who had been newly diagnosed with rheumatic fever at the Cardiac Clinic of the Harriet Lane Home (Department of Pediatrics) of the Johns Hopkins Hospital were enrolled in the study. A control group comprised the same number of children admitted to the Tuberculosis Clinic. All of these children were “an unselected series of new patients taken in the order of their admission” (6, p. 720). Moreover, the authors emphasized that “control families were selected without regard to a past history of any rheumatic manifestations” (6, p. 721). The parents of the 66 children were asked to provide the names, dates of birth, dates and causes of death (when applicable), and evidence of manifestations of rheumatic fever for the siblings, parents, uncles, aunts, and grandparents of the children. The family histories were obtained independently by at least 2 and, in the control group, 3 persons (2 physicians and a social worker). An effort was made to interview and physically examine in the Clinic as many of the children's uncles, aunts, and grandparents as possible. Histories of rheumatic manifestations in uncles, aunts, and grandparents, which were validated in medical records, were analyzed separately as “verified histories.”

The relatives of the cases and controls were closely comparable in terms of numbers, age, and number who were deceased. Altogether, they represented 20,586 and 22,111 person-years of life in the case and control groups, respectively. “The two groups [did] not differ in regard to those characteristics which measure the duration of exposure to the risk of contracting a rheumatic infection” (6, p. 725).

Their findings, which have been reproduced in Table 1, were provided as proportions. There was an unspecified measure of spread, indicated by plus/minus, that I was not able to reproduce. Some results were also expressed in the text as rates based on person-years. Some results were reported as ratios of prevalence of familial history of rheumatic fever for cases versus controls by relative type, which was rare before the 1950s.

Table 1.

Case-Control Study of Prevalence of Rheumatic Fever Separately for Clinically Validated Histories of Rheumatic Fever and All Histories, Baltimore, Maryland, 1935–1936

Subgroup	Verified Historiesa						All Histories
	No. of Cases			No. of Controls			No. of Cases			No. of Controls
	Rheumatic	All	%	Rheumatic	All	%	Rheumatic	All	%	Rheumatic	All	%
Siblings	17	110	15.5	5	126	4.0
Parents	18	55	27.7b	4	62	6.2b	20	65	30.8	5	65	7.7
Uncles and aunts	14	49	4.4b	4	55	1.2b	29	319	9.1	13	342	3.8
Grandparents	14	42	11.1b	2	47	1.5b	23	126	18.2	3	130	2.3

Adapted from Read and Gauld (9).

^a Histories of rheumatic manifestations in uncles, aunts, and grandparents that were validated in medical records.

^b Percent of all histories within subgroups.

Rheumatic disease was substantially more common in the parents and siblings of rheumatic children than in the controls, as expected. For example, the frequency of rheumatic manifestations was 15.5% among the siblings of the rheumatic patients and 4.0% among the siblings of the control group. The reported ratio was 0.015/0.040 = 3.9, and “the probability of this difference being due to chance [was] only 3 in 1,000 trials” (6, p. 726).

Irrespective of whether or not the family history had been verified, the percentages of rheumatic disease in the case groups versus the control groups were, respectively, 9.1% versus 3.8% among aunts and uncles and 18.2% versus 2.3% among grandparents. For the latter comparison, “in terms of annual rates based on the life experience of the two groups, the rate of rheumatic infection per 1,000 person-years [was] 3.0 in the grandparents of the rheumatic patients and only 0.4 in the grandparents of the control children” (6, p. 733). Table 2 presents a re-analysis of the data using odds ratios. The results consistently showed all statistically significant associations.

Table 2.

Reanalysis of the Case-Control Study of Prevalence of Rheumatic Fever Separately for Clinically Validated Histories of Rheumatic Fever and All Histories, Baltimore, Maryland, 1935–1936

Subgroup	Verifieda Histories						All Histories
	Rheumatic Group		Control Group		OR	95% CI	Rheumatic Group		Control Group		OR	95% CI
	No. of FH+	No. of FH−	No. of FH+	No. of FH−			No. of FH+	No. of FH−	No. of FH+	No. of FH−
Siblings	17	93	5	121	4.4	1.6, 12.4
Parents	18	47b	4	61c	5.8	1.9, 18.4	20	45	5	60	5.3	1.9, 15.3
Uncles and aunts	14	305b	4	338c	3.9	1.3, 11.9	29	190	13	329	3.9	2.0, 7.6
Grandparents	14	112b	2	128c	8.0	1.8, 36.0	23	103	3	127	9.5	2.8, 32.4

Data from Read and Gauld (9).

Abbreviations: CI, confidence interval; FH−, family history, negative family history; FH+, positive family history; OR, odds ratio.

^a Histories of rheumatic manifestations in uncles, aunts, and grandparents that were validated in medical records.

^b All histories from the rheumatic group within the familial subgroup.

^c All histories from the control group within the familial subgroup.

The authors concluded that a “consanguinity” and “common inheritance” (6, p. 736) had to explain why aunts, uncles, and grandparents of the cases were at higher risk of rheumatic fever than were those of the control groups. Sharing the same environment could not be the cause. Overall, they concluded:

“These findings, although not sufficiently definitive to determine the relative importance of heredity and environment, are indicative of the strong familial tendency to this disease. The fact that such a familial tendency has been manifest for at least three generations is strongly suggestive of the existence of a constitutional susceptibility to this disease. This does not eliminate the possibility that exposure is also an important factor” (6, p. 736).

THE AUTHORS

This was the first publication of Frances Elba Myrtle Read (1911–1982) (Figure 1), who earned a doctor of medicine degree at McGill University in Montreal, Canada (1933), before she went to Johns Hopkins. She was a dispensary physician at the time of the study. She later became a pediatrician, and she practiced in Florida (where she later died). In 1948, she became director of the Bureau of Maternal and Child Health of the Florida State Board of Health.

Figure 1.

Frances Elba Myrtle Read (1911–1982) (middle row) in 1940 in a group picture of the Johns Hopkins Hospital Department of Pediatrics resident and intern staff, 1939–1940. Front left to right: Henry P. Goldberg, Frances Read, Kent Zimmerman (front), and Beach Chenoweth (back). Provided by The Alan Mason Chesney Medical Archives, The Johns Hopkins Medical Institutions.

Antonio Angelo Ciocco (1908–1972) (Figure 2) most likely acted as the statistician of the study. He was in the Departments of Epidemiology and Biology and was employed by the US Public Health Service. In 1953, Ciocco became professor and founding chair of the Department of Biostatistics at the Graduate School of Public Health at the University of Pittsburgh.

Figure 2.

Antonio Angelo Ciocco in 1953 at a meeting of the Parran Alaska Health Survey Team. Provided by C. Earl Albrecht papers, Archives and Special Collections, Consortium Library, University of Alaska Anchorage.

Helen Brooke Taussig (1898–1986) most likely gave access to the clinic for the recruitment of the case series. Many photographs of her can be found online. She was a pioneer of pediatric cardiology, and in 1944 she became associated with the Blalock-Taussig operation to repair a leaking cardiac septum defect that caused “blue baby syndrome,” otherwise known as anoxemia (7). In 1965, she accomplished one of many firsts when she was the first woman to become the president of the American Heart Association.

EVOLUTION OF THE METHODS

There were at least 2 relevant reports on the same topic preceding the one by Read et al. In 1888, Garrod (Sir Archibald E. Garrod) and Cooke compared 500 patients “who had not themselves suffered from rheumatic fever” with 100 “other patients who had at some period of their life suffered from rheumatic fever” and found the histories of rheumatic fever in their “immediate families” to be 19.8% and 35%, respectively (3, p. 110). As was typical of a 19th century report, these group comparisons involved 2 convenience and independent samples of patients. The second relevant report was the British Medical Research Council's 1927 study (5) that was described in the Introduction: a case-control study with children recruited in the same hospitals and comparisons stratified by hospitals. Its weaknesses were the use prevalent cases and the lack diagnostic validation.

An outstanding aspect of the study by Read et al. was the clear and tidy implementation of the rational for separating hereditary from environmental causes. Moreover, recruiting “incident” cases was unusual for that period. The attempt to validate the parents' reports of rheumatic fever in the uncles, aunts, and grandparents was another strength. Even though not explicitly stated, the choice of control children admitted to a tuberculosis clinic must have enhanced the comparability of the cases and controls in terms of their social environment, because both rheumatic fever and tuberculosis were diseases of the poor. Indeed, infant mortality rates of the siblings, a marker of poverty, were similar among cases and controls. The authors stressed that it was “safe to conclude that the two groups of relatives [were] closely comparable” (6, p. 725).

In a follow-up report in 1939 (8), the case group had been expanded from 33 to 96 white children consecutively admitted who were compared with the same 33 white children of the control group. It essentially confirmed the findings of the 1938 study, leaving open the possibility of both hereditary and infectious origins of rheumatic fever. In 1940, Read and Gauld made available the data set for their first 96 cases and relatives but unfortunately not for their controls (9).

The papers by Read et al. (6) and Gauld et al. (8) show a remarkable understanding of the logic of the case-control design when it was still novel. As of today, we are aware of only 9 case-control studies for which results were published before 1938, not including autopsy-based case-control studies (5, 10–17). More studies might be resurrected through a still-to-be-done systematic exploration of the literature of the 20th century. Nonetheless, in this context, the emphatic acknowledgement of the work of Wade Hampton Frost by the authors in both their 1938 and 1939 papers is telling. Frost never co-authored the report of a case-control study. He made it a principle not to co-sign the papers of his students, which he had not written himself (18, 19). However, this rheumatic fever study (and probably another one by Ramsey, as discussed in a previous article (20)) proves that he was familiar with the conduct of case-control studies.

FINDINGS

The small sample size limited the potential inference from the study. For example, the strongest effects were observed in grandparents, which was contrary to what would be expected with a genetic model because grandparents share 25% of the genes with the children, whereas siblings share 50%. In the presence of a single major gene, there should have been a fall in incidence as the degree of relationships moved from parent-offspring or full siblings to grandparent-grandchild or aunts/uncles. The findings are compatible with this genetic hypothesis because the recomputed confidence intervals are vastly overlapping, but the effects are too strong to have been purely genetic.

Even though the choice of tuberculosis as the disease in the controls had the advantage of selecting children with similar social environments as the cases, it made the study vulnerable to the effects of factors related to tuberculosis rather than rheumatic fever. Still, in the study, the authors found a familial aggregation of rheumatic fever, which could argue for a genetic basis. Modern developments of this approach include work by Risch (21) and Schliekelman and Slatkin (22).

MEMORY LOST

The study by Read et al. has been cited 28 times in journal articles as a contribution to the understanding of the heritability of rheumatic fever but never for its pioneering methodology. In 1940, Maxted mentioned as a limitation that “the relatives probably belong[ed] to a similar walk of life as the patients and would thus [have been] exposed to a similar if not the same environment” (23, p. 181). Khoury et al. did not mention the study in their textbook of genetic epidemiology (24).

Gauld, Ciocco, and Read said in the 1939 follow-up report that the findings of the 1938 study “were not conclusive, but did suggest that inherited predisposition may be an important factor in the development of the disease” (8, p. 213). Today we would perhaps replace “inherited predisposition” with multifactorial inheritance in which many different (independent) genes contribute to risk. Nonetheless, almost 80 years later, the genetic bases of the strongly suspected heritability of rheumatic fever have still not been elucidated despite imaginative study designs (25).