Neonatology on the Web

Retrolental Fibroplasia: A Modern Parable – Notes and References

Retrolental Fibroplasia: A Modern Parable – Notes and References


Peller (1948) coined the term “perinatal” to refer to the period around and including birth. In recent years the adjective has been used to refer to the unity of fetal and newborn medicine. Chain was quoted in Annotation (1975). Lillian Hellman (1976) has pointed to a peculiarly American characteristic:

We are a people who do not wish to keep much of the past in our heads. It is considered unhealthy in America to remember mistakes, neurotic to think about them, psychotic to dwell upon them.
See Popper (1962) for the advice concerning critical rationalism.

Chapter 1 — A New Affliction in Premature Infants

Doctors Clifford and Chandler kindly responded to my request to recall the details of their first encounter with RLF. Mr. James H. Staton, Director of the Boston Hospital for Women, sent photocopies of the 1940 Boston Lying-In Hospital records of the “first” two infants who were discovered to have RLF (Figs. 1-1 and 1-2), on condition that patients’ names be omitted in print. History buffs are referred to Spencer (1955) for these censored details. For the original description of RLF see Terry (1942); he reported a large series of patients three years later (see Terry 1945). Krause (1946) described an association between RLF and congenital malformations of the central nervous system, but there is some doubt about the diagnosis of RLF in this series. For the first description of the serial changes of RLF seen with a direct ophthalmoscope see Owens (1948). Figure 1-3 is a print of the colored plate which appeared in this classic article. The extreme peripheral areas of the curved retina (where the earliest changes of RLF occur) are difficult to visualize with the direct ophthalmoscope. In recent years, the technique of indirect ophthalmoscopy has been used to examine the eyes of young premature infants. With the latter method the observer sees a larger field of view than with the direct ophthalmoscope, though at smaller magnification and the image is inverted. The earliest changes of RLF (see p 192) seen in the terminal portions of the retinal vessels precede those seen with the direct ophthalmoscope. For a description of the changes seen with modern equipment see Figure 9-1. Figure 1-4 is taken from Patz (1957a): copyright American Academy of Pediatrics 1957.

Chapter 2 — Evolvement of Care for Feeble and Prematurely Born Infants

See Harris (1977) for the anthropologic evidence concerning infanticide. Ancient Roman Law (The Fourth Table of the Justinian Code contains the relevant passage) was quoted in Sandars (1876). Tauber (1958) reviewed the history of infanticide in Japan. See McKeown (1976) for the decline of infanticide as a factor in modern population increase. For a description of early English caretaking, see Commentary (1897). The old quatrain, from Sainte-Marther’s book “The Art of Nursing and Rearing Children,” was quoted in Lipton (1965). Chaussier’s experiment was cited by Zuntz (1906). For the use of mother’s milk in enemas and baths see Meissner (1838). The Russian incubator was described by Fuerst (1887); the French version, shown in Figure 2-1, is in Denuce (1857). See Dunham (1957) for an account of the history of premature infant care. Ballantyne (1902) described the growth of interest in premature infants in England. Budin’s book (1900) was the primer for turn-of-the-century physicians who were interested in this subject. Figure 2-2 is taken from Fuerst (1887); it was reproduced in my article on incubator-baby side shows (see Silverman 1979) and appears here with permission of the American Academy of Pediatrics (Copyright 1979). The quote from the French Academy appeared in Commentary (1897). See Winckel (1882) for the womb-like device, Figure 2-3, and Hess (1922) for the Colerat citation. The quotes from Budin appeared in his book (1900). I reported the story of incubator-baby side shows (Silverman 1979); Figure 2-4 appears in that article; copyright American Academy of Pediatrics 1979. According to an 1896 article in The Strand Magazine of London, Doctor Alexandre Lion established premature infant exhibits (Oeuvre Maternelle des Couveuses D’Enfants) in a number of cities in France prior to Couney’s first show in Berlin. If this account is true, it casts doubt on Couney’s story about the origin of the side-show phenomenon.

Chapter 3 — The First Decade of RLF

See Owens (1949a) for the Baltimore study, Hess (1934a) for the Chicago experience, and Kinsey (1949) for the U.S. survey. Reese (1949) found an RLF-like condition described by Collins (1925) as “an opaque membrane behind the lens”. Between 1925 and 1937 sporadic examples of (probable) RLF occurred under these typically obscure terms: “metastatic retinitis,” “extrauterine endophthalmitis and iridocyclitis,” “congenital falciform fold,” “shrunken fibrous tissue cataract,” “congenital connective tissue formation in the vitreous chamber,” and “fibrous tissue cataract.” Unsworth (1948) cited a description resembling RLF published in 1820. Table 3-1 is from Lowenfeld (1947) and Figure 3-1 is redrawn from his later article (Lowenfeld 1959). The Boston details and Figure 3-2 (redrawn) are taken from Kinsey (1949). For the vitamin E experience in Baltimore see Owens (1949b) and, in Boston, see Kinsey (1951). After a hiatus of more than 20 years this early experience with vitamine E (alpha tocopherol) prophylaxis has suddenly assumed new importance. Beginning in 1974 (see Johnson 1974 and Phelps 1977), re-investigation of a possible protective effect of vitamin E has raised new hopes for this substance which acts as an antioxidant. For observations of spontaneous regression see Owens (1953). In 1943, Terry suggested that the exposure of the incompletely developed eye to light was, possibly, a cause of RLF. The role of light was discussed for nine years by almost all observers, and anecdotal experiences were cited repeatedly. For example, the frequency of occurrence of RLF appeared to increase with the number of examinations of the eyes of infants with the bright light of the ophthalmoscope. However, no one had examined the question by means of a formal trial. Finally, Locke and Reese (1952) covered one eye in each of 22 premature infants in the nursery of Babies Hospital, New York. The occlusive patch was applied within 24 hours after birth and remained in place until the infants were sent home, 22-75 days later. There was no evidence that RLF could be prevented by occluding light to one eye. Both eyes were then patched in 33 small infants; again, the results failed to implicate light.

Chapter 4 — The Oxygen Hypothesis

See K. Campbell (1951) for the quote and the Melbourne data in Table 4-2; Crosse (1951) for the Birmingham data in Table 4-1, and Evans (1951) for the comment about a political evil. Ryan (1952) commented on the introduction of an efficient cot in Melbourne. The contrary experience in Oxford was described by Houlton (1951). The oxygen-lack proposal appeared in Szewczyk (1951). The thesis was roughly the same as one proposed earlier by Ingalls (1948 and 1952) who considered RLF to be a type of congenital malformation caused by inadequate oxygenation before birth or at delivery. Ingalls induced ocular malformations by exposure of pregnant mice to low-oxygen environments, but the changes produced in the developing eyes were not accepted by all critics as typical of RLF. The use of supplemental oxygen to “treat” early RLF was reported by Szewczyk (1952). For the New Orleans report see Exline (1951). The actual concentration of oxygen in the incubators of the New Orleans nursery was not measured. The convective-type incubators used there were of a primitive design. In retrospect, there is reason to suspect that open louvres in these incubators prevented accumulation of significant amounts of supplemental oxygen. For the Paris report and Table 4-3 see Lelong (1952). In a letter to me, Doctor Patz related his conversation with Hoeck, the response to his application, and the nurses reactions. A description of his initial trial, Table 4-4, and the concluding statement appear in Patz (1952).

Chapter 5 — The Eye and Oxygen

The embryonic development of the eye was described in Mann (1928). The hyaloid artery forms the fetal intraocular blood system. It arises from the ophthalmic branch of the internal carotid artery, enters the embryonic eye and courses through the vitreous to the posterior surface of the lens. There it breaks up into capillaries which form a network around the lens (the tunica vasculosa lentis). During the sixth month, all vessels of the tunic atrophy, except the hyaloid artery trunk. In the seventh month of gestation, the hyaloid artery shrinks; about the time of birth the artery has disappeared and the permanent retinal circulation is established. Michaelson (1948 and 1954) described the development of patent vessels. (His injection method of study revealed only the vessels which had formed a channel, not the cellular events which preceded this stage of development.) Later it was demonstrated (by means of retinal digest and special staining techniques) that the retinal vessels do not develop by a process of budding from the hyaloid artery system, but by preliminary invasion of primitive mesenchymal cells which first appear in the vicinity of the hyaloid artery (see Ashton 1970). The mesenchymal cells differentiate and form a capillary network. According to the present concept, the retinal arteries and veins arise from the capillaries (not the reverse as previously supposed). For a present-day description of the “timetable” of retinal vessel growth see Cogan (1963). Ashton (1970) notes that the current concept concerning perivascular capillary-free zones in the retina is different from that put forth by Michaelson. These clear areas appear to develop through capillary retraction rather than by inhibition of capillary growth. Nonetheless, the fact that the zones can be narrowed and widened by lowering and raising ambient oxygen concentration supports the earlier view that oxygen is “antivasculogenic”. See F.W. Campbell (1951) for the low-oxygen observations in rats; also Ingalls (1952). High-oxygen effects in mice were described by Gyllensten (1952); salt and water effects in kittens were noted by Hepner (1952). For the clearest report of the oxygen effect see Ashton (1953, 1954 and 1966). Vaso-obliteration in the kitten began in less than 12 hours of exposure to high oxygen and was complete by 36 hours. The obliterative effect was directly related to the duration of exposure in short periods (12, 24 and 36 hours), but was not conclusively shown in studies of longer periods of exposure. Vasoconstriction produced by oxygen exposure was reversible, but total vaso-obliteration, once induced, remained unaltered. Later studies, using a trypsin-digest method to study the capillaries of the retina, indicated that vaso-obliteration begins with capillary closure after about 6 hours’ exposure to hyperoxia. This is soon followed by degenerative changes in the cells of the developing vessels. The pathologic changes of blood vessels in human eyes was described by Reese (1951). Figures 5-1 and 5-2 are redrawn from illustrations in Scientific American (Silverman 1977). In a later study, Ashton (1965) found that the proliferative phase was the result of capillary obliteration and not a specific effect related to hyperoxia. When the newborn kitten’s retinal vessels were occluded by glass beads, proliferation of vessels into the vitreous occurred (exactly as found after obliteration produced by oxygen exposure). Other animal studies were described by Patz (1953 and 1957a).

Chapter 6 — The National Cooperative Study

There is no reliable accounting of the RLF epidemic. Dependable statistics on blindness in infants were (and continue to be) lacking in the United States. The estimate, “10,000 RLF blind,” is an educated guess based on a review of the available, but incomplete, reports from the U.S. and abroad. The details of the Cooperative Study were recorded in the preliminary report (Kinsey 1955) and final report (Kinsey 1956a). It was estimated that about 750 infants would be available for study in the cooperating hospitals in one year. This fixed sample size determined the calculations of the practical limits of the trial. The two outcomes-of-interest, RLF and survival, presented different time problems: the presence or absence of RLF could not be established until each infant was followed for at least three months, but the observation period for survival was arbitrarily set at 40 days. In order to minimize the number of infants exposed to each of the two risks, a complex strategy was devised. Two rates of RLF-incidence were postulated:

“Routine (unrestricted) oxygen”10-20%
“Curtailed oxygen”2%

At these rates, a calculated minimum of 50 infants would need to be enrolled in the “routine (unrestricted) oxygen” group in order to obtain results which could be distinguished from chance fluctuation (confidence level of 0.01 percent). It was also assumed that 18 patients would be lost because of death or lack of follow-up. Thus, the number 68 (50 plus 18) was determined for the size of the routine oxygen group. A. Bradford Hill (see p 133) suggested that one infant be assigned to the routine oxygen group for each two infants in the curtailed oxygen group during the first three months of study. This procedure was expected to permit an early-as-possible check of the pretrial postulates concerning the RLF incidence rates in the two groups. If the difference in rates was found to be less than expected, more infants would be assigned to routine oxygen throughout the remainder of the year. The design considerations with respect to survival were different. Here it was expected that week-by-week monitoring of mortality rates in the two groups would quickly disclose a gross disadvantage to infants allotted to “curtailed oxygen” and permit an adjustment of the ratios of infants assigned to the two groups. “Routine (unrestricted) oxygen” was defined as oxygen (concentration over 50 percent) for 28 days. Return to an air environment at or after 28 days by reducing the flow rate of oxygen by one-third over each of three successive days. This regimen was common in the care of small premature infants in 1953. In “curtailed oxygen,” supplemental oxygen was administered only if, in the pediatrician’s opinion, the clinical condition of the infant demanded it. Concentration not to exceed 50 percent. The concentration of oxygen in incubators was measured three times a day with a paramagnetic analyzer (Beckman Instruments, Inc.). Birthweight differences, within each set of three, were minimized by allotment within three Birthweight categories: <= 1.0 kg (<21b 4 oz), 1.0-1.25 kg (>21b 5 oz – 21b 12 oz), and >1.25 – 1.5 kg (>21b 12 oz – 3 lb 5 oz). The New York City Health Department memorandum was entitled, “Recommendations to Hospitals with Maternity and Newborn Services.” It was signed by Jean Pakter, M.D., Chief of the Maternity and Newborn Division, Department of Health, New York City. The relevant passage from the minutes of the Pediatric Advisory Committee meeting of March 18, 1954 read as follows:

The meeting opened at 4 P.M. with a brief presentation by Dr. Pakter on the problem of oxygen concentration in relation to retrolental fibroplasia in premature infants. After some discussion by the Committee a motion was made and passed that a recommendation be sent to all hospitals with maternity and newborn services that oxygen should be administered to premature infants only as necessary and then only in concentrations under 40% unless respiratory embarrassment is present. An explanatory statement as to why this recommendation is being made is to be incorporated in the recommendation.

There was no explanation for the fact that the caveat “unless respiratory embarrassment is present”-which appeared in the minutes, was not repeated in the circulated recommendation. For the Bellevue results see Lanman (1954). The mortality data in the Bellevue study were recently reexamined (Day 1979):

LivedDiedTotalDeaths (%)
High Oxygen3694520
Low Oxygen28124030

Day and his collaborators noted that the 50 percent increase in mortality observed in low oxygen would be expected to occur by chance in about 1 out of 5 repeated trials of this size (P=0.21). Another way to examine the result of the 1954 Bellevue study is to ask the question: What were the chances of the study being able to declare a 50-percent increase in death rate as significant? The reviewers computed that only 1 out of 8 trials involving 45 babies in one group and 40 in the contrasting group would be expected to detect that a true increase in death rate of 50 percent had taken place. To increase the likelihood of detecting such an important discrepancy in death rate to 8 chances out of 10, it would be necessary to include 313 patients in each of the study groups-a total study enrollment of 626 babies! For the Philadelphia trial see Bedrossian (1954); the Colorado General Hospital experience was summarized by Gordon (1954a). The initial period of the Cooperative Study was focused on the survival issue. The experience in this three-month period was not disclosed until the trial was completed. The results appeared to confirm the favorable impression gained by week-to-week monitoring of short-term outcome in the two oxygen treatment groups:

Routine (unrestricted) oxygen (July-Dec.): 15/68 = 22% succumbed
Curtailed oxygen (July-Dec.):36/144 = 25% succumbed

All infants were then assigned to curtailed oxygen for the remaining nine months of the study. The results for the final period continued to be reassuring:

Curtailed oxygen (Oct.-June):115/574 = 20% succumbed

A disturbing rumor about this carefully considered design has been circulated in the years since the trial. Unfortunately, the myth was recorded in 1976 (James 1976):

Early in the course of the Cooperative Study, it was learned that RLF could be produced experimentally in newborn animals by administering oxygen … It was immediately decided that no additional infants should be placed in the routine or “control” group (unrestricted oxygen) and all infants enrolled after the first three months of study received “curtailed oxygen” therapy.

This statement is untrue, but it has fueled the fires of controversy about the study. The results presented in Table 6-1 appeared in the final report (Kinsey 1956a). The unexpected associations were found by “dredging” the data collected in the Cooperative Study (post eventum). Since no further controlled clinical trials testing oxygen treatment of human infants have been conducted, the associations have not undergone a proper challenge; they must be viewed with these reservations. In kittens (unlike the results in the Cooperative Study) Ashton’s group found a relationship between concentration of oxygen and vascular changes. The severity of the vaso-obliterative effect of oxygen rose with increasing concentration of the gas; concentrations below 35 percent, even after 21 days of continuous exposure, had no effect on the retinal vessels of newborn kittens. The relationship of these observations (vascular changes in a small number of experimental animals) to contrary evidence concerning cicatricial RLF in observations made on a large number of premature infants (p. 000) is unknown. Controlled studies in mice demonstrated that gradual withdrawal does not prevent vascular changes in that species. On the contrary, both the severity and frequency increased in mice when there was additional oxygen exposure during slow reduction of supplemental oxygen (Gyllensten 1956 and Patz 1957b). The rise and fall of RLF in New York State was reported by Yankauer (1956a).

* From enrollment at age 48 hours to age 40 days

Chapter 7 — Oxygen Treatment Practices in Premature Infants

Early recommendations were given by Budin (1900), Ylppo (1917), Hess (1922) and Bakwin (1923). Figure 7-1 is taken from Hess (1934a); reproduced with permission of the publisher, copyright The University of Chicago Press, 1934. The experience summarized in Table 7-1 appeared in Hess (1934b). See Wilson (1942) for the observations made in Detroit (depicted in Fig. 7-2). Nineteen years earlier Bakwin (1923) had quoted authorities who considered periodic breathing to be due to oxygen lack; Cheyne-Stokes breathing in adults was abolished by giving oxygen. Bakwin observed that when oxygen was administered to cyanosed premature infants, color improved and respirations became more regular. See Smith (1942) for the concept of “subcyanotic anoxia.” In the mid-1930s, Chapple conceived the idea of building an incubator that would (1) protect premature infants from exposure to air-borne micro-organisms, and (2) maintain stable surrounding conditions of temperature, humidity and oxygen. He concluded that “the infant would have to be kept in an incubator which did not need to be opened to care for them.” Access to the infant was accomplished through balloon cloth sleeves. The cabinet was ventilated by fresh air “drawn from out of doors” (filtered, heated and humidifed before it was blown into the closed infant compartment). Oxygen was introduced through a tube connected to a mask or funnel inside of the box. The first experimental model was installed in November 1937 at the Children’s Hospital of Philadelphia (see Chapple 1938). Approximately 35 incubators of this design were used exploratively, in Abington, Baltimore, Boston, Germantown, New Haven and New York City (Brooklyn) in the years before World War II. A commercially marketed version of the individually ventilated Chapple-type incubator became available in the late 1940s. AirShields, Inc. made the drawings of the incubator (Fig. 7-3) and the air-oxygen intake (Fig. 7-4). See Yankauer (1955) for the New York State memorandum. See Lanman (1955) for the 40-percent oxygen quote, Guy (1956) for the paper indicating the possibility of eliminating RLF. See Kinsey (1956b) for the letter cautioning against the emphasis on 40 percent oxygen, Gordon (1957) for the letter pointing out the defect in the Cooperative Study. The decision to limit enrollees in the Cooperative Study to those infants who survived 48-hours was reached with full appreciation of the fact that most of the deaths related to premature birth occur in the first two days of life. These early deaths, it was argued, would not help answer the question of oxygen damage to the developing retinal vessels. This strategy also took into account the fact that no restriction on oxygen administration in the first two critical days would be more acceptable to many nurses who were resistant to a proposed policy of oxygen restriction. Gordon did not indicate in his letter the magnitude of the risk of dying in the first 48 hours of life in 1953-54 among infants weighing <= 1.5 kg at birth. This can be appreciated from the following statement, which appeared in the final report of the Cooperative Study (Kinsey 1956a):

786 survived 48 hours* and were enrolled in the study . . . there were 634 additional premature infants in this birthweight category born in or brought to these hospital nurseries during this [one-year] period. All of these were reported to have died before 48 hours.

*Among the enrollees, 166 died before age 40 days.

Chapter 8 — Consequences of Oxygen Restriction

The totals in Figure 8-1 are based on articles listed in Quarterly Cumulative Index 1942-56, Current List of Medical Literature 1957-59, and Cumulative Index Medicus 1960-62. See Editorial (1974) for a summary of RLF unrelated to supplemental oxygen. The rare examples of congenital RLF observed since Reese and Blodi’s original description in 1951 were, and remain, a puzzle; they are often regarded as variants of oxygen-induced RLF and difficult to classify. On the other hand, Ashton’s demonstration (1965) that vasoproliferative changes occur irrespective of the cause of obliteration of the developing retinal vessels makes it fairly easy to accept these congenital instances as “true” RLF. The infant described in 1951 had typical microscopic changes of RLF. Birthweight was 1.95 kg, and he was thought to be full-term. The baby died soon after birth, and, in addition to RLF, had anencephaly. A later example of congenital RFL was associated with hydrocephalus. See Avery (1960) for the Johns Hopkins evidence (summarized in Table 8-1) and the concluding quote. In 1959 (see Silverman 1961), I reviewed the annual occurrence of hyaline membrane disease at Babies Hospital from 1950 through 1957 (the years before and after oxygen restriction). Crude annual mortality rate and proportion of dead infants examined at autopsy did not vary appreciably from year to year, and annual fluctuations in “incidence” of hyaline membrane disease exhibited no significant trend-change. However, these uncontrolled retrospective observations suffered from the same interpretive limitations as those at Johns Hopkins Hospital. See McDonald (1962) for the English follow-up results, the data shown in Fig. 8-2 and the quote. The Clark polarographic oxygen measuring device was first described in 1956 (Clark 1956). When this instrument became generally available in the early 1960s it was possible to measure oxygen tension of the blood quickly and reliably. To demonstrate that it was oxygen in blood (not in the air surrounding the eye) Ashton (1964) performed these experiments: pure oxygen passed through a cup over one eye of a newborn kitten for three days produced no retinal vessel changes; pure oxygen beathed by kittens for the same period resulted in typical retinal abnormalities. The low tensions of oxygen in arterial blood drawn from the lower aorta of infants with respiratory distress was largely due to shunting of blood from the right to left sides of the heart through fetal channels (see Strang 1961). This blood is unoxygenated because it is diverted to the aorta before there is any contact with the lungs. See Warley (1962) for the Oxford studies. See Silverman (1968) for a report of the 1967 meeting concerning oxygen and RLF.

Chapter 9 — The Determinative Era of Oxygen Treatment

The pure oxygen breathing test (Table 9-1) is described in Roberton (1968). The National Society’s hospital survey was conducted by Mrs. E.M. Hatfield (see Silverman 1969). The distribution of grades of RLF found in the survey (1,117 hospitals which delivered 1000 or more infants each year) was as follows:

Vascular RLF18
Cicatricial RLF14
Stage not reported1

From past experience (Chapter 3) it was expected that in 3 out of 4 infants who develop RLF the blood vessel changes would not go on to scarring (cicatricial stage). Since the survey indicated that about half ended in the vascular stage, I interpreted this to indicate that many infants with early RLF were not examined. In 1976 (Akeson 1976), the Variety Blind Babies Foundation and the Educational Services for Preschool Blind Children were (between them) serving 63 preschool children throughout the State of California. The birth dates of these children were distributed as follows:


The 1969 Cooperative Study involved Columbia, Johns Hopkins, McGill, Vanderbilt and Washington (Seattle) Universities. The study was co-directed by Doctors Arnall Patz and V. Everett Kinsey. The results of this study were published in 1977 (Kinsey 1977). Doctor John T. Flynn kindly supplied the photographs in Fig. 9-1. Flynn’s group (Kushner 1977 and Flynn 1977) has postulated that precursors of the normally developing capillary network of the retina (chords and sheets of mesenchymal tissue which grow out from the optic disc just ahead of vessels) are involved in development of the distinctive “silver line” structure in early RLF. According to this hypothesis, hyperoxia obliterates newly formed capillaries, particularly in the region just behind the advancing border of mesenchyme. The mesenchyme ceases to migrate and piles up as a shelf of tissue. Arteries and veins empty into the structure (by a few remaining bridges of blood vessels) to begin formation of a vascular shunt. In most instances regression occurs when buds of new vessels arise from the advancing edge of the shunt; vascularization proceeds leading, eventually, to near-normal retinal blood supply and a normally functioning eye. Persistence and progression of the abnormal changes, in a minority of instances (Flynn estimates ca. 15 percent), leads to the familiar damaging lesions of cicatricial RLF. See also Baum (1971b), Mushin (1974) and Kingham (1977) for present-day observations of the early changes of RLF by photography, fluorescein angiography and indirect ophthalmoscopy See Baum (1971a) for the residual changes of tortuosity which resemble the findings in the proliferative phase of RLF (esp. hairpin bends in the retinal arteries as in Fig. 9-2, which was kindly supplied by Doctor Baum). Other associations suggested that these minor changes were those of aborted RLF:

a. The temporal branches of retinal arteries were more often involved than the nasal segments (a well-known finding in acute vascular RLF).

b. Arterial tortuosity was associated with low gestational age, and was significantly less evident in those persons known to have had a more advanced gestational age than expected from their birth weight. It is reasonable to presume that the retinal vessels were more mature at birth in this group of individuals than in others of similar birth weight with appropriate gestational age.

c. In another set of observations the frequently occurring vascular changes seen by photography and fluorescein angiography were more common in multiple births (another well-known RLF association).

No definite pattern of refractive error was found in the eyes of the 35 persons with only arterial tortuosity of retinal vessels. Scarring (incomplete) RLF in 10 people was associated with myopia; curvature of the cornea was normal in these eyes. The invention of a device for continuous monitoring of the transcutaneous oxygen by means of a skin electrode (see R. Huch 1973 and A. Huch 1979) has had the effect of perpetuating the preoccupation with the role of oxygen in RLF. In addition to technical questions concerning this approach (e.g. which, if any, skin sites provide the best estimate of the state of oxygenation of the retina?) there remains the unanswered problem of RLF which occurs at “normal” arterial tensions. See Johnson (1974) and Phelps (1977) for vitamin E studies. See Aranda (1975) for observations concerning blood transfusions and RLF. See Cross (1973) for the calculations concerning the cost of preventing RLF and Bolton (1974) for the data shown in Fig. 9-4. Deaths on the day of birth were considered to be related to the policy of uncritical oxygen restriction. The hypothesis predicted that the deviation from a continuous fall in mortality would be trivial in babies born at term (normal birthweight) in whom respiratory troubles are relatively infrequent. This was borne out in the experience in New York State examined by Bolton (1974). There were numerous confounding influences during the years of rising day-of-birth deaths (e.g. influenza epidemics in the 1950s, large number of drugrelated deaths caused by sulfisoxazole and chloramphenicol-see Chapter 10—during the same period) and many major changes in neonatal care during the time of, and subsequent to, the fall in day-of-birth deaths. The conclusion of the 1953-54 Cooperative Study concerning the non-relationship between oxygen concentration and RLF (p 41) was based on after-the-fact analyses. A formal strategy in which infants were allocated in random order to prescribed exposure to specific concentrations of oxygen below 50 percent was not incorporated into the 1953-54 investigation. Recently, the St. Louis Society for the Blind honored Doctor Szewczyk for his work on the RLF problem. He was quoted (Szewczyk 1977) as follows:

. . . since the policy of gradual withdrawal from high oxygen concentration was instituted in 1951 only three out of more than 4800 babies born at the hospital have suffered blindness from RLF, although others have suffered a lesser degree of vision impairment.

See Hatfield (1975) for the prevalence data depicted in Fig. 9-5 (the existing number of affected children per 100,000 pupils). A.G. Jenkins, Director of the Orientation Center for the Blind, Albany, California, told me of the performance of the RLF-blinded man. Doctor Berthold Lowenfeld, former Director of the California School for the Blind, recently recalled his impressions concerning the behavior of RLF-blinded children when they arrived at his school in the 1950s. His experience in teaching blind children goes back to Vienna at the turn of the century when syphilis, gonorrhea and severe bacterial infections accounted for most of the blindness in young children. He said

. . then, early in the century, we knew children were damaged; those children showed mental retardation, inability to get along with others, quite out of the way, but here we had RLF children who showed behavior with which we were not familiar … and, of course, the name “autism” was tagged on these youngsters, and a good deal of observation was done; but no conclusive evidence has been submitted …

Elonen (1964) noted that the irregular pattern of development in blind children frequently leads to a mistaken diagnosis of mental retardation. She listed a number of “pseudo-conditions” in blind children and the circumstances which contribute to these manifestations:

Pseudo-conditionContributing Circumstances
a. Chewing difficultiesProlonged bottle feeding, late introduction of solids
b. Passivity/dependenceWaited on (carried, fed, dressed, guided…)
c. Spatial disorientationRestricted exploration
d. Speech – languageBombarded with sound, over-response to nonverbal signals (family), absent facial-gesture signals
e. ConceptualizationLack of visual challenges (scenes) and clues (depth, size, color)
f. Loneliness/lone-ismLimited experience with other children
g. Dearth of pleasant fantasyIncreased experience with hospitals, surgery, falls, accidents and separations
h. Auto-erotic habitsSeductive behavior of adults

See also Chase (1972) and Fraiberg (1977) for descriptions of the behavior of RLF-blinded children. At the time when plans for the 1953-54 Cooperative Study were discussed, the need for long-term followup of participants was brought up. Unfortunately, the opportunity for systematic observations, as I pointed out earlier, was missed. In 1976, federal support was requested to examine a cohort of RLF-blind young adults and compare their neurologic status (esp. spatial orientation) with controls (other congenitally blind and a group of sighted ex-prematures). The proposal was rejected by reviewers with the statement, “. . . much of the information sought by the study is mainly of historical value.” Patz (1955) reported that extensive examination of all organs in newborn mice, rats and kittens exposed to prolonged high concentrations of oxygen failed to reveal changes in any tissues except the eye and occasionally the lung. On the other hand, Gyllensten (1959) found retarded development of capillaries in the brain of young mice reared in high-oxygen environments. In rats, residual brain damage has been seen only after exposure to very high oxygen tension (4-5 atmospheres). See also Balentine (1966) for oxygen effects on the brain of rats. See Feeney (1976) and Dormandy (1978) for a discussion of the biochemical mechanism of oxygen toxicity.

Chapter 10 — Medical Inflation

A somewhat modified version of this chapter appears in Perspectives in Biology and Medicine, Summer 1980. See Wallace (1950) and Silverman (1961) for demographic information on the incidence of prematurity. Yankauer (1956b) described the New York-Cornell institutes. Doctor Chapple recalled that Chestnut Hill Hospital near Philadelphia was among the first institutions to express an interest in his new incubator (after the initial experimental period, see p 189). He delivered a hand-built model to the hospital in a borrowed pick-up truck in the summer of 1938. A 2 lb 10 oz infant was reared in the device without any swaddling clothes. He and Doctor Aims C. McGuiness (the infant’s pediatrician) spent every spare moment staring at the baby-the first naked premature infant anyone had observed for an extended period of time. Toward the end of the first year of use of the new incubator, Doctor Chapple became aware that the nurses were resisting the practice of caring for nude infants. They seemed ill at ease when a naked boy suddenly sprayed conspicuously. A compromise was struck to preserve modesty: diapers! See Silverman (1961) for the development of premature centers and Taylor (1948) for a report of the Colorado program. See Pakter (1954) for the history of the New York City program for premature infant centers. Some indication of the immediate (hospital) cost for premature infant care in the 1970s is given in a report by Pomerance (1978): the average daily cost for infants weighing less than 1 kg at birth was $534/day/infant in September 1976 (by November 1977 this rate had increased by 31 percent). Murphy (1951) described North Carolina developments. See Bloxom (1950) and Reichelderfer (1956) for the “air-lock” experience. See Ravenel (1953), Silverman (1955) and Briggs (1955) for the detergent-mist details; and Silverman (1956a) for further observations on water-mist. See Smith (1949 and. 1957) for observations of effects of delaying feedings. See Ylppo (1954-55) for the European arguments and Gleiss (1955) for the German evidence against the fasting practice. See Drillien (1961) and D.P. Davies (1978) for the observations on brain damage (esp. spastic diplegia) in relation to first-feed practices. For the New York studies on fat absorption and its implications, see Gordon (1941). See Menkes (1966) for the report of a follow-up study of transient tyrosinemia in the newborn period. The taurine hypothesis was advanced by Gaull (1977). See Clifford (1950) for the prophylaxis, and Silverman (1956b) for the sulfisoxazole experience. See Alexander (1957) and Burns (1959) for the chloramphenicol observations. See Katz (1972) and Insight Team (1979) for the details concerning thalidomide. The “therapeutic programs which evolve . . .” quote appeared in a letter written to me by Doctor William H. Tooley on May 9, 1972. The “Proclaimed Therapies” noted in Table’ 10-2 were widely applied long before formal evaluation was carried out. In most instances, the practices faded away (became “unfashionable”) without a formal test. See Agerty (1952) for the results of a trial of testosterone treatment which failed to show any significant influence on weight gain. Years after the attempts to stimulate the growth of premature infants with this powerful hormone were abandoned, some disturbing animal studies came to light; early treatment with male sex hormone had an effect on sexual-orientation behavior at maturity. Follow-up studies of the testosterone-treated babies were never undertaken. A double-blind controlled study of the effects of thyroid hormone (see Stevenson 1953) indicated that this treatment was without recognizable benefit and was sometimes harmful'(diarrhea and increased pulse rate were observed in treated infants). The DES experience is described on p 158. Masculinization of the female fetus may follow administration of progestins (especially when given prior to the 13th week of pregnancy-see Grumbach 1959). See Allen (1957) for an account of the use of exchange transfusion in the management of erythroblastosis fetalis. The exact criteria for use of this technique to treat jaundice caused by other mechanisms has been debated for years. The use of oxygen for periodic breathing is discussed on p 46. See note above for the issues in initial thirsting and starving. Meyer (1956) described jaundice and kernicterus caused by synthetic vitamin K. See note above for the low-fat-high-protein feeding experience, the sulfisoxazole episode and the chloramphenicol incident. The stomach of newborn infants was suctioned at birth in the hope that this would prevent the development of respiratory difficulty. Several trials were conducted (see Westin 1958); none were able to demonstrate a beneficial effect. Traction on the sternum to stabilize the soft chest wall of infants with respiratory difficulty was proposed, used half-heartedly, but never evaluated formally (see Townsend 1956). For the epsom-salts see note below. A rocking device (see Millen 1955) to prevent respiratory problems was used for several years; there was no convincing evidence that the apparatus was effective. The Alevaire and water-mist reports are noted above. See Chu (1967) for the experience indicating that administration of acetylcholine may be beneficial to infants with respiratory distress syndrome. The authors advised that their uncontrolled study results should not be accepted until a convincing test was carried out using proper experimental design (and they warned that the swiftly acting, dangerous drug should be administered with great caution). Nonetheless, the agent was used by others, but the purported beneficial effects on survival were never put to a critical test by the original observers (nor by anyone else). See Donald (1954) for an early description of the use of a respirator to support premature infants. Belenky (1978) discussed the weakness of the evidence to support the widely used and accepted measure of constant positive airway pressure. See Vengusamy (1969) for results of a controlled clinical trial which demonstrated a higher mortality with gastrostomy than with routine oral feedings in groups of very small infants (birthweights 0.75- 1.25 kg = 1 lb 10 oz – 21b 12 oz). See Miller (1957) for the animal experiments on the use of cold in resuscitation. Ice-water resuscitation was used in delivery rooms for years; it was slowly abandoned after it was shown that the newborn infant makes a metabolic response to cold exposure. Sodium bicarbonate injections were used extensively to treat early acidemia in the hope that this would prevent hyaline membrane disease; Hobel (1972) was unable to confirm these hopes from the results of a well-planned controlled trial involving 90 babies. See Finberg (1967) for the evidence which suggests that rapid injections of alkali are dangerous. See p 105 for the relation of temperature to survival. See Powell (1973) for the description of the dangers of hexachlorophene bathing. See Speck (1975) for the possibility of damage to chromosomal material from exposure to phototherapy. The list of treatments in Table 10-2 is not complete; see Moore (1976) for futher misadventures which followed those presented in this table. The “I am disturbed at . . .” quote appeared in a discussion of the paper by Clifford (1950). See Coleman (1957) for the sociologic study. Soon after the New York Times article announced the effect of epsom-salt enemas, a query was received by the editor of Pediatrics. The writer (see Van Gelder 1965) expressed concern about the safety of dehydrating enemas for desperately ill infants and he was disturbed by the fact that physicians were under considerable pressure to use the treatment as the result of the wide publicity. He requested guidance from the journal (the official organ of to American Academy of Pediatrics). The plea was sent to the originator of the new approach; when a reply was received both letters were printed in the February, 1965 issue. The innovator repeated his rationale (which was explained in full in a later report-see Stowens 1965) and his letter of reply ended with these words, “I am happy to be able to report that since the appearance of the newspaper and magazine stories, numerous physicians around the country, who have not shared Doctor Van Gelder’s querulousness or inability to reach independent conclusions, have reported to me of their successes with this form of therapy.” See Andrews (1965) for the effects of magnesium sulfate in newborn lambs, and Outerbridge (1973) for the report of a death following an epsom-salt enema treatment. The incredible Laetrile situation is described in Sounding Board (1978). Indeed, the emotion surrounding Laetrile threatens the legislative safeguards erected after the thalidomide incident (see p 83). See Naftulin (1973) for the “Doctor Fox” phenomenon.

Chapter 11 — The Price of Progress

The data in Table 11-1 were calculated from information provided by Richard W. Turlington, Information Officer, Division of Research Grants, N.I.H., and from U.S. Bureau of the Census (1975). Table 11-2 was prepared from listings in Quarterly Cumulative Index Medicus 1950 and 1955 and Index Medicus 1960. See Price (1961 and 1963) for analysis of the growth of science. The Russian scientist’s remark is apocryphal. See Pakter (1974) for the birth rate/infant mortality experience in New York City, and Lee (1976) for the Bronx Municipal Center analysis. See Morris (1975) for evaluation of causes of decline in infant mortality in the U.S. Intermediate infant mortality was defined as a rate equal to the observed overall infant mortality plus or minus 10 percent, using data from a 1960 U.S. Live Birth Cohort Study to define the high-risk and low-risk maternal-age/total- birth- order cells shown in Fig. 11-1. The 1960s infant mortality rates for each age/total-birth-order cell were applied to the age/ total-birth-order distributions of births for each of the succeeding years. This calculation provided the estimated number of deaths in each cell for these years under an assumption of constant age/total-birth-order mortality rates within each cell. The sum of the estimated deaths for each year was then compared with the actual number of deaths to measure the contribution of changing age-parity distributions. See Hinds (1974) for the relation between infant mortality and numbers of health-care workers, and St. Leger (1978) for the correlation with “doctors available” (Fig. 11-2 is redrawn from this report). See Neligan (1974) for the Newcastle study. The U.S. studies were reported by Broman (1975), Werner (1967) and Jordan (1970). See Drillien (1967) for the Edinburgh results. The University College Hospital results and subsequent English reports are summarized by P.A. Davies (1976). See McCormick (1977) for the eye survey results. See Sameroff (1975) for the “continuum of caretaking casualty” thesis, Werner (1971) for outcomes at ten years of age in 1000 live births, and Baum (1977) for his comments on families. Also see Stratton (in Chard 1977) for a review of the evidence concerning perinatal circumstances and later development. See Jonsen (1975) for Doctor Nader’s comments.

Chapter 12 — Progress in a Groove

See Chard (1977) for criticism of perinatal medicine. See Whitehead (1925) for his quote. See Silverman (1958) for the temperature results. See MacMahon (1977) for U.S. cause-of-death data and life-expectancy trends; Wegman (1977) for recent infant mortality. See Harris (1977) for a discussion of female infanticide, and Birdsell quoted by Harris. Figures 12-1 and 12-2 are redrawn from Shapiro (1968). See Grove (1968) and National Center for Health Statistics (1977) for the data in Table 12-1. McKeown (1976) examined the causes of modern population rise. See Erbe (1977) for genetic defects statistics (including cystic fibrosis). Chargaff (1973) discusses the revulsion from science. I. Chalmers (1976) described the Oxford criticism; in this regard, see also Chard (1977). The news report in the Bangladesh Times was written by Zagrulla Chowdry See Annotation (1974) for recommendations of the West African workshop. Mumford (1946) discussed a program for survival. See Editorial (1977) for comments about malpractice trials. See Lowenfeld (1975) for an account of the independence movement among the blind. See Commoner (1973) for comments on the social responsibility of science. See Hunter (1978) for the North Carolina study of social chaos in families of ex-premature infants. See Lehrer (1965) for his song, That Was the Year That Was. See Brody (1973) for a discussion of the distinction between technical problems and value problems. Freidson’s book (1972) is an excellent sociologic analysis of the profession of medicine. See WHO report (1976) for the role of physicians in ethical problems related to birth defects. Lawless (1974) commented on “Medical Practice.” The “Values Underlying . . .” conference was reported by Steinfels (1978). See Firth (1970) for the effects of evangelism in Tikopia. Webster (1976) described the Puritan outlook on science and medicine. See Mumford (1970) for brilliant comments on the growth of technology. See Watch Tower (undated) and Frankel (1977) for the attitudes of Jehovah’s Witnesses concerning blood transfusion. See Brody (1973) as in the abovecited quote. G.B. Shaw’s maxim appears in the epilogue of Man and Superman. See von Fritz (1952) for a discussion of relative and absolute values. See Kluckhohn (1951) for the definition and comments about morals in an excellent essay which gives an anthropologist’s analysis of the concept of values. See Beecher (1966) for the classification of clinical studies. The Anti-Vivisection Society’s brochure was sent to me some time in the 1960s. See Schulman (1967) for the presidential address to the Society for Pediatric Research. See Gregory (1971) for the continuous positive airway pressure report, and Surgeon-General (1966) for the Public Health Service memorandum concerning signed informed consent. The “While controlled trials . . .” quote is from Stern (1975). See Bernard (1865) for the “Many physicians attack . . .” quote.

Chapter 13 — The Experimental Method in Clinical Studies of Children

The material in this chapter will appear, in different form, in my book on human experimentation in the Scientific American Illustrated Library series. See Smithells (1975) for the “I need permission . . .” quote. See Singer (1976) for comments about the orientation of bioethecists. Cahn (1972) described the “Pompey Syndrome”; the passage is reproduced with permission of the original publisher, copyright The Johns Hopkins University Press 1964. See Greenwood (1942-43) and Graunt (1662) for the origins of medical statistics. Bronowski (1977) discussed the development of the experimental method. Claude Bernard’s quotes are from his book (1865). See Freidson (1972) for his comments. See Bronowski (1959) for the quote concerning the statistical method. In this regard, no less a personage than Albert Einstein resisted the use of the statistical approach to explore natural (subatomic) events, with his famous remark, “God does not play dice with the world.” The quote attributed to Galen is found in Strauss (1968). The South Pacific story is apocryphal. See Venn (1866) for the most lucid discussion of the logic of chance extant (!); the description of the behavior of observations in a series is taken from his classic book. Gauss’ Law expresses the number of errors (variations) of each size that are expected to occur as the result of chance in many measurable characteristics appearing in nature and in the man-made world. For example, in target shooting, the hits group themselves in a pattern: close to the bullseye they are densely packed, farther from it there are fewer and fewer hits, until there are none at all. The frequency distribution, when plotted out according to the distances that separate the hits from the bullseye, has the physical appearance of the familiar bell-shaped figure: the Gaussian normal curve. The mathematical equation of this curve (which includes the two parameters of the equation: the central value of the distribution expressed as the arithmetic mean and the dispersion of values around the mean expressed as the standard deviation) provides the basis for the arithmetic operations used in making statistical inferences about the variations with a similar distribution which occur in a series of observations. These variations occur in the field of measurements (e.g., stature, concentration of substances in the blood, cognitive tests). They also appear as variations in the proportions of dichotomous characteristics (qualities which divide the “population” into two mutually exclusive groups) in a series seen in medical problems (e.g., “alive” or “dead”, “RLF” or “no-RLF” . . .) and in gambling problems (e.g., “heads” or “tails” in coin-tossing). See Levinson (1939) for a most readable account of gambling and the laws of chance. The probability that two successive events both take place is equal to the probability of the first event multiplied by the second (the latter being computed on the assumption that the first event has already taken place). The probability of “heads” on a single toss of a coin is 1/2 and the chance of throwing two successive “heads” in two tosses is (1/2)^2=1/4. The general rule for calculating the chance of throwing any number of n successive “heads” in n tosses is (1/2)”. And the chance of throwing a number of “heads” in a succession during a long series of tosses is calculated from the formula (1/2)” +- For two successive “heads” the computation is (1/2)^(2+2) =0.0625; thus, in a set of 100 tosses of the coin about 6 runs of two-heads-in-a-row are expected. See Fisher (1925 and 1926) for a description of his field-experiment designs. See Hill (1937 and 1953) for his early writings; a description of the classic streptomycin trial is reprinted in Hill (1962). The MRC streptomycin trial is usually cited as the first of its kind because it kindled considerable interest and led directly to further studies on the same model. It is unfortunate that previous well-planned controlled clinical trials (see Amberson 1931 and Patulin Clinical Trials Committee 1944) did not stimulate widespread interest in the power of this study format. I believe this was related to the fact that the early trials were negative: they failed to demonstrate the effectiveness of proposed treatments (sodium -gold-thiosulphate treatment of tuberculosis in 1931, and a metabolic product of Penicillin patulin, called “patulin,” for the treatment of the common cold in 1944). See Mainland (1952), Feinstein (1977) and Sackett (1975) for discussions of methodologic issues in clinical experimentation. See Mainland (1960) for examples of misuse of statistical arithmetic. See Bronowski (1977) for the comment about Darwin. See Cochrane (1972) for the coronary-care-units story. Byar (1976) discussed the problems of adaptive designs. See T.C. Chalmers (1975) for comments about randomization. The task force findings on risks for human subjects was reported by Cardon (1976). The comments of the Boston group appeared in a letter to the editor (Boston Interhospital 1975). See Fried (1974) for a discussion of legal issues. See Kabat (1975) for his questions. See Selvin (1966) for “data-dredging.” See Kinsey (1977) for the results of the blood oxygen study. See Eisenberg (1977) concerning the social need to foster excellence in medical research. See Popper (1962) for his comment.

Chapter 14 — The Future for Studies Involving American Children

See Price (1961 and 1963) for his analysis of the growth of science; Fig. 14-1A is redrawn from his book (1961) and Fig. 14-1B is redrawn from Durack (1978). Price points out that logistic curves of growth have become well known in numerous analyses of time series, especially those concerning the growth of science and technology. See Mahler (1977) for comments on health care development. See Tuskegee (undated) for conclusions after investigating policies to protect research subjects. See Eisenberg (1977) for his comment. Meier (1972) summarized the experience of the 1954 polio vaccine trials. Comroe (1976) reported the review of the key research efforts leading to the ten most important advances. The “Lyndon Johnson doctrine” phrase is Eisenberg’s (1977). See Bronowski (1977) for a dicussion of society’s involvement with science, and Medawar (1969) for the history of development of the scientific method. Juhl (1977) reported the “epidemiology” of the randomized clinical trial (in gastroenterology), and Tannenbaum (1977) reported the survey in institutions conducting research in children. See Walster (1970) for the editorial policy proposal. See Barber (1973) for discussion of clinical research among the poor. See Jonas (1970) for a discussion of recruitment of subjects for experimentation. In connection with experimentation in the children of physicians, see Altman (1972). Claude Bernard’s maxim appears in his book (1865). See National Commission (1977) for recommendations concerning research involving children, and American Academy (1977) for guidelines developed by the American Academy of Pediatrics. Jonas (1970) has given considerable thought to the agonizing question of whom should be called upon to participate in clinical studies. See Bostock (1962) for the “exterior gestation” thesis, and Klaus (1976) for evidence concerning the dyadic relationship between mother and infant. Ramsay (1970), Bartholome (1977) and Hauerwas (1977) have written about the rights of the fetus and the newborn. See Eisenberg (1977) for his comments about informed consent. Guttentag (1953) discussed the problem of experimentation on human beings years before the present-day crisis. We began to implement his suggestions about 9 months before the National Institutes (1953) guidelines were circulated, see Silverman (1966). T.C. Chalmers (1974) reviewed the DES incident and other examples of the alienation of practicing physicians from research results. Two marketing surveys provided the estimates of DES usage in the 1960-70 period, see Heinonen (1973). Mahler (1977) recommended community trials. Popper’s book, The Logic of Scientific Discovery, was published in German in 1934; it was not translated into English for 25 years (see Popper 1959 and 1962). See Bronowski (1977) for a critique of Popper’s test-by-refutation proposal and Perkinson (1978) for an excellent summary of the Popperian approach. In connection with failures to confirm an hypothesis, the use of the term “no significant difference” introduces confusion, unless otherwise qualified. A. Bradford Hill once advised:

… it is better (particularly in a matter of importance) to take “statistically not significant” as the “non-proven” of Scot’s Law rather than as the “not-guilty” of English Law.

The phrase “no significant difference” may denote (1) a difference which is no greater than would be expected reasonably often in repeated sets of observations of a given size solely as the result of random variation (for example, the true difference in survival between two methods of care is zero), or (2) a difference which is not important to the community [for example, the true difference in survival is not zero, but it is so small that (a), it does not justify investment of the available resources of the community, or (b), for practical purposes the two methods of care may be considered equal alternatives]. If the number of observations is few the true difference may be quite large and the decision “no significant difference” is reached because of an inappropriately small sample size. In clinical trials, estimates of sample size “needed” are determined on the basis of more or less arbitrary answers to questions posed before the trial begins: (1) What shall be considered an important difference, for example, in survival? (2) What risk of errors shall be assumed (a) to declare a “difference” when the true increase in survival is zero, and (b) to declare “no significant difference” when the true difference is as great or greater than the important amount? Concerning a critical test of oxygen curtailment in the first two days of life, it is interesting to turn back to an editorial written in November 1954, two months after the announcement of the results of the Cooperative Study (see Gordon 1954b). After reviewing the evidence concerning oxygen treatment of premature infants up to that time, Doctor Gordon noted that “. . . until carefully controlled studies of oxygen therapy . . . are available, empiric evidence . . . must be used to guide physicians in the treatment of these babies during the first 48 hours of life.” The “carefully controlled studies” were never carried out. Claude Bernard’s quote appears in his book (1865).


See Zacharias (1952) for the information in Appendix A. See Kinsey (1956a) for details in Appendices B and C. See Sackett (1975) for the methodologic standards in Appendix D.

Return to Contents Page