Field trials of tuberculosis vaccines: How could we have done them better?

Field Trials of Tuberculosis Vaccines: H o w Could We Have D o n e T h e m Better? George Wills Comstock, MD, Dr PH, FACE Department of Epidemiology, Johns Hopkins University, Baltimore, Mar~/land

ABSTRACT: Nineteen controlled trials of vaccination against tuberculosis are reviewed. Most

involved very large numbers of participants and represented a wide variety of geographic and socioeconomic conditions. The trials were conducted under field conditions that sometimes verged on the primitive. Length of follow-up for tuberculosis ranged from 3 to 23 years, and up to 28 years for cancer. Under these circumstances, compromises and mistakes were made along with notable successes. With the current interest in the immunology of tuberculosis and other chronic infectious diseases giving rise to renewed hope for more efficacious vaccines, lessons from the past can be useful in planning the long-term evaluations that will be needed as these hoped-for vaccines become available. KEY WORDS: Tuberculosis, clinical trials, vaccines, BCG vaccine Always learn from the mistakes of others--you don't have time to make them all yourself. (Fortune cookie, circa 1970) HISTORICAL INTRODUCTION The story of vaccination to prevent tuberculosis starts with Robert Koch. Already w o r l d famous for his contributions to bacteriology, and especially for his meticulous w o r k in demonstrating that Mycobacterium tuberculosis was the necessary cause of tuberculosis [1], his a n n o u n c e m e n t in 1890 of a "curative agent against tuberculosis" caused great excitement [2,3]. Unfortunately, it was soon f o u n d that "Tuberkulin," as Koch n a m e d it, had no therapeutic or prophylactic value [4]. Koch also tried to i m m u n i z e animals with killed tubercle bacilli but gave u p the attempt as u n p r o m i s i n g [5]. Since then, n u m e r o u s other investigators have used killed organisms with generally favorable results [6], although a large and carefully controlled experiment indicated that the protective value in guinea pigs was only m o d e s t [7]. A controlled trial using killed tubercle bacilli was conducted a m o n g mental hospital patients in Jamaica as far back as 1932 [8]. In keeping with the finding a m o n g guinea pigs, the killed vaccine was f o u n d to confer protection against tuberculosis in h u m a n s on the order of 50% [91.

Address reprint requests to: George Wills Comstock, MD, PhD, John Hopkins Training Center, P.O. Box 2067, Hagerstown, MD 21740-2067. Received July 28, 1993; revised January 21, 1994. Controlled ClinicalTrials 15:247-276(1994) © Elsevier ScienceInc. 1994 655 Avenueof the Americas,New York,New York10010

247 0197-2456/94/$7.00

248

G.W. Comstock A wide variety of naturally occurring and putatively avirulent mycobacteria h a v e also been tried as vaccines but with little success [10]. Only one, the vole bacillus, has been s h o w n to be effective in preventing tuberculosis in h u m a n s [11]. Attenuated tubercle bacilli w e r e also tried, but again only one, BCG, was ever seriously considered for h u m a n vaccination [12[. BCG, the currently official n a m e for Bacille Calmette-Gu6rin, owes its existence to serendipity, the impact of a chance observation on p r e p a r e d m i n d s [13,14[. The physician Leon Charles Albert Calmette and the veterinarian Jean Marie Camille Gu6rin had joined forces in 1897 at the Pasteur Institute in Lille, France. In 1900, they b e g a n their search for a tuberculosis vaccine, trying variations of Pasteur's m e t h o d s for p r o d u c i n g attenuated strains of m i c r o o r g a n i s m s [13]. In their experiments they infected calves by the oral route with a strain of virulent o r g a n i s m s obtained f r o m Nocard that he had isolated from a cow with tuberculous mastitis [15]. To minimize the tendency of tubercle bacilli to c l u m p in suspensions, they a d d e d bile to their culture m e d i u m as a natural detergent. Unexpectedly, the o r g a n i s m s lost their virulence for calves, and later for guinea pigs, rabbits, horses, and monkeys. In 1921, Calmette n a m e d it "Bacille bilie Calmette-Gu6rin," later d r o p p i n g the "bilie" and a d o p t i n g the abbreviation BCG. In 1924, he declared it a "virus fix6," i m p l y i n g that it was a n e w nonpathogenic strain of tubercle bacilli I10,14]. The first use of BCG in a h u m a n was described by Weill-HalI6 and Turpin in the following translation [16[. It was tried in July 1921 by one of us (Weill-Hal6) on a newborn child doomed to infection with tubercle bacilli because he would have to live with a contagious grandmother. On three occasions, on the 3rd, 5th, and 7th post-natal days, the infant was given the vaccine. No ill effects were observed, and the child, brought up in a bacilliferous milieu, has developed normally. In fact, he is in perfect health. For the next 25 years, BCG m e t with a decidedly mixed reception. M a n y tuberculosis workers accepted BCG vaccination, partly because of theoretical considerations and the lack of serious side effects in t h o u s a n d s of vaccinated infants but also because of C a l m e t t e ' s eminence and his forceful personality. The fervor of his belief in the vaccine is s h o w n by the death bed statement attributed to him b y one of his students. "My son, I a m dying. Take care of m y wife. Do not emigrate to America. Fight for BCG!" |17[. Others, concerned b y Calmette's refusal to u n d e r t a k e scientific studies a m o n g h u m a n s before its general use, agreed with G r e e n w o o d ' s scathing criticisms of Calmette's evidence that was based on uncontrolled observations [18]. Still others believed that BCG w a s dangerous, either because the induced tuberculin sensitivity m i g h t h a v e a h a r m f u l effect [19] or because the attenuated o r g a n i s m s could revert to their original virulent f o r m [20]. Fears of regained virulence a p p e a r e d to h a v e been substantiated b y a tragic incident k n o w n as the "Lubeck disaster" [21]. During the period D e c e m b e r 10, 1929 to April 30, 1930, 251 of 412 infants were given three oral doses of "vaccine" during the first 10 days of life. Of those vaccinated, 72 died of tuberculosis, m o s t within 2-5 months, a n d all but one within a year. Another 135 d e v e l o p e d clinical evidence of tuberculosis but survived, and 44 became

249

Tuberculosis Vaccine Trials Table 1

Tuberculosis Case and Death Rates A m o n g Student Nurses at Ullevdl Hospital, N o r w a y , by Tuberculin and BCG Vaccination on Entry, 1927-1946 Tuberculin and Vaccination Status on Entry

Tuberculin-positive, not vaccinated Tuberculin-negative, not vaccinated Tuberculin-negative, vaccinated Source:

Rate/1000 Person-years

No.

Personyears of Observation

Cases

Deaths

668 284 501

1772 687 1450

12.4 141.2 24.1

0 14.6 2.1

Data from Ref. 23. positive tuberculin reactors during a period of 23 days to 3 months but were well otherwise. There were no tuberculosis deaths among the 161 unvaccinated infants during the next 3 years. Fortunately for BCG, virulent tubercle bacilli of the Kiel strain were cultured from the victims. This strain had the unusual characteristic of producing a greenish fluorescence on Sauton's medium. Both the Kiel strain and BCG vaccine had been kept in the same incubator, a practice said not to be u n c o m m o n at that time [22]. Furthermore, it was reported that the labels often fell off the tubes. After careful review, it was concluded that the infants had been given the virulent Kiel organisms by mistake. Although BCG was exonerated, the damage to its reputation was longlasting. Some good did come of this sad affair, however. Since that time, it has been a strict and general rule that BCG vaccine for h u m a n use must never be kept in the same laboratory with other mycobacteria [10]. Interest in BCG vaccination did not become widespread until the end of World War II. The foundation for this interest, however, started much earlier. It was based to a large extent on the work of Heimbeck in N o r w a y (23). In 1924, he began routine tuberculin testing of all entrants to the nursing school at Ullev$1 Hospital. Of the 336 students admitted d u r i n g the next 3 years, 44.9% were positive reactors to the Pirquet tuberculin test. Only two of these y o u n g w o m e n developed tuberculosis during this time, a rate of 0.65 per 100 person-years of observation. In stark contrast, the rate among the 185 students who were tuberculin-negative on entry was 17.3 per 100 person-years. Because of this clear demonstration of the need to protect tuberculin-negative nursing students u n d e r conditions existing in some hospitals at that time, Heimbeck offered them BCG vaccination. At first, there was little effort to gain acceptance since he n e e d e d a nonvaccinated comparison group, but the offer must have become more persuasive with the passage of time judging from the decreasing proportion of refusals. By 1946, Heimbeck had observations on 1453 student nurses. As shown in Table 1, the vaccinated group fared far better than the tuberculin-negative students w h o were not vaccinated. Because of the similar living conditions during training for each of the three groups, Heimbeck's results gave much support for vaccination. Further support came from a so-called natural experiment in Denmark during World War II [24]. In a schoolroom with poor ventilation as a result of precautions against air raids, a tuberculous teacher exposed 305 students. An epidemic of tuberculosis ensued. Because the prior tuberculin and vaccination

250

G.W. Comstock Table 2

Tuberculosis Case Rates A m o n g Schoolchildren Exposed to a Tuberculous Teacher in Denmark, by Tuberculin and BCG Vaccination Status Prior to Exposure, 1943-1955 Postprimary Tuberculosis Cases

Tuberculin and Vaccination Status Prior to Exposure

Number Exposed

Number Infected

No.

% Exposed

Tuberculin-positive, not vaccinated Tuberculin-negative, not vaccinated Tuberculin-negative, vaccinated

105 94 106

__a 70 ?b

9 14 2

8.6 14.9 1.9

qnfected prior to this exposure. bIndeterminate because vaccination caused positive tuberculin reactions. Source: Data from Ref. 24. status of the students was known, attack rates could be calculated according to these characteristics. These rates are s h o w n in Table 2, where it can be seen that the risk was m u c h higher a m o n g unvaccinated tuberculin-negative students than a m o n g the other two groups, the previously vaccinated, and those w h o were positive reactors prior to exposure. Although this "experiment" was neither controlled nor blind, and living conditions were not uniform, it is still reasonable to conclude that BCG vaccination had resulted in considerable protection against tuberculosis. Following the ravages of World War II and the privations of the postwar period, the United Nations Rehabilitation Agency, the United Nations Childrens Emergency Fund, and the World Health Organization were greatly concerned about tuberculosis in m a n y nations. At that time, it was widely believed that the risk of tuberculosis was concentrated in a few years after becoming infected and that survivors were protected by partial i m m u n i t y [25]. Because both observational and animal studies had shown that vaccination could reduce this risk, and because vaccination was the only control measure that could be e m p l o y e d on a scale commensurate with the needs, the International Tuberculosis Campaign was initiated to conduct mass vaccination campaigns in m a n y parts of the world. During the period July 1, 1948 to July 1, 1951, a total of 29,677,380 persons were tuberculin-tested and 13,874,709 persons with negative tuberculin reactions were given BCG vaccine [261. Little of this enthusiasm for vaccination was felt in the United States. Here the tuberculosis control efforts centered a r o u n d case finding and treatment [27]. Chest p h o t o f l u o r o g r a p h y was e m p l o y e d on a mass basis [28]. Persons found to have contagious tuberculosis were usually admitted to sanatoria for rest and isolation. P u l m o n a r y cavities were treated with a variety of surgical procedures aimed at collapsing the diseased portions of their lungs in an effort to encourage healing. This approach to tuberculosis control was evaluated entirely b y uncontrolled observations. Injection of gold salts was the only therapeutic p r o c e d u r e that was properly evaluated. In what appears to have been the first controlled clinical trial, it was shown to cause more harm than good [29]. In contrast, efforts to place vaccination of h u m a n s on a scientific footing

Table 3

S u m m a r y of C o n t r o l l e d Trials of V a c c i n a t i o n A g a i n s t T u b e r c u l o s i s in Humans % Reduction

Ref. 1 2 3 4 5

Study Population 237 Philadelphia infants in tuberculous families (30) 821 eligible and 2439 other mental patients, Jamaica (9) 1092eligible infants in tuberculous families, New York (31) 609 newborn Qu'appelle Crees, Saskatchewan ( 3 2 ) 41,301 newborns, Algiers (33)

6

3008 eligible American Indian children, western USA ( 3 4 )

7

3381 newborns, Chicago (35)

8 9 10 11 12 13 14 15 16 17 18 19

Vaccine a BCG (O) (King) Killed

Starting Year

Years of Follow-up

Total TB Problem

1927

3-4

--

--

83

1932

1-10

17

35

56

1933

5

--

--

8

1933

15

81

81

80

1935

1-11

--

--

(11)c

1936

--

75

82 (36)

1937

9-11 cases 20 deaths 12-23

75

75

83

1938

6--8

59

59

100

1941

19

74

74

100

1947

20

-9

-57

--

1947

13

--

-40

-88

1949

18-20

5

29

38

1950

20

1

6

1950

16-21

1

20

1950

18-20

53

77

1953

3--6

7

100a

1965

3

39

--

1965

3

--

67

1968

15

--

-2

Eligiblesb Cases Deaths

M. tbc.

tiC) BCG (IC) (Calmette & Phipps) BCG tiC) (Montreal) BCG (O) (Algiers) BCG (IC) (Phipps)

BCG (MP) (Tice) 262 American Indian BCG tiC) newborns, N. Central US ( 3 6 ) (Phipps) 451 newborns, Chicago (39) BCG (MP) (Tice) 4839 eligible and 6423 other BCG (MP) children, Georgia (40) (Tice) 1025 eligible mental BCG (MP) retardates, Illinois (41) (Tice) 77,972eligible and 113,855 BCG (IC) other children, Puerto (Birkhaug) Rico (42) 34,767eligible and 29,369 BCG (MP) others, 5+ years, Georgia (Tice) & Alabama (43) 10,877eligible and BCG tiC) 10,679 other villagers, (Danish) Madanapalle, India (46) 32,282eligible and BCG tiC) 21,957 other adoles(Danish) cents, England (45) Vole (MP) 4030 eligible and Vole (MP) 13,667 other miners, Zimbabwe (47) 16,314newly employed BCG (?) miners, S. Africa (48) (Glaxo) 3174 eligible, Haiti (49) BCG (IC) (Frappier) 96,846 eligible and BCG (IC) 116,461 other villagers (Danish; Chingleput, S. India (50) French)

aMethod of vaccination: (IC) intracutaneous infection w i t h needle and syringe; (MP) multiple puncture; (O), oral; (?) not stated. ~Eligible = tuberculin-negative. c(

) Percent reduction in deaths from all causes.

abased on three cases a m o n g controls and n o n e a m o n g vaccinees.

252

G.W. Comstock started in the 1930s. As shown in Table 3, eight controlled trials were initiated between 1932 and 1937 [9,30-36]. Only two, however, seemed to have caught the attention of tuberculosis control workers in the United States [34,35]. Reports of these two trials were published shortly after the end of World War II, and both showed that a high level of protection had been conferred by BCG vaccination. These two reports and the enthusiasm for vaccination they helped engender p r o m p t e d action in two of the few developed nations in which BCG vaccine was not being widely used. The U.S. Public Health Service appointed a committee in 1946 to r e c o m m e n d policies regarding vaccination [37], in part in response to pressure from Mary Lasker, an influential public health crusader, but also from Carroll Palmer's desire to conduct controlled trials of BCG vaccination u n d e r field conditions [38]. In Great Britain, the Medical Research Council appointed a Tuberculosis Vaccines Committee in 1949 [11]. Both groups r e c o m m e n d e d that large-scale controlled trials be instituted. As a result, seven trials were started in the United States during the period 1941-1950 [39-44[, one in Great Britain in 1950 [45], and one in Madanapalle in 1950 under the auspices of the World Health Organization's Tuberculosis Research Office [46]. Since 1950, four additional controlled trials of vaccination against tuberculosis were started [47-50]. It is worth noting that the two in India were supported in large part by the U.S. Public Health Service.

BCG VACCINES

Forty years ago, Anderson and Palmer said, "We have a woefully incomplete knowledge of just what it is that is being used as BCG" [511. This unfortunate situation, which still persists, arose because until the mid-1960s BCG vaccine had to be serially subcultured every few weeks, often on different culture media in different laboratories [52]. Even when freeze-drying m a d e it possible to preserve seedlots from which lots of vaccine could be produced, some subculturing was necessary to produce enough vaccine for a reasonably sized lot. To make matters worse, there is no universally accepted method for assessing vaccine potency. C o m p o u n d i n g confusion, m a n y reports in the literature fail to state the source of their BCG vaccine and its k n o w n characteristics. It is known, however, that BCG strains can and do vary in m a n y respects. Strains have been found to differ in their ability to multiply and survive on laboratory culture media [53], colony characteristics [54], viability counts [551, drug resistance [56], DNA restriction fragments [57], virulence in animals [58-611, protection against tuberculosis and tumors in animals [62-65], the frequency and degree of postvaccinal skin sensitivity to tuberculin in humans [66,67], and in the presence of the MPB64 gene [68]. With all these documented differences between strains, it is unreasonable to believe that there are no variations in their ability to protect h u m a n s against tuberculosis. Clearly BCG is not BCG is not BCG; there are m a n y BCG vaccines. This fact should be kept firmly in mind in any discussion of the efficacy of BCG vaccines and the appropriate use of BCG vaccination.

Tuberculosis Vaccine Trials

253

OTHER VACCINES Although a n u m b e r of mycobacteria other than BCG have been suggested as possible vaccines, only two, killed Mycobacterium tuberculosis and Mycobacterium microti, the vole bacillus, have been used in controlled trials in humans. Killed tubercle bacilli were used in a vaccine trial among mental hospital patients in Jamaica in 1932 [9]. A life table reanalysis of their data over a 10-year period showed that the reduction in tuberculosis cases among persons eligible for vaccination was 35%. Interest in the vole bacillus as a potential vaccine arose as a result of investigations by the Bureau of Animal Population at Oxford into the wide fluctuations that occurred a m o n g bank voles in Great Britain [69]. In 1936, A. Q. Wells was called in to examine trapped voles for disease. He found tuberculosis-like lesions in some animals, isolated M. microti, and was surprised to find it completely avirulent for guinea pigs. By 1942, he was satisfied that it was safe to use vole bacilli to vaccinate humans. In 1950, the Medical Research Council included this vaccine along with BCG in its major trial of vaccination against tuberculosis [11], and in 1953 Paul used vole bacillus vaccine in a controlled trial among miners in Zimbabwe [47]. The results in both trials indicated that its efficacy was similar to that of the better BCG strains. Neither killed tubercle bacilli nor the vole bacillus vaccine is in current use. Killed tubercle bacilli have been found to produce only a modest level of protection in animals [6,7]. It m a y also be that tuberculosis workers never learned about the Jamaica trial since it was reported in a journal unlikely to reach tuberculosis physicians, w h o throughout much of the 20th century were all too often isolated along with their patients in remote sanatoria. The high degree of efficacy shown by vole bacillus vaccine might have m a d e it an acceptable alternative to BCG except for two considerations. By the time this efficacy was demonstrated, BCG had become widely accepted and vole bacillus vaccine had no clear-cut advantages over its established predecessor. Its death knell as a vaccine was probably assured by the discovery that some vaccinated persons developed sarcoid-like lesions at the vaccination site. Strongly allergenic lots were more likely to cause these lesions than lots that p r o d u c e d a lower proportion of tuberculin reactors [71].

C O N T R O L L E D TRIALS A total of 21 controlled trials of vaccination against tuberculosis have been reported. A succinct s u m m a r y of the 19 that were continued to a planned conclusion is given in Table 3, which also gives the most recent references to the findings. It is clear that there is a wide variety in the nature and location of the study populations, vaccines and methods of administration, types of outcome, and estimates of efficacy and usefulness. The trials also exemplify a considerable variety of experimental designs. No attempt will be m a d e to review all features of these trials. In the author's opinion, there are no defects in any of them that could have appreciably affected the results, although there are differences of opinion on this conclusion [72]. However, the purpose of this presentation is to indicate some

254

G.W. Comstock problems that arose in these long-term trials as conducted u n d e r field conditions. It is h o p e d that these past experiences, not widely k n o w n among the controlled trial community, m a y help in future trials aimed at evaluating the new tuberculosis vaccines that are expected to result from the recent advances in immunology.

Study Populations Ten of the 19 trials in Table 3 were done in the United States, two in India, and one each in Algeria, Canada, Haiti, Jamaica, South Africa, United Kingdom, and Zimbabwe. The study populations were composed of infants (7 trials), children and adolescents (4 trials), general c o m m u n i t y populations (4 trials), mental patients (2 trials), and miners (2 trials). Most of the populations were selected because they were considered to be at higher than average risk of developing tuberculosis. The exceptions are the trials in Georgia and Alabama where the tuberculosis problem was similar to that for the United States as a whole. In addition to the trials in Georgia, Alabama, and Puerto Rico, the U.S. Public Health Service started two other trials. Only a preliminary report was published [44]. A m o n g approximately 20,000 mental patients in Ohio, 43 cases of tuberculosis were diagnosed a m o n g the vaccinated population during the period 1948-1952, compared to 36 cases expected from the experience of the controls. Some 27,000 American Indian schoolchildren were entered into a trial in 1949. During the 3-year follow-up period, no tuberculosis deaths occurred a m o n g the vaccinated whereas three deaths would have been expected. In both groups, approximately 80% of the cases occurred among persons w h o were initially tuberculin-positive. Because of b u d g e t a r y limitations, these studies were discontinued in 1952 in favor of the larger studies in Puerto Rico and in Georgia and Alabama.

Prevaccination Tuberculin Testing In the brief description of the study populations in Table 3, the word "eligible" signifies that the subjects so labeled were negative tuberculin reactors and had no contraindications for admission to the trials. It is assumed that all n e w b o r n infants were negative reactors to tuberculin. Because there is no evidence that vaccination with any of the vaccines used in these trials confers any benefit on persons w h o are tuberculin skin test reactors, and because in the earlier days there was some concern that persons w h o reacted to tuberculin would have unacceptable vaccination reactions, prevaccination tuberculin testing was usually done. The major exceptions were trials a m o n g infants. At this age, especially if newborn, they are pres u m e d to be uninfected and hence to be negative reactors to tuberculin. While it is possible that at all ages some negative reactors m a y have been infected but had not yet d e v e l o p e d demonstrable tuberculin hypersensitivity, in only one trial has this been considered as a possible explanation for the outcome, in this instance a failure to demonstrate benefit from vaccination [73]. In the trial a m o n g South African miners, only a sample was tuberculin-tested; but the results were not used to exclude the tuberculin reactors from the trial [48].

255

Tuberculosis Vaccine Trials Table 4

Tuberculosis Cases in the Trial Populations with the Effect of Vaccination R e m o v e d and the Percentage of the Potential Tuberculosis Problem Accounted for by Initial Tuberculin Reactors

Trial Jamaica (9) Georgia schools (40) Puerto Rico (42) Georgia and Alabama (43) Madanapalle (46) England (45) Zimbabwe (47) Chingleput (80)

Observed, Observed, Other Reactors Nonvaccinated 161 36 1400 207 135 274 39 1276

67 6 315 40 47 248 3 47

Reactor Cases Expected, ~s ~ of Vaccinated a Potential Total f' 69.4 3.2 260.7 34.1 41.0 374.2 3.5 94.0

56.0 79.6 70.9 73.6 60.5 30.7 85.7 90.0

aCalculated by applying nonvaccinated (control) rate to vaccinated population. bObserved cases in reactors divided by sum of cases in first three columns of figures and multiplying by 100. Even t h o u g h prevaccination tuberculin testing is no longer considered necessary in routine vaccination p r o g r a m s , it is i m p o r t a n t in a controlled trial because the exclusion or inclusion of previously infected persons, i.e., those w h o h a v e positive tuberculin reactions, affects the m e a s u r e being determined. A c o m p a r i s o n of the incidence of disease a m o n g negative tuberculin reactors in the vaccinated and control g r o u p s is a m e a s u r e of vaccine efficacy, the ability to protect against disease in the p o p u l a t i o n that is capable of being benefitted. But if positive tuberculin reactors are included in each group, the m e a s u r e d e p e n d s not only on vaccine efficacy but also on the difference in risk of disease a m o n g controls and positive reactors, and on the p r o p o r t i o n of reactors in the s t u d y population. This is different f r o m the case of acute c o m m u n i c a b l e diseases in which infected survivors are no longer at risk and their inclusion in a trial does not affect estimates of efficacy [79]. In tuberculosis, a disease in which the risk a m o n g tuberculin reactors persists for m a n y years a n d usually differs m a r k e d l y f r o m that a m o n g nonreactors to tuberculin, the inclusion of reactors a m o n g vaccinees and controls will tend to yield a m e a s u r e (i.e., reduction in the total tuberculosis p r o b l e m or usefulness) that is lower than the true efficacy of the vaccine [75,76]. A major surprise f r o m the trials conducted shortly after World War II was the high p r o p o r t i o n of tuberculosis that d e v e l o p e d a m o n g persons w h o were already tuberculin reactors. This is s h o w n in Table 4. Only in the British trial a m o n g school leavers did the cases a m o n g initial reactors c o m p r i s e a minority of the potential tuberculosis problem. In all the others, a majority, and in s o m e instances a very large majority, of the cases arose a m o n g reactors, for w h o m vaccination offered no benefit. In several of the trials, this situation has been s h o w n to persist for u p to 21 years [40,42,43,46]. The first e x a m p l e of h o w the p r o p o r t i o n of cases coming from tuberculin reactors could cause the reduction in the total p o p u l a t i o n caseload to be less than the efficacy of the vaccine w a s reported f r o m the trial in Jamaica [9]. The efficacy of the vaccine in p r e v e n t i n g tuberculosis cases w a s 35%. As s h o w n in Table 5, the reduction in the total tuberculosis p r o b l e m w a s only 17% because

256

G.W. Comstock Table 5

Potential I m p a c t of Vaccination on the Tuberculosis Problem in a Jamaican Mental Hospital, 1932-1941 Tuberculosis Cases

Impact

Reactors

Observed numbers of cases Expected cases, no eligibles vaccinated Expected cases, all eligibles vaccinated Reduction in tuberculosis problem by vaccinating all eligibles

232 87 232 87 232 52 409 - 338 _ 17% 409

Source:

Controls

Vaccinees 54 90 54

Total 373 409 338

Data from Ref. 9.

75% of the population were positive tuberculin reactors and not eligible for vaccination. In addition, unlike the situation in m o s t c o m m u n i c a b l e diseases, they were at continuing risk of b e c o m i n g cases. If no vaccination had been done, the attack rate a m o n g vaccinees w o u l d h a v e been the s a m e as that a m o n g controls, p r o v i d i n g 90 cases a m o n g vaccinees. Conversely, a p p l y i n g the observed attack rate a m o n g vaccinees to the control p o p u l a t i o n gives 54 expected cases if all eligibles have been vaccinated. The total potential tuberculosis problem, 409 cases, w o u l d h a v e been reduced to 338 cases had all eligibles been vaccinated. This illustrates a c o m m o n difficulty with vaccination of a general p o p u l a t i o n as a tuberculosis control p r o c e d u r e - - i f a large proportion of tuberculosis cases arises a m o n g tuberculin reactors, even a highly efficacious vaccine will have only a slight impact on the total problem. This p r o b l e m is m i n i m i z e d b y limiting vaccination to y o u n g children and avoided completely by vaccinating newborns. During the last 4-1/2 years of the Jamaica trial, 184 persons w h o did not react to the 0.01-mg dose of old tuberculin but did react to the 1.0-mg dose were admitted to the trial. In this s u b g r o u p , the reduction in tuberculosis attributable to vaccination w a s 46%. This was an early indication that the mycobacterial infection(s), which had caused the w e a k sensitivity in these persons that could only be detected b y a strong dose of tuberculin, had not p r e e m p t e d some of the efficacy of the vaccine, in contrast with later evidence in guinea pigs and current w i d e s p r e a d belief [77,78]. In the study a m o n g South African miners, the initial p r o c e d u r e was to vaccinate all n e w e m p l o y e e s a n d to rely on the old e m p l o y e e s as a comparison g r o u p [48]. W h e n the a v e r a g e annual case rates a m o n g the two g r o u p s were compared, there w a s 60% less tuberculosis a m o n g the vaccinated g r o u p (Table 6). H o w e v e r , it was then realized that the "old timers" had higher rates of tuberculosis than n e w l y e m p l o y e d miners, thereby vitiating the comparison. The p r o c e d u r e was changed so that n e w e m p l o y e e s with even c o m p a n y n u m b e r s were vaccinated and those with odd n u m b e r s were left unvaccinated. H o w e v e r , tuberculin testing a m o n g a s a m p l e of controls indicated that as m a n y as half of the new e m p l o y e e s could have been infected with tubercle bacilli on e m p l o y m e n t . Consequently, the c o m p a r i s o n of incidence rates during the second period gives an estimate of the effectiveness of the vaccination p r o g r a m in reducing tuberculosis in this particular population and not the efficacy of the vaccine.

257

Tuberculosis Vaccine Trials

Table 6

Tuberculosis Cases A m o n g BCG-Vaccinated and Unvaccinated E m p l o y e e s of a South African Mine During T w o Study Periods

Employment and Vaccination Status

Average Number of Employees per Year*

Tuberculosis Cases No.

Av. Ann. Rate, %

New employees vaccinated; old employees are "'controls'" Vaccinated new employees 4550 54 Unvaccinated old employees 3165 98

0.40 1.03

New employees allocated on basis of company numbers Vaccinated new employees 8317 29 Unvaccinated new employees 7997 45

0.12 0.19

,Includes unknown proportions of persons previously infected with tubercle bacilli (positive tuberculin reactors). Source: Data from Ref. 48. In addition to the South African trial, no prevaccination tuberculin testing was d o n e in seven other trials because their participants w e r e infants [3033,35,36,39]. In three others, prevaccination testing was done to select eligible persons for allocation to vaccinated or control groups, but the positive reactors w e r e not followed and their risk is u n k n o w n [31,34,41]. In the remaining eight, the s u b s e q u e n t incidence of tuberculosis w a s recorded for vaccinees, controls, and positive reactors, thereby allowing both efficacy and the potential overall impact of vaccination (usefulness) to be determined. If a small p r o p o r t i o n of s u b s e q u e n t tuberculosis in the s t u d y population comes from persons w h o had been p r e v i o u s l y infected, the positive reactors at baseline, then an efficacious vaccine can m a k e a considerable reduction in the overall tuberculosis problem. As can be seen in Table 3, this w a s the situation in the British Medical Research Council Trial [45[. On the other hand, if m o s t of the s u b s e q u e n t tuberculosis comes from previously infected persons, then no vaccine can h a v e a major i m p a c t on the problem. The Chingleput Trial in South India, in which 90% of the tuberculosis came from persons infected at baseline, exemplifies this situation [79].

Effect of Nontuberculous Mycobacterial Infections The dose of tuberculin used in prevaccination testing has turned out to h a v e introduced another complicating factor into the interpretation of trial results. At the time of the earlier trials, it was generally believed that a negative reaction to a strong dose of tuberculin (100 or 250 tuberculin units, TU) was required to rule out previously acquired tuberculous infection. Then, d u r i n g the 1950s, it w a s discovered that w e a k tuberculin sensitivity (so w e a k that it required a strong dose of tuberculin to detect it) tended to be caused by infections with nontuberculous mycobacteria, s o m e t i m e s called atypical or e n v i r o n m e n t a l mycobacteria [80]. Populations that had been screened by 100 or 250 TU of tuberculin w o u l d tend to h a v e persons with nontuberculous mycobacterial infections r e m o v e d from both vaccines and controls. Those screened only with 5- or 10-TU doses w o u l d h a v e excluded nearly all persons w h o had p r e v i o u s l y been infected with tubercle bacilli (Mycobacterium tuber-

258

G.W. Comstock Table 7

Estimated Vaccine Efficacya by Frequency of N o n t u b e r c u l o u s Mycobacterial Infections, and Degree of Tuberculin Sensitivity, a n d M e t h o d of Vaccination b

Nontuberculous Mycobacterial Infections Not common

Common

Newborn Infants 81%IC(32) 59% IC ( 3 6 ) 75%MP (35) 74%MP (39)

Negative Tuberculin Reactors 100--250 TU c 5-10 TU c 75%IC(34) - 4 0 % M P (41) 77% IC (45) -57% MP (40)

29% IC (42) 6%MP (43) 20% IC (46) 67% IC (49) -2% IC (50)

aExcludes trials using oral vaccination and those not reporting efficacy based on incident cases. blC, intracutaneous; MP, multiple puncture. cTU, tuberculin units; 100 or 250 TU tests were given only to persons with negative reactions to a preliminary 5- or 10-TU test.

culosis, or M. bovis), but these vaccinee and control g r o u p s w o u l d contain u n k n o w n p r o p o r t i o n s of persons infected with the other mycobacteria that are n o r m a l l y harmless. These n o n t u b e r c u l o u s mycobacteria have been found to be v e r y c o m m o n in m o s t tropical countries, m o d e r a t e l y c o m m o n in the t e m p e r a t e zone, and essentially nonexistent in the Arctic. The hypothesis w a s then raised, and confirmed b y an elegant s t u d y a m o n g guinea pigs, that n o n t u b e r c u l o u s mycobacterial infections conferred s o m e protection against tuberculosis [77]. Further, w h e n guinea pigs with nontuberculous mycobacterial infections w e r e vaccinated with BCG, there was no additive effect. Vaccination merely raised their partial i m m u n i t y against tuberculosis to the level attained b y vaccinated animals that had had no p r e v i o u s n o n t u b e r c u l o u s mycobacterial infections. The a p p a r e n t efficacy of a vaccine w o u l d thus v a r y inversely with the level of partial i m m u n i t y conferred by an earlier n o n t u b e r c u l o u s mycobacterial infection. From this, it follows that vaccine efficacy should a p p e a r greatest a m o n g s t u d y populations f r o m which persons with n o n t u b e r c u l o u s mycobacterial infections had been r e m o v e d by screening with a strong dose of tuberculin. By the s a m e token, one w o u l d expect that vaccine efficacy as m e a s u r e d in a controlled trial w o u l d be least in areas w h e r e nontuberculous mycobacterial infections are m o s t c o m m o n a n d better in areas w h e r e they are u n c o m m o n . This has been found to be generally true [81]. Unfortunately, the effect associated with g e o g r a p h y is c o n f o u n d e d by w h e t h e r or not persons with w e a k tuberculin sensitivity w e r e excluded by prevaccination tests with a strong dose of tuberculin, and also by the age of the s t u d y populations. Only n e w b o r n infants in these trials can be p r e s u m e d to h a v e been free of all mycobacterial infections. A further complicating factor is that different methods of administering the vaccines were also e m p l o y e d , in addition to the use of different strains in m o s t of the trials. The degree of confounding s h o w n in Table 7 is one indication of the impossibility of d r a w i n g reasonable general-

Tuberculosis Vaccine Trials Table 8

259

Apparent Efficacy of BCG Vaccine in Puerto Rico Trial According to Degree of Reaction to a Strong Dose of Tuberculin (100 TU) at the Time of Vaccination Diameter of Induration to 100 TU PPD

Vaccinees

0 1-5 6+

18 19 22

Average Annual Tuberculosis Rates per 100,000 Controls Efficacy (%) 25 27 29

28 30 24

Source:Data from Ref. 42. izations from such variegated data. For example, all trials involving newborn infants were done in areas in which nontuberculous mycobacterial infections are not common, while all the trials that did not exclude persons with such infections by a strong tuberculin test were conducted in areas where these infections are common. Doubt is cast on the potential preemptive protection afforded by nontuberculous mycobacterial infections by findings in the trial in Puerto Rico [42]. In this trial, both the vaccinees and the controls were negative to a 10-TU test prior to vaccination and also received a 100-TU test at the time of vaccin ~tion. If the p r e e m p t i v e effect were important, one w o u l d expect the measured efficacy of BCG vaccine to be least a m o n g those w h o reacted with 6 m m or more of induration to the 100-TU test dose (those most likely to be infected only with nontuberculous mycobacteria), better a m o n g those with reaction sizes of 1-5 mm, and best a m o n g those with no induration. As can be seen in Table 8, this expectation is not borne out. Additional information on this point should come from the trial in the Chingleput area of South India, since participants were tested with 5 TU of the international standard tuberculin (PPD-S) and also with a sensitin prepared from Mycobacterium avium-intracellulare, PPD-B. Unfortunately, the results of this dual testing as they relate to subsequent case rates a m o n g vaccinees and controls have not yet been reported. Allocation M e t h o d s

H o w eligible participants are allocated to vaccinated and control groups is n o w widely recognized to be crucial for the validity of results from controlled trials. In five of the trials in Table 3, allocation was by alternation [9,31,3435,39], in one partly by alternation and partly "at r a n d o m " [36], in three it was by the terminal digit of an identification n u m b e r [33,45,48], and in three by year of birth [40,42,43]. In two trials, the allocation p r o c e d u r e was not described [41,47], and in one trial there was a " r a n d o m distribution of vaccines and vaccine lots" [49]. In the Madanapalle trial, the assignment to vaccinated and control groups was r a n d o m except in instances where unexpected procedural deviations occurred [46]. Only in the Chingleput trial would the allocation p r o c e d u r e and its description be considered acceptable by m o d e r n standards [80]. The others used a variety of methods, some of which provided instructive examples.

260

G.W. Comstock Table 9

Tuberculosis Deaths A m o n g BCG-Vaccinated and Control Children, N e w York City, During T w o Periods of the Trial Vaccination Status

No. of Children

No.

Tuberculosis Deaths %

1926-1933 (before alternate allocation to study groups) Vaccinees 445 3 Controls 545 18

0.7 3.4

1933-1944 (after alternate allocation to study groups) Vaccinees 566 8 Controls 528 8

1.4 1.5

Source: Data from Ref. 31.

Alternate allocation by definition ought to yield nearly identical n u m b e r s of vaccinees and controls, as should the use of o d d and even identification numbers. This goal was n e v e r achieved and in only one instance is an explanation given. The N e w York City trial [31] recruited children from tuberculous households w h o were seen in the Health D e p a r t m e n t clinics. Eligible children had to be negative tuberculin reactors unless they were u n d e r one m o n t h of age; almost all were u n d e r one year of age. From 1926 to 1933, clinic physicians w e r e instructed to vaccinate half of the eligible children seen in their clinics. From 1933 to 1944, children were allocated to vaccinee or control g r o u p s by the central office. The results during a 5-year follow-up are s h o w n in Table 9. In the first period, the reduction in tuberculosis deaths associated with vaccination was 80%. Although this w a s a statistically significant result (p ~ 0.01), it was not epidemiologically valid. After central office allocation was adopted, there was no d e m o n s t r a b l e difference. Unfortunately, two things changed at once. BCG vaccine was obtained f r o m Calmette at the start of the trial and m a i n t a i n e d in serial cultures locally but this strain b e c a m e c o n t a m i n a t e d in 1932. BCG was recovered only after treatment with antiformin and aniline dyes. The recovered vaccine was used in the second s t u d y period together with vaccine obtained from Phipps Institute, the strain s h o w n by Aronson et al. to have been efficacious a m o n g American Indians [34]. A change also likely to have affected the o u t c o m e involved the nature of the participants. During the first period, vaccinated children tended to come from patients actually seen by the clinic physicians, i.e., those w h o kept their a p p o i n t m e n t s best. As a result, vaccinees came from cooperative families with higher socioeconomic status and less serious tuberculosis exposures. During the second period, vaccinees and controls were similar in these respects. As can be seen in Table 9, the initial favorable results a p p e a r to be due m o r e to selection for vaccination than to vaccine efficacy. In the earliest of the trials in Table 3, participants were recruited from children born into families in which there was a case of tuberculosis at that time [30]. Vaccination by the oral route was given to children in which the three-dose schedule could be completed in the first 10 days of life while they were still in the hospital. Controls were older infants with similar h o m e conditions. A m o n g the vaccinees 52% were subsequently exposed to persons

Tuberculosis Vaccine Trials

261

w h o still had demonstrable tubercle bacilli in their sputum, compared to 50% a m o n g controls. Although no mention is m a d e of whether or not those w h o diagnosed tuberculosis as cause of death knew the vaccination status of the participants, 11 of the 12 deaths a m o n g controls were d u e to miliary tuberculosis and one to generalized lymphatic disease with six deaths being confirmed pathologically, while the only death a m o n g the vaccinees was due to p u l m o n a r y disease following measles and did not have bacteriologic or pathologic confirmation. The major problem with this trial is that some of the older unvaccinated controls could have been infected by their household contacts before being enrolled into the trial whereas the vaccinated children at least had a start toward developing postvaccinal cellular i m m u n i t y before being exposed to infection by virtue of being kept in the hospital for 10 days. In this connection, it should be recalled that on average it requires a considerable duration of exposure for infection to occur [86]. Even with this problem, and an inadequate description of several features of the trial, it is hard to claim that vaccination conferred no protection. However, these defects have caused this trial to be generally ignored. A unique m e t h o d of allocation was e m p l o y e d by Ferguson and Simes in their vaccination trial in the Qu'Appelle Valley of Saskatchewan, Canada [32]. At the start of the trial, families were grouped in pairs, with each pair consisting of families as similar as possible "in respect of housing, sanitation, and certain other economic and social factors likely to affect the health of children." One family of each pair was r a n d o m l y assigned to g r o u p A and the other to g r o u p B. Vaccinees were all children born into g r o u p A families during one year and into group B families the following year. This alternation was continued t h r o u g h o u t the study, yielding 306 vaccinated and 303 unvaccinated control children. The tuberculosis case rates were 3.0 and 15.8 per 100 person-years of observation, respectively, yielding an estimate of efficacy of 81%. A measure of the similarity among the two groups is the probability of dying of nontuberculous causes during the follow-up period. This can be calculated as 22% a m o n g the vaccinees and 20% among the controls. H o w ever, the study cannot be considered doubly m a s k e d - - a u t o p s i e s were performed on 64% of the deaths a m o n g vaccinees and only 25% of those a m o n g controls. An experience in the trial in Madanapalle, South India should give pause to e v e r y o n e w h o believes they have a foolproof system of allocation [821. In this rural setting, it was not considered practicable to assign serial numbers to the participants. The numbers on chest photofluorograms could not be used because transportation problems sometimes resulted in these examinations being done after vaccination. Birth years and dates of birth were u n k n o w n to most villagers at that time. To avoid these problems, a simple randomization scheme was devised. Half of a pack of index cards that were used to record tuberculin test and photofluorographic results were marked with an X on the back. The deck was then shuffled before use. If a tuberculin test was negative, the card was to be turned over; persons whose cards showed an X were to be vaccinated. Although the system appeared to be working well, a considerable n u m b e r of errors were discovered only at the time of analysis. The results from a 20% sample of completed records are s u m m a r i z e d in

262

G.W. Comstock T a b l e 10

Distribution of a Sample of Field Records in the Madanapalle Trial, by Tuberculin and Vaccination Status, and by the Presence or Absence of an X to Indicate Selection for Vaccination if the Tuberculin Test was Negative X on card

Source:

Tuberculin-Negative Vaccinated Not Vaccinated

TuberculinPositive

Present Absent

1009 12

158 1257

939 1384

Total

1021

1415

2323

Data from Ref, 82.

Table 10. If the system had w o r k e d perfectly, there would have been no vaccinee records without an X, no control records with an X, and nearly equal numbers of records of the positive reactors with and without X's. This goal was obviously not achieved. After considerable investigation, the major causes for deviations from the desired results were identified. First, it was noted that these deviations increased as the intake for the trial progressed. It did not take long for the field workers to realize that since positive tuberculin reactors did not get vaccinated, marking X's on their cards was wasted effort. When pressed for time, they did not mark the cards until the tuberculin tests had been read. Then half of the cards for negative reactors were marked, leaving cards for positive reactors unmarked. This, however, did not account for discrepancies a m o n g the negative reactors. In this instance, it was found that nearly all of the discrepancies occurred w h e n the photofluorographic unit visited the village before the arrival of the tuberculin testing and vaccination team. When chest photofluorograms were taken first, none of the cards had X's. If this was not p r o m p t l y noted by the tuberculin testing team, persons with cards from which X's were missing for this reason would be allocated to the control group. The excessive n u m b e r of control cards with X's was not explained; one w o n d e r s if these were not persons who had contraindications for vaccinations or those who slipped away between the "placebo" test with the 100-TU test and vaccination. In any case, it is difficult to see how these deviations from strict randomization could have had any major effect on the results over a 20-year follow-up period. The principal lesson appears to be that even with a simple, straightforward system, constant vigilance is the price of freedom from error.

U s e of a Placebo

While few w o u l d question the desirability of using a placebo in evaluating the efficacy of a vaccine, only 3 of the 19 controlled trials of vaccination against tuberculosis did so. One of the early trials among American Indians used an intracutaneous injection of physiologic saline [34], the trial in Haiti used smallpox vaccine [491, and the Chingleput trial used the vaccine diluent [79]. In the trials in Georgia and Alabama [40,43], a placebo was seriously considered but only in the context of mimicking BCG vaccination. For that purpose, nothing seemed suitable. Not until the analysis of results was

Tuberculosis Vaccine Trials

263

u n d e r w a y was another p u r p o s e of a placebo injection noticed. When persons returned to have their tuberculin tests read, tuberculin reactors and other persons born in even years were given a report of their findings and excused. Nonreactors to tuberculin born in odd years were seen by other team members and offered vaccination. Persons w h o refused vaccination and those w h o were found to have medical contraindications to vaccination were thus rem o v e d from the vaccinee g r o u p whereas similar persons among the controls were not identified and remained a m o n g the controls. Fortunately, there were relatively few such irregulars and their subsequent tuberculosis status was ascertained in the same w a y as other participants. Corrections for their experience could be made; these slightly decreased the already distressingly low estimates of efficacy [40,43]. However, had a placebo been used in these trials, refusals and medical contraindications would have been identified similarly in each group, and no post hoc estimates would be necessary. This problem has not been discussed in reports of other trials. In two trials, another p r o c e d u r e was used that not only a d d e d useful information but acted in some ways as a placebo. In both Puerto Rico and Madanapalle [42,46], all negative reactors to the initial screening dose of tuberculin were given a 100-TU test dose. In this way, both vaccinees and controls were seen by a second unit of the team; both had similar opportunities to refuse and to have medical contraindications noted. In addition, the frequency and size of reactions to the 100-TU dose p r o v i d e d estimates of the frequency of infections with nontuberculous mycobacteria.

Vaccines and Methods of Administration Still other variables between trials were the vaccines and the methods of administering them (Table 3). In addition to heat killed tubercle bacilli of u n k n o w n parentage used in the Jamaica trial, substrains of vole bacillus were used in the British and Z i m b a b w e trials and 11 different strains of BCG were administered in the other trials. The intracutaneous route of administration was used in nine trials, the multiple puncture (percutaneous) methods in six, and the oral route in two. One did not state the m e t h o d of administration. The British trial used two methods: intracutaneous for BCG and multiple p u n c t u r e for the vole bacillus [83]. Comparisons between strains and m e t h o d s of administration are difficult because only infrequently have two strains been used in the same trial and never have two methods of administration of the same vaccine been e m p l o y e d in this way. In the N e w York City study [31], two BCG vaccine strains were used but the results were not reported by strain. In Haiti, the conventional Montreal strain and its isoniazid-resistant variant caused reductions in total tuberculosis incidence of 65% and 70% respectively, w h e n c o m p a r e d with the placebo g r o u p [49]. In the Chingleput trial, results were not reported separately for the Danish and French strains but it was stated that each gave similar results [79]. In the British trial, strains of the vole bacillus vaccine causing low and high degrees of postvaccinal tuberculin sensitivity gave virtually identical reduction in tuberculosis [84]. Another interesting finding from the Chingleput trial was that there has been no demonstrable difference in efficacy between the usual dose of BCG

264

G.W. Comstock vaccine and a dose only one-tenth as large, although demonstrating such a difference is admittedly difficult w h e n neither dose showed any evidence of benefit. The weaker dose did cause smaller postvaccinal tuberculin reactions. Although the original oral route of administration has been out of favor for a long time because it had to be given on at least three different days, the full dose was not always swallowed, and the production of postvaccinal tuberculin sensitivity was unreliable, there have been suggestions of its efficacy that cannot be entirely ignored. In addition to the previously cited small study a m o n g infants in Philadelphia [30], a large trial was conducted among newborn infants in Algiers in the 1930s [33l. This appears to be a good illustration of h o w to conduct a large longitudinal study on a limited budget, since it was based on original and foUow-up records routinely collected for other purposes. As births were registered, allocation for vaccination was determined by registry number. BCG vaccine was given at h o m e by the oral route. Follow-up consisted solely of linking deaths a m o n g infants and children to the records of vaccinees and controls. Over an l 1-year period, the overall death rate a m o n g vaccinated children was 11% lower than among control children. Assuming that the allocation procedure was properly followed, it seems reasonable that linkage errors and population losses from emigration would be distributed similarly a m o n g the two study groups and therefore would have no effect on estimates of reduction in total mortality. If one assumes that the death rates from tuberculosis and other causes in Algiers were similar to those a m o n g Canadian Indians in Saskatchewan [32] and that the efficacy of the vaccine used in Algiers was 80%, the expected reduction in total mortality turns out to be 11%, the same as that actually observed in Algiers. Although m a n y details of considerable interest are omitted from the reports of this trial, it is again hard to escape the conclusion that oral vaccination had at least a modest effect, as was also suggested by the experience a m o n g Philadelphia infants. Although both trials are open to criticism, it should be incumbent on critics to show h o w the supposed defects could reasonably lead to important differences in the outcome. This is rarely done in critiques of epidemiologic studies.

Postvaccinal Tuberculin Sensitivity For decades, it has been part of the d o g m a of tuberculosis that the efficacy of a vaccine is reflected by its ability to create tuberculin sensitivity among the vaccinated, even though it has long been k n o w n that immunity and tuberculin sensitivity can be dissociated in animals [85]. More convincing evidence is p r o v i d e d by the controlled trials of vaccination in h u m a n s [86]. The correlation of postvaccinal sensitivity to an intermediate dose of tuberculin and vaccine efficacy is illustrated in Fig. 1. Although within each vaccine strain there is a slight tendency for lots that p r o d u c e d the highest levels of tuberculin sensitivity to be more efficacious, the overall results show a negative correlation. Strains that caused the highest frequency of tuberculin reactors tended on average to be less efficacious. While, as noted earlier, the m a n y differences between controlled trials make any comparison uncertain, the findings s h o w n in Fig. 1 give no support to the still current practice of

265

Tuberculosis Vaccine Trials

I00-

02c

90 Q; °

2F @ @2D

C o 0

80

0

60"

I

(D E3

O9 O9

°

~

0

0t , , -

8A~ 8D Q

• 9A

70'

°2B

°

°

@IOA

02A @2E

50 40"

1

30-

0 7

I0"

@6

I

¢.0

O"

elB

~

-I0"

0 Q~ rr

- 20

cO

-.'50

0 Q~ EL

@80

lOB@

20"

I-¢-

8C 9B

-40"

@4

-50" "3

-60 0

,o

do

40

70

80

,oo

Percent Postvaccinal Conversions to Tuberculin Figure 1.

Postvaccinal conversion rates and corresponding percentage reduction in tuberculosis attributable to vaccination in 10 controlled trials (areas of the circle are proportional to the square root of the combined numbers of vaccinated and control subjects). 1. Killed Mycobacterium tuberculosis, Jamaica (8,9); 2. BCG lots,* western United States (104,105); 3, BCG, Georgia children (40,106); 4. BCG, Illinois mental retardants (41); 5. BCG, Puerto Rico (42,76); 6. BCG, Georgia-Alabama (43,76); 7. BCG, Madanapalle, India (46,82); 8. BCG lots,* England (84); 9. Vole vaccine lots,* England (84,107); 10. BCG lots, Haiti (49); 11. BCG lots,* Chingleput, India (50,79). Asterisk indicates combination of lots as grouped in references.

266

G.W. Comstock evaluating vaccine efficacy or vaccination effectiveness by postvaccinal sensitivity [87].

Follow-up Procedures Methods of follow-up also varied. In six trials, the method was not stated, although a reasonable presumption is that it depended on routine examinations for tuberculosis contacts [30,31;35], mental hospital patients [9,4l], and miners [47]. In six, active follow-up methods were applied either to the individual participants [32,34,36,39,49] or by periodic community chest photofluorographic surveys [46]. In another five, case detection depended solely on record linkage of routinely collected official reports: death from any cause in Algiers [33], tuberculosis cases and deaths in the U.S. Public Health Service Trials [40,42;43], and cases detected in periodic examinations for silicosis in South American miners [48]. In two other trials, both active and passive methods were employed. For the first 10 years of the British trial, participants were visited and examined at approximately 14-month intervals, including examination of the vaccination sites on the early visits, while for the last 10 years, cases were identified through linkage with official case and death reports [88]. Ascertainment by the latter method was almost as complete as by active follow-up [45]. In the Chingleput trial, a chest clinic was established to provide diagnostic and treatment facilities to an area where none had previously existed. Case finding also depended on chest photofluorographic surveys at approximately 2-1/2-year intervals [79]. While most reports of the trials gave few or no details of the follow-up procedures, those that did are worthy of consideration. Both active and passive methods of follow-up had potential problems with respect to maintaining independence of diagnoses and vaccination status. When active follow-up consisted of efforts to screen the entire community in which the trial participants lived or worked, maintaining independence appears to have been simple as long as record linkage was performed only after the diagnoses had been made. A more important problem is created when tuberculin test and vaccination records are used in the field for follow-up of individual participants. This was done in two of the few trials that provided details on follow-up procedures, namely, the trial among U.S. Indian children and the British trial [11,34]. To minimize the bias that knowledge of vaccination status by the field workers might have introduced, both studies had chest radiographs and other records reviewed without any information regarding vaccination. While bias is thus removed from the diagnosis process, there still remains the possibility that requests for interim examinations by the field personnel could be influenced by vaccinee or control status. The situation is aggravated when attention is called to this status by having vaccinee and control records in different colors as was done in the U.S. Indian trial, or when status is determined by the terminal digit of the serial number on the record as in the British trial. However, it is again difficult to see how the potential for bias could be sufficiently great to cause meaningful changes in the results. In a third trial, that among Chicago infants, it is stated that diagnoses were independently reviewed, but details were not given [39]. The possibility of biased case ascertainment with passive follow-up varies

Tuberculosis Vaccine Trials

267

with the method of allocation, the separation of study personnel from those doing the case finding, and the proportion of study cases in the total caseload. If allocation to vaccinee or control status is truly random, and records of status in the trial are available only in a central office away from the work in the field, as was the case in Chingleput, there should be no diagnostic bias [79]. In Puerto Rico, the study office was completely separated from the providers of case and death reports [76]. Although allocation for vaccination based on year of birth made it theoretically possible for outsiders to "break the code," it seems unlikely that this was actually done. The system of allocation was never made public, vaccinee and control cases made up only one-sixth of the 1901 cases among trial participants, and these study cases composed an even smaller proportion of the total caseload seen in the busy clinics. Had the original goal of enrolling all the children on the island in the trial been achieved, the situation would have been different. In the trials in Georgia and Alabama, things were somewhat more problematic because allocation was based on birth year and the group that conducted the trials also provided chest clinic services to the two counties included in the trial. Again, the method of allocation was not publicized. In the chest clinics, initial screening was by means of chest photofluorography. At these examinations, the available history included only identification data, reason for examination, and the presenting symptom, if any. The workload, approximately 20,000 screening examinations each year in addition to the provision of other clinic and hospital services, was not conducive to speculation about trial status, especially when cases among vaccinees and controls only composed approximately 3% of the total caseload. Estimates of vaccine efficacy were similar whether based on screening photofluorograms or subsequent diagnoses. More reassuring was the fact that an independent review of all films of cases occurring in the first 7 years of the trial showed no evidence of bias [76]. Careful reviews of possible problems with the Puerto Rico and Georgia-Alabama trials indicated that biases sufficient to cause important changes in the conclusions were highly unlikely [89,90]. Nevertheless, random allocation, similar to that in the Chingleput trial, probably would have been worth the effort. With current emphasis on detailed informed consent, random allocation would be the only acceptable method. In future trials, maintaining independence of diagnoses and vaccination status may be somewhat more difficult. In much of the world, it is now considered unethical in tuberculosis case finding to expose persons to even the slightest radiation from chest radiography unless they have been demonstrated to be positive tuberculin reactors. Because vaccinees will have received a vaccine that is likely to make them positive reactors, and controls will not, more vaccinees than controls will have chest radiography and the additional history that is likely to go with it. However, because tuberculosis patients will almost certainly be tuberculin reactors whether or not they had been vaccinated, this potential source of bias is not likely to be important if evaluation is based on manifest symptomatic cases.

Duration of Protection As with any vaccination procedure, there has been considerable interest in the duration of protection conferred by BCG vaccination against tuberculosis.

268

G.W. Comstock Relevant findings have been reported from nine of the trials [33--45,40, 42,43,45,46,50]. In only two has there been any appreciable variation during follow-up periods ranging from 11 to 23 years. In the British trial among adolescents, efficacy showed a significant decline over a 20-year period, with no demonstrable efficacy after 15 years [45]. In the Madanapalle trial, the cumulative estimate of efficacy with each successive 3-year period after vaccination ranged from a high of 83% for the period 7-9 years to only 20% at the end of 21 years. Much of this variability can be attributed to the small numbers of cases among vaccinees and controls--only 33 and 47, respectively, after 21 years [46]. In the trials in Algiers infants [33], U.S. Indian children [34], Georgia and Puerto Rican children [40,42], and the general population in Georgia, Alabama, and Chingleput [43,50], overall efficacy, whether high or low, has remained essentially constant for up to 20 years.

Forms of Tuberculosis

Whether vaccination has a differential protective effect against various forms of tuberculosis is also a matter of great interest. In these trials, the belief that it is effective in preventing miliary and meningeal tuberculosis is definitely confirmed. In eight of the nine trials in which miliary or meningeal tuberculosis was reported among vaccinees or controls, the rates were clearly higher among controls and vaccine efficacy against these forms of tuberculosis tended to be higher than against other manifestations of the disease [30,32,39,41,42,45,46,91[. Only in the trial in Georgia and Alabama was the situation reversed, although this was based on a total of only two cases among vaccinees and one among controls [43[.

Outcomes

Except for the trial in Algiers, in which the efficacy of BCG was assessed only by comparing total mortality among the vaccinated and control groups, all estimates of efficacy were based on reduction in incident tuberculosis cases or tuberculosis deaths among the vaccinees compared with controls. Although 10 of the trials also reported the degree to which vaccinated persons developed positive tuberculin reactions (Fig. 1), in none was an estimate of efficacy based on this outcome. Deaths from nontuberculous causes were reported only from the trial among adolescents in England [71]. After 15 years of follow-up, total mortality was 9% lower among the vaccinated group than among the controls. There were only 17 cancer deaths among the vaccinated and 12 among the controls, with the vaccinated showing a slight deficit of neoplasms of the lymphatic and hematopoietic organs, and a slight excess of other malignant neoplasms. Mortality from violent causes was similar in the two groups, while other diseases were slightly less common among the vaccinated. A later report based on 27 years of follow-up showed no difference in leukemia mortality between vaccinees and controls [92]. Cancers developing among vaccinated and control populations were stud-

269

Tuberculosis Vaccine Trials

Table 11

Observed and Expected Cancer Cases A m o n g Vaccinated Persons in Controlled Trials of BCG Vaccination in Puerto Rico, Georgia, and Alabama a (93,94) Puerto Rico

GeorgiaAlabama

Site

ICD-8

O

E

O

E

O/E Ratio b

Stomach Colorectal Lung Uterus c Ovary Prostate Bladder Lymphatic Hodgkin's Total

151 153,154 162 180,182 183 185 188 200,202 201

5 3 2 70 4 0 2 9 9 150

1.9 5.6 0.0 72.9 1.8 0.0 0.0 1.8 3.7 142.6

13 34 34 114 17 15 17 8 7 429

4.2 44.8 31.5 126.7 11.0 12.5 9.41 2.7 4.5 379.4

2.97 0.73 1.14 0.90 1.64 1.20 2.02 3.71 1.94 1.11

95% Confidence Limits 1.76, 0.51, 0.80, 0.74, 1.02, 0.67, 1.22, 2.16, 1.11, 1.02,

4.69 0.99 1.57 1.10 2.51 1.98 3.15 5.94 3.14 1.20

~Limited to sites with a total of 20 or more observed and expected cases in the two trials combined, and those sites where association was in the same direction in both trials. bFor combined trials. cCervix and corpus not reported separately for the Puerto Rico trials. Source: Data from Refs. 93 and 94.

ied in the Puerto Rico and G e o r g i a - A l a b a m a trials after follow-up periods of 23 and 28 years, respectively [93,94]. The results are s h o w n in Table 11 for cancer sites with more than 20 observed and expected cases in both trials combined and in which the direction of association with vaccination was in the same direction in both trials. In the G e o r g i a - A l a b a m a trial, a significant difference in total cancer cases between vaccinees and controls did not develop until the later years of the 28-year follow-up period. However, the excess cancer a m o n g vaccinees in the Puerto Rico trial decreased with the passage of time. Several conclusions come out of these observations. First, as noted by Sutherland, is the m u c h greater trustworthiness of results from controlled trials than from observational studies in which one can never have adequate assurance that the g r o u p s being c o m p a r e d are alike in all respects except for the characteristics being studied [92]. In studies dealing with tuberculin reactors and nonreactors, it is important to keep in mind that the risk factors for becoming infected with M . tuberculosis are risk factors for m a n y other diseases as well [95]. For example, in the trial a m o n g English adolescents, tuberculin reactors had cancer rates nearly twice those of nonreactors [71]. Hence, in experimental or observational studies, the analysis needs to be stratified on the tuberculin reaction status of the subjects. Most importantly, this experience emphasizes the desirability of preserving records of controlled trials for investigations of long-term outcomes which may, as in the case of the two American trials, be the opposite of what was anticipated initially. Unfortunately, the Puerto Rico records have been destroyed, although it m a y still be possible to look again at the experience of the G e o r g i a - A l a b a m a s t u d y population. Because the records of A r o n s o n ' s trial a m o n g Indian children in

270

G.W. Comstock the western United States were preserved, a nearly completed 58-year followup will allow a look at a n u m b e r of long-term outcomes, including cancer (M. DeBoer, personal communication).

FUTURE EVALUATIONS With respect to the future, several things are clear. First, there is a definite need in most of the world for an inexpensive vaccine for tuberculosis. One might hope too that its efficacy w o u l d approach the levels attained by m a n y of the vaccines against acute communicable diseases. An efficacy of 80%, or conversely a failure rate of 20%, is hardly an optimal goal. Undesirable side effects of current BCG strains, such as local ulceration and lymph node involvement, have been acceptable. Systemic disease caused by BCG is very rare [96]. Only in areas w h e r e chemoprophylaxis is a feasible c o m p o n e n t of a tuberculosis control p r o g r a m is the creation of tuberculin sensitivity by BCG vaccination undesirable. In this connection it should be noted that the administration of a vaccine of u n k n o w n efficacy in hospitals and similar settings of potential exposure to tuberculosis will make it impossible to identify situations where infections are occurring. Unidentified, they are not likely to be corrected. Only a vaccine of very high efficacy could justify the decision to destroy the means to identify and remove exposure hazards. A major problem, therefore, is to evaluate the efficacy of current and future vaccines. More controlled trials like those in the past seem out of the question because of their expense and duration. Observational studies, whether cohort or case-control, measure p r o g r a m effectiveness, which only partially depends on vaccine efficacy [97,98]. As was shown in Fig. 1, the degree of postvaccinal tuberculin sensitivity does not reflect vaccine efficacy. The ability of animal test systems to reflect efficacy in h u m a n s has been called into question by a study that s h o w e d m a r k e d disagreements among animal systems in current use [99]. There are ways, however, that could remove m u c h of the present uncertainty. One was p r o p o s e d recently, the other 20 years ago [43,86]. Both d e p e n d on finding or creating a situation in which the relative efficacy of vaccines can be compared. If two or more strains with different efficacies among h u m a n s can be found, an animal test system that reflects such differences could be used to rank vaccines in current use or those yet to be developed. Case-control studies of vaccination effectiveness can measure relative vaccine efficacy in areas in which vaccination is performed at birth or during early infancy and in which the vaccine has been recently changed, as long as there is no other change in the vaccination p r o g r a m [86]. In two such instances, the Japanese strain, and possibly the Glaxo strain also, were found to be superior to local modifications of the Danish and French BCG strains [100,101]. If seed lots of these vaccines are still available, the appropriate animal studies can be done at any time to ascertain which animal test systems best reflect the situation in humans. Only then could results in animals be reasonably extrapolated to humans. The relative efficacy of c u r r e n t - - o r future--vaccine strains could also be

Tuberculosis Vaccine Trials

271

assessed on a large scale, with consequent shortening of time to obtain a meaningful answer [43]. Several countries that have a need for vaccination and that vaccinate at birth could be recruited to use their own vaccine in even years and a different vaccine in odd years without making any other changes in their vaccination programs. Merely accumulating reported cases and deaths among persons born in these even and odd years would reflect effects of the vaccines. If the vaccines differed appreciably in efficacy, this should become apparent relatively soon, even if diagnoses of tuberculosis can be made in only a portion of the vaccinated population. Several cooperating countries would be needed because of the possibility that some of them might be testing vaccines of comparable efficacy. Both of these methods of evaluation avoid some pitfalls of the past. In each country, the two vaccines would be administered similarly to the same kinds of populations, avoiding the confounding that arises when each vaccine is tested under different circumstances. Vaccines would be assessed under actual field conditions. Changing vaccines once a year should not be a major problem. Deviations of a few days would be inconsequential; larger deviations could be taken into account in the analysis if dates of birth are known for the reported cases. Differences in complications could also be determined. Finally, the expense of such comparisons should be minimal; for slight additional expense, cases and a sample of controls could be investigated more thoroughly in a case-control approach. There is another situation in which it may be possible to obtain some estimate of the relative protective potencies of tuberculosis vaccines. This would involve trials of vaccination among persons infected with the human immunodeficiency virus at high risk of becoming infected with tubercle bacilli. Such individuals have an extremely high risk of developing manifest tuberculosis within a short time [102]. Because use of a placebo is likely to be considered unethical, comparisons of two or more vaccines seems more likely. However, deciding what to do about persons previously infected with tubercle bacilli will be important. While this problem will not arise among newborn infants, a quicker answer could probably be obtained by including older persons in the trial population. Unfortunately, detecting tuberculous infections among HIV-infected persons is unreliable because of the anergy that develops as CD-4 cell counts go down [103]. It is almost certain, therefore, that controlled trials of vaccination among HIV-infected persons will include at least some persons already infected with tubercle bacilli. For this reason, trials in these populations will yield a measure that is likely to be less than the vaccine's inherent efficacy.

CONCLUSIONS Controlled trials of vaccination against tuberculosis have involved a wide range of times, places, and persons. Methodologies and vaccines have been equally diverse. Although a review of such a polyglot collection cannot be expecfed to produce a single overall estimate of the value of vaccination against tuberculosis, much can be learned about the problems of controlled trials under field conditions and how to avoid those problems.

272

G.W. Comstock R e c e n t a d v a n c e s in t h e i m m u n o l o g y of t u b e r c u l o s i s h a v e l e d to e x p e c t a t i o n s of n e w a n d b e t t e r v a c c i n e s . But u n l e s s s o m e s y s t e m to e v a l u a t e b o t h c u r r e n t l y a v a i l a b l e a n d n e w v a c c i n e s among humans is f o u n d , w e w i l l b e n o b e t t e r off t o m o r r o w t h a n w e a r e t o d a y . Supported in part by Research Grant CA 47503 from the National Cancer Institute and Research Career Award HL 21670 from the National Heart, Lung and Blood Institute.

REFERENCES 1. Koch R: Die Aetiologie der Tuberkulose. Berlin Klin Wochenschr 19:221-230, 1882 (For an English translation, see Pinner B, Pinner M: A m Rev Tuberc 25:285323, 1932) 2. Koch R: Weitere Mitteilungen fiber ein Heilmittel gegen Tuberculose. Dtsch Med Wochenschr 16:1029-1032, 1890 3. Edwards PQ, Edwards LB: Story of the tuberculin test from an epidemiologic viewpoint. A m Rev Respir Dis 81 (no 1, pt 2):1-47, 1960 4. Dubos R, Dubos J: The White Plague: Tuberculosis, Man and Society. Little, Brown, Boston, 1952 5. Zinsser H, Enders JF, Fothergill LD: Immunity: Principles and Application in Medicine and Public Health. Macmillan, N e w York, 1939 6. Weiss DW: Vaccination against tuberculosis with nonliving vaccines. I. The problem and its historical background. A m Rev Respir Dis 80:340-468, 495-509, 676-688, 1959 7. Tuberculosis Program, Public Health Service, USA: Experimental studies of vaccination, allergy, and immunity in tuberculosis. 3. Effect of killed BCG vaccine. Bull W H O 12:47-62, 1955 8. Opie EL, Flahiff EW, Smith HH: Protective inoculation against human tuberculosis with heat-killed tubercle bacilli. A m J Hyg 29 (Sect B):155-164, 1939 9. Wells CS, Flahiff EW, Smith HH: Results obtained in man with the use of a vaccine of heat-killed tubercle bacilli. A m J Hyg 40:116-126, 1944 10. Irvine KN: B.C.G. Vaccination in Theory and Practice. Charles C Thomas, Springfield, IL, 1949 11. Medical Research Council: B.C.G. and vole bacillus vaccines in the prevention of tuberculosis in adolescents. Br Med J 1:413-427, 1956 12. Pinner M: Pulmonary Tuberculosis in the Adult: Its Fundamental Aspects. Charles C Thomas, Springfield, IL, 1945 13. Gu6rin C: The history of BCG. In Rosenthal SR, ed. BCG Vaccine: TuberculosisCancer. PSG, Littleton, MA, 1980 14. Sakula A: BCG: Who were Calmette and Gu6rin? Thorax 38:806-812, 1983 15. Grange JM, Gibson J, Osborn TW, Collins CH, Yates MD: What is BCG? Tubercle 64:129-139, 1983 16. Weill-Hall6 B, Turpin R: Premiers essais de vaccination antituberculeuse de l'enfant par le bacille Calmette-Gu6rin (BCG). Bull M6m Soc Med H o p Paris 49:1589-1601, 1925 17. S~ienz A: La i m m u n i d a d antituberculosa y la vacuanaci6n con el B.C.G. Rev Tuberc Habana 9:5-23, 1945 18. Greenwood M: Professor Calmette's statistical study of B.C.G. vaccination. Br Med J 1:793-795, 1928 19. Anderson AS, Dickey LB, Durfee ML, Farber SM, Jordan LS, Jordan KB, Kupka E, Lees HD, Levine ER, McKinlay CA, Marshall MS, Meyerding EA, Myers JA,

Tuberculosis Vaccine Trials

20. 21. 22. 23. 24. 25. 26. 27. 28.

29. 30. 31. 32. 33.

34. 35.

36. 37. 38. 39. 40.

41. 42. 43.

273

Ornstein GG, Slater SA, Stoesser AV, Sweany HC: The case against B.C.G. Br Med J 1:1423-1430, 1959 Kraus R: Present knowledge of the innocuousness of BCG. Am Rev Tuberc 24:778-784, 1931 Wilson GS: The Hazards of Vaccination. Athlone Press, London, 1967 Anonymous (Berlin correspondent). The L6beck trial. Lancet 2:1038, 1931 Heimbeck J: BCG vaccination of nurses. Tubercle 29:84-88, 1948 Hyge TV: The efficacy of BCG-vaccination. Tuberculosis epidemic in a state school with an observation period of 12 years. Danish Med Bull 4:13-15, 1957 Hilleboe HE: Controversial issues in tuberculosis control. Public Health Rep 61:1561-1563, 1946 International Tuberculosis Campaign: Final report of The International Tuberculosis Campaign, July 1, 1948-June 30, 1951. UNICEF, Geneva, 1951 Hilleboe HE, Morgan RH: Mass Radiography of the Chest. Year Book, Chicago, 1945 Anderson RJ: Rationale and results. In Community-wide Chest X-ray Survey, Public Health Service Publication No. 222. United States Government Printing Office, Washington, DC, 1-19, 1952 Amberson JB, McMahon BT, Pinner M: A clinical trial of sanocrysin in pulmonary tuberculosis. Am Rev Tuberc 24:401-35, 1931 Aronson JD, Dannenberg AM: Effect of vaccination with BCG on tuberculosis in infancy and in childhood. Am J Dis Child 50:1117-1130, 1935 Levine MI, Sackett MF: Results of BCG immunization in New York City. Am Rev Tuberc 53:517-532, 1946 Ferguson RG, Simes AB: BCG vaccination of Indian infants in Saskatchewan. Tubercle 30:5-11, 1949 Sergent E, Catanei A, Ducros-Rougebief H: Pr6munition antituberculeuse par le BCG. Campagne contr6116e poursuivie a Alger depuis 1935. Troisi6me note. Arch Inst Pasteur d'Alg6rie 38:131-137, 1960 Aronson JD, Aronson CF, Taylor HC: A twenty-year appraisal of BCG vaccination in the control of tuberculosis. Arch Intern Med 10i:881-93, 1958 Rosenthal SR, Loewinsohn E, Graham ML, Liveright D, Thorne MG, Johnson V: BCG vaccination against tuberculosis in Chicago. A twenty-year study statistically analyzed. Pediatrics 28:622-641, 1961 Aronson JD: Protective vaccination against tuberculosis with special reference to BCG vaccination. Am Rev Tuberc 58:275-276, 1948 Anonymous; Report of a conference on BCG vaccination. Public Health Rep 62:346-350, 1947 Palmer CE: Prospectus of research in mass BCG vaccination. Public Health Rep 64:1250-1261, 1949 Rosenthal SR, Loewinsohn E, Graham MI, Liveright D, Thorne MG, Johnson V: BCG vaccination in tuberculous households. Am Rev Respir Dis 84:690-704, 1961 Comstock GW, Webster RG: Tuberculosis studies in Muscogee County, Georgia. VII. A twenty-year evaluation of BCG vaccination in a school population. Am Rev Respir Dis 100:839-845, 1969 Bettag OL, Kaluzny AA, Morse D, Radner DB: BCG study at a state school for mentally retarded. Dis Chest 45:503-507, 1964 Comstock GW, Livesay VT, Woolpert SF: Evaluation of BCG vaccination among Puerto Rican children. Am J Public Health 64:283-291, 1974 Comstock GW, Woolpert SF, Livesay VT: Tuberculosis studies in Muscogee

274

G.W. Comstock

44. 45.

46.

47. 48. 49.

50. 51. 52. 53.

54.

55. 56. 57. 58.

59. 60. 61.

62.

63. 64.

County, Georgia. Twenty-year evaluation of a community trial of BCG vaccination. Public Health Rep 91:276-280, 1976 Palmer CE, Shaw LW: Present status of BCG studies. A m Rev Tuberc 68:462-466, 1953 Hart PD, Sutherland I: BCG and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life. Final report to the Medical Research Council. Brit M J 2:293-295, 1977 Frimodt-Mctller J, Acharyulu GS, Kesava Pillai K: Observations on the protective effect of BCG vaccination in a South Indian rural population: fourth report. Bull Int Un Tuberc 48:40-50, 1973 Paul R: The effects of vole bacillus vaccination of African mine workers in the Northern Rhodesian copper mines. Br J Ind Med 18:148-152, 1961 Coetzee AM, Berjak J: B.C.G. in the prevention of tuberculosis in an adult population. Proc Mine Medical Officers' Assoc 48:41-53, 1968 Vandiviere HM, Dworski M, Melvin IG, Watson KA, Begley J: Efficacy of Bacillus Calmette-Gu6rin and isoniazid-resistant Bacillus Calmette-Gu6rin with and without isoniazid chemoprophylaxis from day of vaccination. II. Field trial in man. A m Rev Respir Dis 108:301-313, 1973 Tripathy SP: Fifteen-year follow-up of the Indian BCG prevention trial. Bull Int Un Tuberc Lung Dis 62(3):69-72, 1987 Anderson RJ, Palmer CE: BCG. 143:1048-1051, 1950 Osborn TW: Changes in BCG strains. Tubercle 64:1-13, 1983 Dubos RJ, Pierce CH: Differential characteristics in vitro and in vivo of several substrains of BCG. I. Multiplication and survival in vitro. A m Rev Tuberc 74:655666, 1956 Pierce CH, Dubos RJ: Differential characteristics in vitro and in vivo of several substrains of BCG. II. Morphological characteristics in vitro and in vivo. A m Rev Tuberc 74:667-82, 1956 World Health Organization: Co-operative studies of methods for assay of BCG vaccine. Bull W H O 31:183-221, 1964 Hesselberg I: Drug resistance in the S w e d i s h / N o r w e g i a n BCG strain. Bull W H O 46:503-507, 1972 Collins DM, DeLisle GW: BCG identification by D N A restriction fragment patterns. J Gen Microbiol 133:1431-1443, 1987 Pierce CH, Dubos RJ, Schaefer WB: Differential characteristics in vitro and in vivo of several substrains of BCG. IlL Multiplication and survival in vivo. A m Rev Tuberc 74:619-717, 1956 Bunch-Christensen K, Ladefoged A, Guld J: The virulence of some strains of BCG for golden hamsters. Bull W H O 39:821-828, 1968 Sher NA, Chaparas SD, Pearson J, Chirigos M: Virulence of six strains of Mycobacterium bovis (BCG) in mice. Infect Immun 8:736-42, 1973 Dubos RD, Pierce CH: Differential characteristics in vitro and in vivo of several substrains of BCG. IV. Immunizing effectiveness. A m Rev Tuberc 74:699-717, 1956 Frennette M, Portelance V, Beaudet R: Comparison of the in vivo multiplication of 10 BCG substrains in three strains of mice: No correlation with their antitumour activity. Microbios 36:145-146, 1983 Ladefoged A, Bunch-Christensen K, Guld J: The protective effect in bank voles of some strains of BCG. Bull W H O 43:71-90, 1970 Smith D, Harding G, Chan J, Edwards M, Hank J, Muller D, Sobhi F: Potency of 10 BCG vaccines as evaluated by their influence on the bacillemic phase of experimental airborne tuberculosis in guinea pigs. J Biol Stand 7:179-197, 1979

Tuberculosis Vaccine Trials

275

65. Bartlett GL, Kreider JW, Purnell DM, Katsilas DC: Augmentation of immunity to line 10 hepatoma by BCG. Cancer 46:488-496, 1980 66. Jacox RF, Meade GM: Variation in the duration of tuberculin skin sensitivity produced by two strains of BCG. Am Rev Tuberc 60:541-546, 1949 67. Vallishayee RS, Shashidhara AN, Bunch-Christensen K, Guld J: Tuberculin sensitivity and skin lesions in children after vaccination with 11 different BCG strains. Bull W H O 51:489-494, 1974 68. Li H, Ulstrup JC, Jonassen TO, Melby K, Nagai S, Harboe M: Evidence for absence of the MPB64 gene in some substrains of Mycobacterium bovis BCG. Infect Immun 61:1730-1734, 1993 69. Irvine KN: BCG and Vole Vaccination: A Practical Handbook. National Association for the Prevention of Tuberculosis, London, pp 18-19, 1954 70. Birkhaug K, Darricarrere R: Effect of time and temperature on antigenic potency of BCG vaccine. Am Rev Tuberc 68:96-102, 1953 71. Medical Research Council: BCG and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life. Fourth report to the Medical Research Council by its Tuberculosis Vaccines Clinical Trials Committee. Bull W H O 46:371-385, 1972 72. Clemens JD, Chuong JJH, Feinstein AR: The BCG controversy. A methodological and statistical reappraisal. JAMA 249:2362-2369, 1983 73. Levine MI: BCG (letter). Am Rev Tuberc 58:581-582, 1948 74. Orenstein WA, Bernier RH, Hinman AR: Assessing vaccine efficacy in the field. Further observations. Epidemiol Rev 10:212-241, 1988 75. Smith PG: Retrospective assessment of the effectiveness of BCG vaccination against tuberculosis using the case-control method. Tubercle 63:23-35, 1982 76. Palmer CE, Shaw LW, Comstock GW: Community trials of BCG vaccination. Am Rev Tuberc 77:877-907, 1958 77. Palmer CE, Long MW: Effects of infection with atypical mycobacteria on BCG vaccination and tuberculosis. Am Rev Respir Dis 94:553-568, 1966 78. Wiegeshaus EH, Smith DW: Evaluation of the protective potency of new tuberculosis vaccines. Rev Infect Dis 11(Suppl 2):$484-$490, 1989 79. Tuberculosis Prevention Trial, Madras: Trial of BCG vaccines in South India for tuberculosis prevention. Indian J Med Res 72(Suppl):l-74, 1980 80. Palmer CE, Edwards LB, H o p w o o d L, Edwards PQ: Experimental and epidemiologic basis for the interpretation of tuberculin sensitivity. J Pediatr 55:413-429, 1959 81. Fine PEM: The Kellersberger Memorial Lecture, 1985. The role of BCG in the control of leprosy. Ethiop Med J 23:179-191, 1985 82. Frimodt-M611er J, Thomas J, Parthasarathy R: Observations on the protective effect of BCG vaccination in a South Indian rural population. Bull WHO 30:545574, 1964 83. Hart PD, Pollock TM, Sutherland I: Assessment of the first results of the Medical Research Council's trial of tuberculosis vaccines in adolescents in Great Britain. Ad Tuberc Res 8:171-189, 1957 84. Hart PD, Sutherland I, Thomas J: The immunity conferred by effective BCG and vole bacillus vaccines, in relation to individual variations in induced tuberculin sensitivity and to technical variations in the vaccines. Tubercle 48:201-210, 1967 85. Rich AR: The Pathogenesis of Tuberculosis. Charles C Thomas, Springfield, IL, 1944 86. Comstock GW: Identification of an effective vaccine against tuberculosis. Am Rev Respir Dis 138:479-480, 1988

276

G.W. Comstock 87. Sedaghatian MR, Kardouni K: Tuberculin response in preterm infants after BCG vaccination at birth. Arch Dis Child 69:Suppl 309-311, 1993 88. Medical Research Council: B.C.G. and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life. Third report to the Medical Research Council by their Tuberculosis Vaccines Clinical trials Committee. Br Med J 1:973-978, 1963 89. Comstock GW: Correspondence: Community trials of BCG vaccination. Am Rev Respir Dis 81:932-933, 1960 90. Comstock GW: Letter to the editor: BCG vaccination. Tubercle 41:227-229, 1960 91. Stein SC, Aronson JD: The occurrence of pulmonary lesions in BCG-vaccinated and unvaccinated persons: Am Rev Tuberc 68:695-712, 1953 92. Sutherland I: BCG and vole bacillus vaccination in adolescence and mortality from leukemia. Stat Med 1:329-335, 1982 93. Snider DE, Comstock GW, Martinez I, Caras GJ: Efficacy of BCG vaccination in prevention of cancer: an update. JNCI 60:785-788, 1982 94. Kendrick MA, Cornstock GW: BCG vaccination and the subsequent development of cancer in humans. JNCI 66:431-437, 1981 95. Comstock GW, Cauthen GM: Epidemiology of tuberculosis. In Reichman LB, Hershfield ES, eds. Tuberculosis: A Comprehensive International Approach. Marcel Dekker, New York, 1993 96. Lotte A, Wasz-H6ckert O, Poisson N, Dumitrescu N, Verron M, Couvet E: BCG complications. Ad Tuberc Res 21:107-193, 1984 97. Greenland S, Frerichs RR: On measures and models for the effectiveness of vaccines and vaccination programs. Int J Epidemiol 17:456-463, 1988 98. Comstock GW: Vaccine evaluation by case-control or prospective studies. Am J Epidemiol 131:205-207, 1990 99. Wiegeshaus EH, Harding G, McMurray D, Grover AA, Smith DW: A co-operative evaluation of test systems used to assay tuberculosis vaccines. Bull WHO 45:543-550, 1971 100. Sutrisna B, Utomo P, Komalarini S, Swatriani S: Penelitian efektifitas vaksin BCG Dan beberapa faktor lainnya pada anak yang menderita TBC berat di 3 rumah sakit di Jakarta 1981-1982. Medika 9:143-50, 1983 101. Shapiro C, Cook N, Evans D, Willett W, Fajardo, I, Koch-Weser D, Bergonzoli G, Bolanos O, Guerrero R, Hennekens CH: A case-control study of BCG and childhood tuberculosis in Cali, Columbia. Int J Epidemiol 14:441-446, 1985 102. Selwyn PA, Hartel D, Lewis VA, Schonebaum EE, Vermund SH, Klein RS, Walker AT, Friedland GH: A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeficiency virus infection. N Engl J Med 320:545-550, 1989 103. Graham NMH, Nelson KE, Solomon L, Bonds M, Rizzo RT, Scavotto J, Astemborski J, Vlahov D: Prevalence of tuberculin positivity and skin test anergy in HIV-l-seropositive and -seronegative intravenous drug users. JAMA 267:369-373, 1992 104. Aronson JD, Parr EI, Saylor RM: BCG vaccine. Its preparation and the local reaction to its injection. Am Rev Tuberc 42:651-666, 1940 105. Aronson JD, Palmer CE: Experience with BCG vaccine in the control of tuberculosis among North American Indians. Public Health Rep 61:802-820, 1946 106. Shaw LW: Field studies on immunization against tuberculosis. I. Tuberculin allergy following BCG vaccination of school children in Muscogee County, Georgia. Public Health Rep 66:1415-1426, 1951