Answer by Tirumalai Kamala:
1. Prior exposure of trial population to Non-Tuberculous Mycobacteria (NTM) could explain non-effectiveness in two of the trials (1, 2).
2. The third trial, likely unwittingly but effectively, excluded individuals with prior NTM exposure (3).
To fully understand the differences between these discordant US BCG trials, we need to understand the history and details of some of the reagents they used, specifically the materials used in the. In a vaccine trial, how do we decide whom to vaccinate? Typically, we prefer to include only “naive” individuals, i.e. individuals not previously exposed to either the vaccine or disease causing organism. Premise is to be able to attribute trial results solely to the vaccine. This entails some type of pre-screening test to exclude those already exposed. In the case of TB, Mantoux skin tests are used, and responders above a certain cut-off are excluded. The standard Mantoux consists of intradermal injection of 5 Tuberculin Units (TU) on the volar surface of the arm (10). 48 to 72 hours later, the reaction is read as the size of the induration, the hard central nodule of the reaction, excluding the erythema, the redness around the nodule. Obviously, both injection and test reading are error prone. Even more perplexing, different trials used different TU lots and doses, and different cut-offs. Typically cut-off is 5mm at higher latitudes while tropical countries have 10 to 15mm cut-off. What is this and how did these metrics come into existence?
What is Tuberculin, the material injected into the skin to assess TB exposure?
In 1891, reported his discovery of Tuberculin as a cure for TB (4). His claim was disproved in short order but speedily recognized its potential value in TB diagnosis (5). This spurred further research into the topic, and Rich and McCordock made the next key observation (6), namely, individuals with small or moderate skin reactions to Tuberculin were less likely to develop TB disease compared to strong reactors. Thus, BCG trials typically pre-screen with Tuberculin and exclude strong reactors.
Koch‘s Tuberculin, later called Old Tuberculin, was eventually phased out of use because it gave too many variable skin test results. Consensus arose around the need to improve its preparation method to yield a more consistent product. Intensive research on Tuberculin preparation ensued, with, then a young biochemist at the University of Chicago, at the forefront.
New Tuberculins, the Purified Protein Derivatives (PPD)
Seibert precipitated Koch‘s Old Tuberculin with a variety of chemicals to derive products she called Purified Protein Derivative (PPD) (7). However, these products were neither pure nor entirely proteins, rather they are protein-polysaccharide mixtures of varying degrees (8).
Pure or not, in 1952, the WHO adopted a crude mixture prepared from heat-sterilized Mycobacterium tuberculosis culture filtrates, Seibert‘s lot 49608, also called PPD-S (S for Seibert), as the international reference standard for all PPDs (9). The International Unit (IU) for PPD is defined as the biological activity contained in 0.000028mg of PPD-S consisting of 0.00002mg of PPD and 0.000008mg of salts. 1 Tubeculin Unit (TU) is thus defined as 0.00002mg of PPD-S (9). The WHO adopted 5TU as the standard screening dose of PPD from the practical necessity of using a single concentration, and also because almost everyone reacts to Tuberculin if given large enough doses (9). So much for Mantoux skin test specificity! The US BCG trials we discuss here pre-date these guidelines.
Two US BCG trials showed little or not efficacy
initiated and/or was involved in two US BCG trials (1, 2). Both trials pre-screened with a Mantoux skin test using the same PPD batch, lot RT19-20-21.
One trial (1) was conducted in 1950 in Muscogee County (and Russell County, Alabama) using BCG Tice on a total of 64136 persons over the age of 5 years (16913 vaccinees, 17584 controls, 29369 Mantoux positive reactors). Pulmonary TB rates were similar between vaccinees and controls at a 20 year follow-up.
Another trial (2) was conducted from 1949 to 1951 in Puerto Rico using BCG Birkhaug on 191827 children. Pulmonary TB rates were similar between vaccinees and controls over a 18 to 20 year follow-up.
The many differences (population ages, BCG sub-strains, TU used for skin test, and test cut-off size) between these two trials don’t matter much since neither trial showed BCG protection against pulmonary TB. What else? Something about the trial location? The Georgia trial had almost as many pre-trial PPD responders (29369) as not (16913+17584 = 34497) as not, suggesting strong mycobacterial exposure among the trial population. Could NTM prevalence at these trial sites explain the data? Two kinds of data sets support this idea.
1. In later studies, individuals at some of these trial sites responded more strongly to skin tests with PPDs made from NTM compared to TBPPD (PPD-S).
2. More NTM were found at some of these trial sites compared to elsewhere.
Location Does Matter
I. Skin Test Responses to NTM in one BCG trial location (South-Eastern US) are stronger compared to elsewhere
In the 1950s, two microbiologists/epidemiologists, Lydia B. Edwards and Carroll E. Palmer were trying to develop a diagnostic tool for atypical (non-TB) mycobacterial infections. Taking a leaf out of Koch-Siebert, they prepared PPDs from a number of mycobacterial species. They had the key insight that simultaneously testing with two or more PPDs, one from TB and the others from NTM, could differentially diagnose NTM versus TB exposure. How did they do this? They isolated NTM from nasal swabs of volunteers and dual tested them with TB or NTM PPD. The largest reaction was usually to the antigen prepared from the isolated NTM (11, 12). With this insight, from 1958 to 1969, Edwards and Palmer undertook an extensive study (13) of Mantoux skin test patterns to two PPDs, PPD-S and PPD-B, in a very large group of individuals (all male naval recruits numbering about 100000 per year). We already know PPD-S. They prepared PPD-B from an NTM they called the Battey strain. We now call it Mycobacterium avium. As the name suggests, it was first isolated from a bird. The Edwards-Palmer study incontrovertibly showed that place of origin predicted skin test response patterns. Strongest PPD-B skin test responses were from those who came from the South-Eastern US.
Edwards and Palmer also found organisms responsible for such sensitivity were not transmitted from human to human but were readily isolated from the environment (13). Palmer was also among the first to speculate that NTM could influence response to BCG (14). In 1974, in an extensive study of patients and employees in hospitals spanning 8 US states (15), Wijsmuller and Erickson also found stronger skin test responses to PPD-B from people from the Southern US states. In 1970, Abrahams replicated this result in Queensland, Australia. He skin-tested school children with PPD-S and PPD-B. Those who originally responded more strongly to PPD-B retained same response pattern 4 years later, prompting him to coin the phrase Original Mycobacterial Sin, i.e. primary mycobacterial exposure imprints a stable (skin) immune response pattern (16).
Location Does Matter
II. Are there more NTM in Southern US compared to elsewhere?
NTM are found in mud, compost, wet soil, surface water, rivers, estuaries, piped water supplies, taps and shower heads (17). Some species are more hydrophobic and can readily aerosolize to sensitize people who inhale them (17). A study particularly relevant for the US BCG trials is part of a series of studies done by a group led by Joseph O. Falkinham, III, at the University of Virginia Tech, Blacksburg, VA. Following up on the Edwards-Palmer skin test study (13), Falkinham’s group compared NTM distribution in South-Eastern and North-Eastern US soil and water. They found NTM species were more abundant in the South (18, 19).
NTM prevalence and exposure could explain the Georgia BCG trial. What about the Puerto Rico trial? Here the clue comes from examining the war of words (20) that broke out between Comstock and Sol Roy Rosenthal, another prominent US TB researcher, when these trial results were published. The key text from Rosenthal is, “Puerto Rico has a high incidence of atypical mycobacterial infection“. Today we call Rosenthal‘s atypical mycobacteria NTM. Thus, as with Georgia, we can safely infer high NTM prevalence in Puerto Rico as well.
Is BCG ineffective against TB in the US? No, it worked in the earliest US BCG trial.
The last US BCG trial (3) I discuss was actually the first one undertaken in the US. From December 1935 to February 1938, 3025 American Indians and Alaskan natives with normal chest radiographs and skin test response to 250TU of PPD (yes, 250TU!) got one dose of intracutaneous dose of BCG in a placebo-controlled trial. Study area included southeast Alaska, Arizona, North Dakota, South Dakota, and Wyoming. They used early BCG strains obtained directly or indirectly from Calmette or Guerin. In a 1992 to 1998 follow-up, they found a vaccine efficacy of 52%. Not bad for a 50 to 60 year follow-up. How to explain this baffling result? By pre-screening with an extremely high (250TU) PPD dose, they likely excluded from vaccination those with prior NTM exposure. I interpret this to suggest that even with a sub-optimal (skin) route, BCG can protect against TB when given to people previously unexposed to mycobacteria.
5. von Pirquet, C. Verlauf der tuberkulose allergie bei einem fälle von masern und miliar tuberkulose. Wien Klin. Wschr., 1908; 21: 861-865.
6. Arnold Rice Rich, and Howard A. McCordock. “An enquiry concerning the role of allergy, immunity and other factors of importance in the pathogenesis of human tuberculosis.” Bull. Johns Hopkins Hosp 44.273 (1929): 228.
7. Seibert, Florence B. “The isolation and properties of the purified protein derivative of tuberculin.” Am. Rev. Tuberc 30 (1934): 713-720.
12. Palmer, C. E., and L. B. Edwards. “Geographic variations in the prevalence of sensitivity to tuberculin (PPD-S) and to the Battey antigen (PPD-B) throughout the United States.” Bull Int Un Tuberc 32.July (1962): 373-383.