Pathogenesis of tuberculosis: the 1930 Lübeck disaster revisited

Background

The chief individuals involved in the genesis of the 1930 Lübeck events were George Deycke, physician and director of Lübeck General Hospital, and Ernst Altstädt, chief tuberculosis treatment physician of Lübeck Internal Medicine Services. Both had extensive experience of tuberculosis management and treatment research and were respected by colleagues and the Lübeck community.

Deycke trained as a physician and worked initially at Eppendorf Hospital in Hamburg. He assisted in establishing a medical faculty in Istanbul in 1898 and was director and chief physician of the Royal Osman Guilane Hospital from 1903 to 1907 [5]. During 1908, he was director of leprosy services in British Guyana and evaluated a leprosy treatment based on a mycobacterial fat nastin, that appeared to ameliorate immune responses in some patients [6]. Returning to Hamburg, he worked at Eppendorf Hospital until 1913 when he was appointed director of Lübeck General Hospital.

Altstädt, a student of tuberculosis researcher Hans Much, trained in Hamburg and moved to Lübeck with Deycke in 1913. Deycke and Altstädt worked closely with Much developing immunological tuberculosis treatments based on eliciting antibodies to certain M. tuberculosis antigens known as partigens [7, 8]. Partigen treatment of tuberculosis was used for several decades with varying results [9–11]. Partigen-related research continued following Deycke's Lübeck appointment; a consequence of this was the presence of the M. tuberculosis “Kiel” strain in the Lübeck General Hospital laboratory [12].

In 1930, Albert Calmette summarised events preceding the first German BCG vaccination programme in the old Hanseatic town of Lübeck [13]. On 24 July 1929, the Lübeck Health Office, on behalf of Altstädt, requested a BCG culture from Institute Pasteur for vaccination among the socially uninsured Lübeck population. On 27 July 1929, the Institute Pasteur Vaccine Section dispatched BCG vaccine 374 to Lübeck. This strain was used in France in August 1929 to vaccinate 573 infants and also supplied to Mexico and Latvia; no adverse events were recorded. Calmette and Camille Guérin maintained the vaccine on glycerine–potato medium and used liquid Sauton's medium when preparing the vaccine. In contrast, Deycke used solid culture media during vaccine preparations undertaken by an assistant who had worked with Deycke for 17 years preparing partigens but was not professionally qualified for laboratory work [14]. Furthermore, disregarding Calmette's recommendations to minimise risk of contamination, the Lübeck laboratory utilised two partially separated areas; BCG cultures were incubated in the smaller area A and the Kiel strain was incubated in space B where BCG vaccine and partigens were prepared (figure 1). The incubators were not locked. There was no separate designated area for culture and vaccine reconstitution. Further, once prepared, the vaccine did not undergo animal evaluation confirming lack of virulence.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Photograph of the original bacterial incubator from the Lübeck city hospital (left; courtesy of Medizinhistorisches Museum Hamburg, Germany, Mr. Henrik Essler. Photo: Mrs. Claudia Ketels, Photo- und Graphik Department, University Hospital Hamburg Eppendorf, with permission) and a poster from the Lübeck public health authorities advocating for oral bacille Calmette–Guérin vaccination in the first days of life of newborn babies (right). One passage of the text states: “This means of protection is completely harmless, it does not result in any health disorders”.

The vaccination programme was approved by the Lübeck Health Council and Medical Practitioners Council. In late February 1930, Lübeck posters advertised the programme (figure 1) and newspapers urged public support. Although initially intended for infants from a tuberculosis milieu (persons living under conditions where transmission events of M. tuberculosis are likely to occur, e.g. under crowded housing conditions), vaccination was later offered to all newborns. Following Calmette's recommendations, the infants were to receive 10 mg of vaccine orally on three separate occasions within 10 days post-birth.

In March 1927, the German Government Health Advisory Council discussed BCG vaccination but considered the evidence insufficient to recommended BCG [15]. Nevertheless, BCG was occasionally used and several hundred children in Bleialf in Eifel, Breslau and Berlin were vaccinated [16, 17].

The Kiel strain

The Kiel strain was isolated in 1927 from a 4-year-old child with hip tuberculosis cared for by Hans Opitz in the Charité Hospital, Berlin. Initially known as H29 or “Werner”, Opitz studied this mycobacterium and considered it had M. bovis characteristics. Both Opitz and Bruno Lange of the Robert Koch Institute thought its virulence was markedly attenuated [18]; Lange commented that “Werner” was the same organism as the Kiel strain [19] that on 29 September 1929 was forwarded from Kiel to Deycke for partigen preparation [12, 20]. It had an unusual characteristic in that it frequently stained Sauton culture media green, facilitating its later identification.

The unfolding tragedy

The vaccination campaign began in late February 1930, but three infants were vaccinated earlier. On 12 December 1929, an infant whose mother had severe pulmonary tuberculosis and was sputum microscopy-positive for acid-fast bacilli received the first vaccine doses; it was later established that these doses were contaminated by the Kiel strain. This infant presented on 15 January 1930 with cervical lymphadenopathy. Biopsy culture revealed virulent human M. tuberculosis; unfortunately, it was assumed that the infant had congenitally acquired tuberculosis. Two more infants were vaccinated on 30 December 1929 and 10 February 1930.

On 17 April 1930, an infant vaccinated 33 days previously died; permission for a post mortem examination (PME) was refused. Another death occurred on 20 April 1930; on PME extensive respiratory tuberculosis was found with mesenteric nodal infiltration. Pulmonary tuberculosis was diagnosed as the mother had pulmonary tuberculosis. However, the death of two more vaccinated infants on 25 and 26 April 1930, both with severe abdominal tuberculosis, placed the matter beyond doubt [21]. The vaccine emulsions prepared in Lübeck were probably contaminated by virulent bacilli.

The response to the vaccination accident

As the nature of the tragedy became apparent, live-vaccine opponents claimed reversion of BCG to virulence [22]. Simultaneously, detailed documentation of events commenced, supported by the Ministry of Internal Affairs coordinated by Moegling of the Health Ministry, Bruno Lange and Ludwig Lange from the Robert Koch Institute were responsible for microbiological aspects, and PMEs were undertaken or supervised by Paul Schürmann from Berlin.

The course of events following the initiation of the BCG vaccination programme is summarised in figure 2. Of 412 infants born in Lübeck, 251 (60.9%) received BCG; 77 (30.7%) vaccinated infants died, but 10 from causes other than tuberculosis. A PME was carried out on 72 (93.5%) infants; the parents refused permission in five instances. A 2-year-old boy also inadvertently ingested vaccine while accompanying his mother and newborn brother for vaccination. He became tuberculin skin test (TST)-positive and later mesenteric calcifications were identified on abdominal radiography [23].

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

The disposition of infants born in Lübeck, Germany, during the 1930 bacille Calmette–Guérin (BCG) vaccination programme. Clin.: clinical; Radiol: radiological; Mtb: Mycobacterium tuberculosis; PME: post mortem examination; TB: tuberculosis, TST: tuberculin skin test; XRC: chest radiograph. ^#: PME refused by five parents, assessment by history, clinical and radiology findings.

On learning of the vaccination accident, Calmette systematically described all means by which BCG contamination could have occurred and precautions taken in the Pasteur Institute [13]. Considerable numbers of infants were already BCG-vaccinated (in France alone 242,250) without complications or reversion to pathogenicity.

Documentation of the tragedy included recording details of the consequences of probable infection of 251 newborn infants with uncertain amounts of M. tuberculosis. A card system recorded every event, and the infants were regularly evaluated clinically, chest and abdominal radiographs taken, and TST sensitivity documented. Hans Kleinschmidt from Cologne supervised clinical evaluation and his assistant Mina Böcker was responsible for TST [24]. Hermann Jänasch and Gertrud Remé coordinated radiological evaluations [23].

Vaccine contamination

Detailed microbiological investigations, including guinea pig experiments, rejected BCG reversion to pathogenicity [12, 25]; M. tuberculosis “Kiel” strain contamination of some BCG vials was confirmed [12]. By 14 July 1930, deaths appeared to segregate into four periods [21]. Of 159 infants immunised between 9 December 1929 and 27 March 1930 and 8–12 April 1930, 53 (33.3%) had died. Notably, there were as yet no deaths among 91 infants immunised between 28 March and 7 April 1930 and 14–25 April 1930; six deaths eventually occurred in the latter groups (six among 91 (6.6%) infants), suggesting that vaccine contamination had occurred even during relatively safe periods when fewer infants died.

A detailed, wide-ranging analysis taking into account death, but also disease severity and infection risk, classified clinical outcomes and BCG contamination into a range of steps of disease severity. Disease severity and death appeared associated with probable ingested dose of virulent bacilli, but it remained difficult to delineate clear-cut differences in disease occurrence and specific vaccination periods in individual infants [1].

Epidemiological aspects of the Lübeck disaster

The first death of a partially vaccinated infant, at age 5 days as a result of prematurity, occurred on 31 March 1930. In April 1930, six infants died as result of tuberculosis. By end of June 1930, 47 infants had died and worse was feared. Intensive surveillance was planned until 1932–1933. However, in July 1930 there were 16 deaths and in August there were eight deaths. Thereafter, in one of the most striking observations resulting from the tragedy, the infant mortality rate declined dramatically; only six further tuberculosis-related deaths occurred from September to December 1930. No additional tuberculosis deaths occurred throughout the 5-year post-vaccination observation period. Viewed in terms of infant age, by completion of age 3 months 50% of eventual deaths had occurred and 87.5% by completion of age 5 months. No relapses or new tuberculosis manifestations occurred among survivors neither following a year after vaccination nor at time of publication of the final report in 1935. In 1942, Kleinschmidt followed up 122 survivors and had reliable information regarding another eight children; neither further tuberculosis-attributed deaths nor recurrences were observed [26]. The occurrence of death and the numbers of children presenting in Moegling's disease categories during the 3 years post-tragedy are summarised in table 1.

After careful reading of the pathology and microbiology reports, we revised the number of tuberculosis-attributed deaths from 72 [27] to 67; 10 deceased infants did not die of tuberculosis. Two neonates died of neonatal complications; the remaining eight died following accidents (two infants), other infectious causes and in one infant pylorospasma.

In 39 (15.5%) infants, no TST evaluation was undertaken. The remaining 212 (84.5%) infants were evaluated by percutaneous TST with Moro or Ektabin reagents. Overall, 190 (89.6%) responded positively to one of these TSTs; among the 22 (10.4%) with negative results, six infants were in disease severity group “6”. Böcker commented that the remaining TST-negative infants may not have been infected with virulent bacilli. Interpretation of these results is further complicated by the fact that there is no doubt that oral BCG vaccination will lead to TST-positivity in a proportion of neonatally vaccinated infants and could also cause necrosis in lymph-node tissues and primary foci with subsequent calcification [28, 29]. It could therefore be questioned whether all the infants, with one exception, as claimed by Moegling, were fed contaminated vaccine.

No long-term complications were reported with the exception of six children with varying degrees of deafness following tuberculous otitis media. Further plans for intensive surveillance were consequently shelved.

Anatomical pathology findings

Twelve early PMEs, before the nature of events was fully appreciated, were performed by local medical personnel and these results and clinical findings were reviewed by Schürmann; he was solely responsible for 53 PMEs and assisted at another seven [30]. Schürmann preferred reporting detailed findings of only PMEs in which he participated, consequently reports frequently refer to varying PME numbers providing only percentages regarding some findings.

Primary foci

Given the gastrointestinal infection route, it was expected that almost any site accessible from the gastrointestinal canalicular system might have primary foci. The sites of primary foci among 60 infants for whose PME Schürmann was responsible are summarised in figure 3. The four main sites were the small bowel, the stomach and oesophagus, the oropharynx, and the lungs; multiple foci were common. In 50% of cases there were two loci, in 29.9% three, in 18.3% a single locus and in one infant (1.7%) there was a focus in each of the four loci.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3

The site of primary tuberculosis foci and secondary lesions in 60 Lübeck infants who died following ingestion of bacille Calmette–Guérin vaccine contaminated by Mycobacterium tuberculosis within 10 days of birth and closely examined during post mortem examination [30]. Foci were found: 1) in the small bowel in 59 (98.3%) infants and the stomach or oesophagus in 11 (18.3%); 2) in the oropharynx in 47 (78.3) infants; 3) in the tympanic cavity in six (10%) infants; and 4) in the lungs in 15 (20.8%) of 72 infants. 5) Secondary lesions arising as result of lympho–haematogenous spread were found in the brain in 34 (47.2%) of 72 Lübeck infants, of whom 19 (26.4%) had tuberculous meningitis at death.

Mesenteric nodal changes were identified by Schürmann in all 60 Lübeck infants in whose PME he participated and were accompanied by mucosal primary foci in 59 (98.3%). These changes varied from 70 ulcers to single foci only microscopically visible and were concentrated close to the ileocaecal valves within Peyer's patches. Mesenteric nodes were often greatly enlarged and already healing and calcifying soon post-infection; the earliest calcification being noted 140 days post-infection. In only one child with limited mesenteric nodal tuberculosis changes were no primary intestinal primary lesions found despite serial sections over 0.75 m of intestine.

Even 44 days post-infection, fresh, deep intestinal wall ulcers might still be visible, but fibrotic changes were already developing; in older children who died 91–128 days post-infection, signs of healing were frequent and gut wall tuberculosis changes of primary foci more difficult to find. No primary foci were found distal to the ileocaecal valves. In older infants, secondary caecal or rectal ulcers were frequent, but unaccompanied by typical primary lymph node lesions.

The oesophagus, stomach and duodenum are usually considered relatively immune to tuberculosis and their frequent involvement was unexpected. The duodenum did appear relatively resistant to tuberculosis and a primary focus was present in only three (5.0%) of 60 infants. However, among 27 infants with tuberculosis lesions in the stomach, 11 (40.7%) lesions associated with regional nodal tuberculosis infection were clearly primary. Vomiting and retching causing mucosal tears might have facilitated establishment of these foci.

Similarly, oesophageal tuberculosis lesions were unexpectedly present in 11 (19.3%) of 57 infants; in three, regional lymph node involvement supported a primary focus diagnosis. As in the stomach, these lesions were ascribed to overwhelming infectious doses and mucosal damage following retching and vomiting.

The oropharynx was the second commonest site of primary foci. Tuberculous cervical lymph nodes were found in 52 (89.7%) of 58 infants on detailed PME. In six (11.6%), the features were those of post-primary tuberculosis, but in 46 (88.5%) they were part of a primary complex. The commonest sites of primary foci were the oropharyngeal tonsils where tuberculous changes were observed in 42 (80.7%) infants and the palatine tonsils with changes in 29 (55.7%). Changes were typical of a primary focus in 26 (50%) and 22 (75.9%) infants in the oropharyngeal and palatine tonsils, respectively. In one instance, the nucleus of a tooth harboured a primary focus. All oropharyngeal primary lesions were ulcerous, frequently involving tonsils, but oropharyngeal healing occurred rapidly and often leaving no macroscopically visible signs.

In two instances, death was accidental and oropharyngeal primary lesions were limited to small, calcified, deep cervical nodes found only on histological examination. One of these infants died 91 days post-infection and intestinal mucosal lesions were also found but were visible only on histology; however, in the mesenteric nodes, macroscopic, confluent, caseating, fibrous-enclosed lesions were present with scattered, scarce tubercles in the spleen; BCG was cultured from gastric-wash specimens. The second child died 498 days post-infection. All tissue cultures were negative and no intestinal wall lesions were noted, but several small centrally calcified caseous mesenteric nodes were found.

Complex anastomoses involving all cervical nodes and lymph vessels were frequent, complicating identification of primary infection sites. Most often deep cervical nodes but especially the upper group were affected. Frequently affected nodes were bilateral, irrespective of the primary focus site. When lower cervical nodes were involved, there were often anastomoses with axillary nodes. Schürmann emphasised that nodes should be examined histologically; a tuberculous complex might be recognised only by histological examination.

Otitis media tuberculosis was present in 21 (29.2%) of 72 infants of whom nine (42.9%) died. An important potential consequence was petrous bone involvement leading to facial nerve palsy. This was found on PME in five children with tuberculous otitis media and was associated with a poor prognosis. Of nine children with tuberculous otitis media who died, five died as result of disseminated tuberculosis and four as result of tuberculous meningitis.

A primary pulmonary complex was present in 15 (20.8%) of 72 infants and more than one focus was found in five infants. Extensive thoracic lymph node enlargement and caseation was present in every case and the intestines and oropharynx had regional primary nodes at similar developmental stages suggesting simultaneous establishment of these foci. Frequently, the focus was clearly associated with a bronchus and caseous bronchitis often caused obstruction. Occasionally, main bronchi, and once the trachea, were compressed by surrounding nodes. The fluid nature of the infectious agent and spoon feeding probably contributed to the many primary lung foci in these very young infants. In six infants, their noses were held during vaccine administration; four vomited after all three vaccine doses. The finding of pulmonary primary lesions in 20% of deceased infants related to bronchogenic aspiration of ingested material is thus not surprising.

In two children, late secondary pneumonia with cavitation developed and some cavities contained stomach contents. Gastrointestinal stasis and reflux or vomiting associated with peritonitis, ileus or meningitis leading to aspiration probably caused these lesions. The absence of pulmonary nodal involvement indicated their secondary nature.

Among surviving infants, a pulmonary primary lesion was infrequent and detected on chest radiology in only 11 (6.4%) of 173-survivors.

Lympho-haematogenous M. tuberculosis dissemination

Generalised dissemination of M. tuberculosis outside the primary focal entry point was prominent. Schürmann examined 54 infants closely for features of lymphohaematogenous dissemination and reported such features in more than 60% of these infants involving lymph nodes, the spleen, the lungs, the liver, the bone marrow, the brain and its membranes, and the kidneys. In a minority of infants, the adrenal glands, the thyroid, the skin, the pancreas and hypophysis were infiltrated. The simultaneous involvement of multiple organs involving more than five systems was common and considerable variation in focal size of lesions suggested intermittent, repeated showers of bacilli, a condition best described as protracted haematogenous tuberculosis [31, 32]. In 23 (42.6%) infants, early generalised dissemination was sufficiently severe to be alone identified as cause of death 44–127 days post-infection.

Closely associated with lymphohaematogenous dissemination, tuberculous meningitis was a major cause of death; 19 (35.2%) infants had tuberculous meningitis at death, but in two the primary cause of death was respectively perforation peritonitis and bleeding from an intestinal ulcer. On PME, inconspicuous solitary foci, frequently caseous, or enclosing fibrosis or calcification were found lying in the brain substance, the choroid plexus or meninges or simultaneously in various stages of development or regression in several sites. In a further 15 infants, similar central nervous system lesions were noted but were not associated with development of meningitis. In some of these latter infants, calcification of the intracranial lesions was already developing.

Among the 19 children with tuberculous meningitis, frequent inflammation across the brain base and choroid plexus might have suggested that it was the main point of penetration into the cerebrospinal fluid (CSF), but Schürmann considered that despite the overwhelming nature of these lesions, the solitary caseous foci were more relevant for tuberculous meningitis pathogenesis and localisation of the majority of lesions at the brain base and in the choroid plexus was determined by CSF flow from an entry point at other sites.

The first symptoms of meningitis among the 19 children with tuberculous meningitis occurred 70–310 days post-infection while death occurred between 72–313 days post-infection.

The immediate causes of death in the Lübeck infants

Based on the PME findings in 72 Lübeck infants, lymphohaematogenous-disseminated tuberculosis was considered sufficiently severe to have alone caused death in 24 (33.3%) of the infants. Given the known susceptibility of infants to haematogenous spread of M. tuberculosis and its consequences this is not surprising [33, 34]. Seventeen (23.6%) infants died as result of tuberculous meningitis and a further two had meningitis at death, but the final cause of death was ileus and gastrointestinal haemorrhage, respectively. Gastrointestinal complications and their consequences, such as ileus, perforation peritonitis, intestinal haemorrhage and resulting anaemia and their consequences, difficult to diagnose accurately and manage with facilities of the time, caused death in 24 (33.3%) infants. The speed with which dissemination caused death is notable; the first death following widespread infection dissemination occurred 44 days post-infection; after 127 days, there were no further deaths directly attributable to dissemination.

Radiological findings among survivors

Among the 173 surviving infants, only 11 (6.4%) had specific pulmonary tuberculosis radiological features that were typical of a primary lung complex [23]. In four cases, these changes were the sole abnormal radiological finding [23]. During the first 6 months post-infection, infiltrates, probably representing mild dissemination, appeared among a minority of surviving infants but regressed spontaneously. It was noteworthy that features of a primary complex were seen only in the first 2–3 months of life or when calcification appeared. Among surviving infants, abdominal radiographs 4 years post-vaccination were available from 173 (98.9%); in 126 (73%), calcifications were present, in 40 (31.7%) these were extensive, but in 86 (68.3%) they were slight. In only two cases was splenic calcification visible; there were no hepatic calcifications. Despite extensive abdominal calcifications among survivors, there were few associated clinical symptoms.

The trial

Deycke and Alstaedt were charged with negligent manslaughter and the resulting trial lasted from October 1931 until February 1932. Its procedures were closely followed, not only in Germany, but also internationally. Both physicians were found guilty. Deycke was sentenced to 2 years and Altstaedt to 15 months imprisonment, respectively. It was found that Deycke had not followed the procedures recommended by Calmette in preparing BCG and failed to evaluate the prepared vaccine in animal hosts. The vaccine was also prepared in an inadequate laboratory where cultures of pathogenic mycobacteria were also maintained. One of the lessons learned was the necessity of a rigorous adherence to laboratory protocols. In this aspect the Lübeck disaster had an important effect on the development of scientific rigor and manufacturing protocols.

The tragical events in Lübeck also resulted in misgivings regarding the irresponsible and unethical manner in which certain aspects of medical research were carried out [35] and they continue to be part of an ongoing debate about the ethical aspects of medicine [36, 37].