Tuberculosis and War. Группа авторов
As an example, a fairly complex picture emerges for the population of Berlin from 1910 through 1946, as shown in Figure 5. Between 1910 and 1920, there were 4 censuses, an unusually high frequency for taking a census. The extrapolations from one census to the next have thus high credibility. The first complexity starts after the census in October 1919 and the next census taken in June 1925. On October 1, 1920, the population size doubled when the surface of the newly designated city of Greater Berlin increased more than tenfold. This raises the question of how one might appropriately calculate a rate for the year 1920 when both the numerator and denominator changed during the year in a complex manner. Even if both the numerator (cases) and denominator (population) were correct, the appropriate technical approach is not easy to decide upon. The population characteristics of the old and new parts of the city are likely to be substantially different, probably more urban versus relatively less rural, judging from the population density (i.e., the same population size in less than a tenth of the surface in the old compared to more than 90% of the greatly expanded new city surface). The changed city definition was also likely to affect TB case rates, which often disproportionally affected urban dwellings. Thus, even if all numbers and calculations were correct, a decrease in the rates in the same city might result from a calculation from 1 year to the next, even if the problems remain unchanged. During the entire period of WWII, no census was taken in Berlin. Whether the population really increased during the war years, as graphically suggested here, and then actually dropped by more than one third at some time in 1945 before the August 1945 census may be anyone’s best guess. While there are good reasons to construct the estimates in this way, they remain estimates rather than measures. If the postulated mass exodus that led to this drop occurred, then how would the TB case rate have been determined; in other words, how would the numerator data (case numbers) have been obtained? The entire war period for Berlin poses almost insurmountable problems in obtaining both credible numerator and denominator data. While in other parts of Germany the changes might have been less abrupt than for the capital, a correct enumeration of both the incident TB cases and the population from which they arose likely created substantial problems throughout most of Germany. A large proportion of the male population was in the military service, and the military is often assumed or known to have a tendency to be rather discreet about the magnitude of the TB problem among its rank and file [54].
Fig. 5. Population data for Berlin, 1910–1946, obtained from census data (indicated by hollow circles) and extrapolated for intercensal years. The dashed line shows the political decision to expand the city of Berlin to “Greater Berlin” in 1920 [81].
Heaf reports on pulmonary TB cases that were “notified or otherwise known to local health authorities” for England and Wales from 1938 to 1941 [55]; he showed a decrease of 6.1% from 1938 to 1939 that was followed by an increase of 4.9% the following year, and of 8.7% from 1940 to 1941, which coincided with attacks by Germany towards the end of June 1940 and onwards. Otherwise reported morbidity data from the United Kingdom remain scarce in Tubercle, its main easily accessible publication. Heaf himself does not put much trust in notifications of TB cases: “until the incidence is found out by a comprehensive survey in various groups of the community, we can only rely on mortality figures as a guide to the rise or fall of the incidence in the general population” [55].
Morbidity data are scarce during much of WWII, but these considerations about the calculation of rates also apply to mortality. This will be addressed in more detail in the next chapter on ascertainment of TB deaths.
Death from M. tuberculosis
Case fatality designates the proportion of people with TB who succumb to it: thus, the numerator is deaths, and the denominator is TB cases among whom the deaths occurred. Tuberculosis mortality has the same numerator as fatality, but the denominator is the entire population (rather than just TB cases) from which the TB cases arose and it has an observation time (usually 1 year). These 3 components make the definition of a “rate” [56].
Tuberculosis case fatality from untreated sputum smear-positive TB was large in the pre-chemotherapy era. One of the earliest long-term observations from a sanatorium in Switzerland showed a cumulative fatality of 70% after 17 years of follow-up (Fig. 6) [57]. What was designated “open tuberculosis” are patients excreting tubercle bacilli, and is thus largely synonymous with sputum smear-positive TB, because cultures were rarely used in routine testing at that time. By contrast, “closed tuberculosis” referred to any other form of TB (both pulmonary and extrapulmonary). “Sputum” is also almost uniformly “direct,” which means spontaneously produced rather than induced or via bronchoscopy. Clearly, the “closed” form with 18% fatality was substantially less deadly than “open” (pulmonary) TB. Similarly, high fatality ratios were determined in Denmark (68% after 10 years) [58], in Sweden (84% after 18 years) [59], and in the United Kingdom (86% after 11 years) [60]. A meta-analysis put the weighted mean after 10 years at 70% [61]. To obtain a reasonable estimate of case fatality, a longer observation period improves the estimate, but it is not required. For instance, it is perfectly justified to have only an approximate observation time for each case in treatment outcome analysis. Therefore, it suffices to state that a certain number of all patients who should have completed treatment died prematurely from M. tuberculosis. According to the WHO/Union definition, those who died of other causes are added to this numerator. Without observation time, we then get a simple proportion rather than a rate.
Fig. 6. Fate of untreated pulmonary tuberculosis, long-term follow-up, Barmelweid Sanatorium, Switzerland, reproduced by permission from Springer [57].
The variation in case fatality at the end of long observation time is relatively large: 68–86% in the four pre-chemotherapy studies mentioned here. For better comparability of these studies, the observation end point was fixed at ten years as in the meta-analysis. The median is then very narrow: between 1.2 and 1.7 years (Fig. 7) [57–60]. In other words, about half of the deaths occur within the first 1.5 years, while the other half occurs over the next 8.5 years. As the final cumulative fatality is so high, it cannot get much higher under the influence of an adverse situation, such as wartime. What could change substantially was the speed at which patients died, as suspected by Rist [28]. Whether this hypothesis can be confirmed or has to be refuted will be examined shortly.
Fig. 7. Cumulative case fatality from untreated sputum smear-positive