in four studies from Denmark [58], Sweden [59], Switzerland [57], and the United Kingdom [60], with permission from Elsevier.
The determination of case fatality poses no problems with the denominator (patients with TB) as the cases are known and all that is needed is to learn about their outcome (e.g., the number of deaths among the cases). In contrast, the determination of mortality can be complex as it requires knowledge of both the correct numerator and denominator, that is, the accurate enumeration of deaths (numerator) and population (denominator) from which the cases arose as discussed above.
The famous compilation of data by Redeker about mortality in Germany (Fig. 4 in chapter 1), includes the years 1892–1940 [62]. In many ways, these are the most remarkable data. For one, there is a huge spike commencing shortly after the beginning of WWI in 1914, then rapidly accelerating to a peak lasting from 1917 to 1919, then precipitously falling to less than pre-war levels by 1921; the second smaller peak denotes the period of hyperinflation in 1923. It is difficult to reconcile what actually happened. One must always consider the succinct possibility of errors in the numerator and denominator, as discussed above. The last pre-war census was held in 1910, enumerating a population of 65 million. Two wartime censuses were held in 1916 and 1917, both covering a population of around 62 million. A census in 1919 reported a population of 61 million and the next regular census of 1925 counted 62 million inhabitants (https://en.wikipedia.org/wiki/Census_in_Germany). During this 15-year interval, the population appeared to have been much too stable to account for a gross denominator error in the calculation of the mortality rates. Potentially more biased might have been the numerator data, obtained by counting deaths from M. tuberculosis, particularly among males. The Spanish influenza pandemic beginning in early 1918 cost millions of lives and could have contributed to misattributing influenza deaths to TB or actually be co-responsible for causing TB deaths [63, 64] or that indeed the H1N1 1918 influenza epidemic removed TB sources of transmission [65]. Tuberculosis mortality in Germany had actually peaked in 1917, one year before the influenza pandemic, thus this type of misclassification could not account for the observation in that year, but influenza could have accounted at least partially for the subsequent decline in TB mortality. If one removes the years 1915 through 1923 and regresses linearly on the remaining data, the regression line suggests that the pre-war secular trend resumed its projection from 1924 onwards, as if the 10-year epidemic surge had had no influence whatsoever.
If we recall the discussion of the epidemiologic transitions in the model discussed at the beginning of this chapter, it suggests probably only one rational explanation: the decline in the transmission of M. tuberculosis at the core of the TB epidemic continued uninterrupted throughout the period of WWI and afterward. There was no ratchet effect in that a declining trend was resumed at a higher level after the 10-year epidemic was over: it continued as if the epidemic had never happened. This may imply that in the final analysis, any excess sources of infection that might have arisen as a result of the war, influenza, and post-war miseries did not contribute equally to the excess transmission as one would normally have expected. As morbidity data are unavailable for Germany during this period, one must thus conclude that either there was no excess morbidity, only excess fatality (as Rist implied [28]) or that both “usual” plus excess cases all died at an accelerated speed. The latter hypothesis is considerably more likely as the untreated case fatality of 70% [61] can simply not double, which would have to be postulated in the former case.
Case-to Death Ratio
To take the potential uncertainty about the correct denominator out of the equation, we can examine only numerators where they exist. For example, we can examine either the ratio of TB cases to TB deaths or the ratio of the respective rates, as both likely used the same denominator, and thus cancel out the influence of potentially faulty denominators. One problem with this is that for simplicity we have to use incident cases from a given year and deaths from the same year. Yet, patients dying from M. tuberculosis were incident cases not just in the current but also in earlier years as we know from the survival curves discussed earlier.
Fig. 8. Tuberculosis case-to-death ratio in 3 different United Kingdom jurisdictions in the pre-chemotherapy era [66–69], reproduced with permission from Elsevier. Underlying data to calculate ratios were abstracted from graphs, where those for England and Wales showed numbers, and those for Lancaster and Worcestershire rates derived from much smaller numbers than for England and Wales.
While morbidity data are available only for some countries in their entirety, some sub-country jurisdictions have notifications of both pulmonary TB cases and deaths for multiple sequential years. Three jurisdictions with such data were chosen from the United Kingdom to exemplify the use of case-to-death ratios. These were England and Wales over 2 time periods [66, 67], Lancaster [68], and Worcestershire [69]. None of the 3 studies revealed a marked change during the war years as far as they were included (Fig. 8). The data from England and Wales are based on large case numbers (60,000–70,000 per year), whereas the data from Lancaster and Worcestershire yielded rates based on much smaller case and population numbers (cities as opposed to 2 entire regions of the United Kingdom). The case-to-death ratio is remarkably constant over all the observation years with an average of about 1.6 for England and Wales, and Lancaster. The ratio in Worcester is higher at approximately 2.0 on average, but it also fluctuates much more as might be expected for the smallest among the 3 jurisdictions. This example demonstrates that in the pre-chemotherapy era, the case-to-death ratio – an approximation to the reciprocal of case fatality – was fairly constant in the inter-war years and was similar during wartime when TB, in Worcester for instance, did not increase [69].
Lewis-Faning examined specifically the interval between TB notification and death from M. tuberculosis and the effect of WWII on it, but could not find any difference, suggesting an effect of the war on acceleration of progression to death in Middlesex County, United Kingdom [70]. Rist noted that TB case notifications in Paris continued to decline during WWII, yet reported that the number of deaths increased in parallel [28]. Unfortunately, he did not provide the numbers to define this relationship more precisely.
Fig. 9. Reported notifications of cases of infectious tuberculosis