Jump to ContentJump to Main Navigation
Death in ChildbirthAn International Study of Maternal Care and Maternal Mortality 1800-1950$

Irvine Loudon

Print publication date: 1992

Print ISBN-13: 9780198229971

Published to Oxford Scholarship Online: October 2011

DOI: 10.1093/acprof:oso/9780198229971.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017

(p.534) Appendix 5 The Problem of Streptococcal Virulence

(p.534) Appendix 5 The Problem of Streptococcal Virulence

Death in Childbirth
Oxford University Press

This appendix is concerned with the problem of streptococcal virulence and its relevance to the history of puerperal fever. I shall put forward three propositions. First, that fluctuations in streptococcal virulence played a very important and largely unrecognized part in determining trends in maternal mortality. Secondly, that different strains of Streptococcus pyogenes caused puerperal fever and erysipelas on the one hand and scarlet fever on the other. Thirdly, that changes in virulence may have resulted from changes in the prevalence of certain varieties, strains, or in modern terminology certain streptococcal serotypes. Because these propositions are relevant to several chapters, and because some of what I will say is speculative, it seemed sensible to relegate this section to an appendix rather than the main text.

What is not speculative but established beyond reasonable doubt, is the following.

  1. 1. Until the late 1930s, puerperal fever was the most common cause of maternal mortality.

  2. 2. A large majority of puerperal fever deaths (as opposed to cases) were due to infection with the β-haemolytic streptococcus Lancefield group A (hereafter referred to as ‘the streptococcus’). When epidemics of puerperal fever occurred they appear to have been purely streptococcal in origin.

  3. 3. The shorter the time between childbirth and the onset of puerperal fever, the higher the mortality. Rapid onset and high mortality were common if not invariable features of epidemics.

  4. 4. Epidemics of puerperal fever could occur in towns and villages as well as in lying-in hospitals. When this happened such epidemics were frequently associated with concomitant outbreaks of erysipelas. Gordon of Aberdeen in 1795 and West of Abingdon in 1815 were two of many examples quoted in Chapter 4. Both described the simultaneous arrival, peak periods, and departure of epidemics of puerperal fever and erysipelas. No authors describing these outbreaks mentioned simultaneous outbreaks of scarlet fever or sore throats although both were well-recognized conditions. Epidemics of puerperal fever were common in the eighteenth and nineteenth centuries but they had become rare by the beginning of the twentieth century.

  5. 5. Wide fluctuations in annual mortality rates of both puerperal fever and erysipelas were a feature of the nineteenth century. In contrast, fluctuations were comparatively slight in the period 1900–35 when epidemics of puerperal fever and erysipelas had become rare. See Figures 13.1, 13.2, and 13.3.

  6. 6. Secular trends in deaths due to erysipelas and puerperal fever followed each other very closely, while the trends in deaths due to scarlet fever bore only a loose relationship to those of puerperal fever and erysipelas. This can be seen in Figure 4.3 and its significance is discussed below.

  7. 7. Studies carried out in the inter-war period showed that when streptococcal (p.535) illness was prevalent the asymptomatic carrier rate (that is, the proportion of the population who were not ill but carried streptococci in their noses or throats, often for long periods) usually rose significantly.

  8. 8. Not all asymptomatic carriers were liable to act as a source of streptococcal infection. Some authors have drawn a useful distinction between carriers and ‘dispersed’, the latter being those carriers who were infectious and thus the unwitting source of streptococcal cases and epidemics. Most dispersers were nasal carriers rather than throat carriers.1

There are numerous strains of Streptococcus pyogenes, differentiated by varying possession of two sets of antigens, the M and T antigens. These provide a means of identifying individuals as the source of epidemics. The identification of strains, however, has not provided any consistent correlation between particular strains and specific types of streptococcal disease. The complexity of the streptococcus, as studied in the laboratory, prompted Christie to comment that ‘it is unlikely that a simple epidemiological pattern for streptococcal infections could emerge and, in practice, this has been the case’.2

In spite of the absence of supporting laboratory evidence, clinical and epidemiological evidence suggests that at least two distinct groups of strains exist, or at least existed in the past. One group was associated with scarlet fever, streptococcal sore throat and tonsillitis, and mastoiditis. Scarlet fever is due to a particular strain which produces erythrogenic toxins; if the patient has no protective antibodies against these toxins they produce the characteristic skin rash.3 The other group of streptococcal strains was the causative agent in erysipelas, most cases of severe puerperal fever, and probably of streptococcal wound and skin infections. The evidence for separating streptococcal infections into two groups is as follows.

Clinically, the two groups of diseases are distinct. Scarlet fever and streptococcal tonsillitis and sore throats are predominantly diseases of children. Scarlet fever is associated with a rash and confers immunity. Until the second half of this century this group was associated with the serious complications of acute rheumatic fever and acute nephritis. Both complications are now rare in affluent Western countries, but they are still quite common in developing countries.

In the nineteenth century erysipelas was a disease of infancy and old age. During the twentieth century, cases in infancy gradually disappeared and it is now a disease of later life. Neither puerperal fever nor erysipelas are associated with a rash. Neither disease confers immunity. Indeed, erysipelas is well known for its tendency to recur. It was uncommon for erysipelas to act as the precursor of rheumatic fever or acute nehpritis, and as far as I can discover there are no records of these complications following puerperal fever. The response to the sulphonamides was different. In the late (p.536) 1930s and early 1940s before penicillin became available, the sulphonamides were highly successful in the treatment of puerperal fever and erysipelas, but they were described as ‘disappointing’ in the treatment of scarlet fever. These are the clinical differences which suggest that the strains of the streptococcus which cause scarlet fever may not be the same as those which cause erysipelas and puerperal fever.

The epidemiological evidence comes from several sources. Deaths from streptococcal diseases show a marked seasonality. During the late nineteenth century the peak of deaths from scarlet fever, erysipelas, and puerperal fever always occurred during the winter. By the 1920s, the peak for scarlet fever had moved to the spring while the peak for deaths from erysipelas and puerperal fever remained in the winter.4 The existence of two groups of strains is also suggested by the similarity between the secular trends for puerperal fever and erysipelas and the rather separate pathway followed by scarlet fever. This can be seen in Figures 4.2 and 4.3.

In 1935 Hektoen wrote a paper entitled ‘The Specificness of Certain Haemolytic Streptococci’ in which he quotes the evidence of Ruth Tunnicliff who, in 1931, suggested ‘the streptococci of scarlet fever and of erysipelas form distinct groups’. Hektoen pointed out that in contrast to the frequent coexistence of epidemics of erysipelas and puerperal fever, he found no records of epidemics in which some patients suffered from scarlet fever and some from erysipelas. He instanced the Faroe Islands where, in 1873, scarlet fever was introduced. There had been no cases of the disease in the previous fifty-seven years. During the epidemic which followed 38.3 per cent of the population had scarlet fever but there were no cases of erysipelas.5

Although I have been unable to find any historical accounts of simultaneous outbreaks of scarlet fever and puerperal fever, it is true that following the discovery of the streptococcus it was thought to be as unwise to attend a lying-in woman after attending a case of scarlet fever as it was after attending a case of erysipelas. That seemed, on common-sense grounds, to be sound practice because of the a priori assumption that there would be a high risk of cross-infection; but there is no evidence the assumption was correct.

One of the most persuasive lines of evidence for the existence of two different groups of streptococcal illness comes from the study of their fatality rates. This can be seen in Table A5.1 and Figure A5.1. It will be recalled that the fatality rates for puerperal fever are meaningless before the end of the 1920s (and still flawed in the 1930s). For notification purposes the disease was defined as puerperal septicaemia which was widely interpreted as meaning the end-stage of the disease. Mild cases were seldom notified and puerperal fever as a whole was under-notified to such an extent that calculations of fatality rates are totally unreliable. The apparent fall in the fatality of puerperal fever shown in Table A5.1 between 1918 and 1925 simply reflects increasing rates of notification.

The important comparison is therefore between the fatality rates of erysipelas and scarlet fever. The table and the figure show that between 1918 and 1934, the former (p.537)

Table A5.1 Deaths as a Percentage of Notifications: Scarlet Fever, Erysipelas, and Puerperal Pyrexia. England and Wales, 1918–1945


Scarlet fever


Puerperal pyrexia
















































































































(a) The apparent abrupt fall in fatality rates for puerperal fever in 1926–7 is an artefact due to the introduction of a new definition of puerperal fever.

Source: Annual reports of the Registrar General and Registrar General's Statistical Reviews.

rose and the latter declined. They went, in other words, in opposite directions. Note also that from 1937, when sulphonamides first became available, there is a pronounced fall in the fatality of erysipelas and puerperal pyrexia, but only a slight downward step in the already declining fatality of scarlet fever. This confirms contemporary clinical (p.538)
Appendix 5 The Problem of Streptococcal Virulence

Fig. Appendix 5.1 England and Wales, 1911–1939. Fatality rates of scarlet fever and erysipelas

Source: Registrar General for England and Wales, Decennial Supplements.

(p.539) opinion that the sulphonamides were ‘disappointing’ in the treatment of scarlet fever.6 It is important to note in passing that confident statements during the inter-war period that streptococcal virulence was declining were based on the trend in the case fatality rate of scarlet fever. Few, if any, undertook comparative studies of the different pathway followed by erysipelas.7

This brings us to the vexed question of virulence on which Turk has observed:

The term virulence is often used in an attempt to quantitate pathogenicity, but caution is needed here. It is sometimes convenient to be able to describe a strain as highly virulent, or of reduced virulence or avirulent; but we can give mathematical expression to virulence only if we define carefully the conditions under which it is measured.8

The ‘conditions’ in this context are the health and resistance to infection of the population. If we think of virulence as one side of a coin, susceptibility (or its reciprocal, ‘resistance’) is the other. An unusually serious outbreak of an infectious disease may be due to a highly virulent strain, or to a population weakened by malnutrition and poor health, or both. Thus it is sometimes said that the pandemic of influenza in 1918–19 killed so many people because of a low state of health following four years of war and deprivation—an unlikely hypothesis if only because the death rate was as high in non-combatant as combatant countries. A simple analogy would be the risk of death from a bite by a venomous snake. The chance of dying would be determined in part by whether the victim was large, strong, and healthy, or small, weak, and frail; and in part by the variety of snake (corresponding to the type of streptococcus) and also by the amount of venom injected (corresponding to the prevalence or ‘dose’ of infection, discussed below).

Historians, noting changes in fatality rates, usually turn to the health and nutrition of the population and search for correlations with periods of affluence or deprivation. They do so because of the implicit assumption that the toxicity or virulence of the infective organism is a constant. Yet a moment's reflection shows that in the nineteenth century it is unlikely that the susceptibility (or resistance) of the population could have varied so widely and rapidly that it accounted for the rapid swings in mortality due to puerperal fever. Still less is it possible to correlate peaks of deaths from streptococcal disease with periods of economic decline. It therefore seems likely that observed changes in fatality were due to changes in bacterial virulence rather than changes in (p.540) host susceptibility or resistance. If we assume from the evidence above that the same streptococcal strains were implicated in erysipelas and puerperal fever, we can take the fatality rates of erysipelas as proxy for the fatality rates of puerperal fever. This would lead to the conclusion that the strains responsible for causing puerperal fever were increasing in virulence between 1910 and 1934, accounting in part for the rise in maternal mortality which was a feature common to most Western countries.9

What mechanisms may be involved in an increase in virulence in this particular context? ‘Virulence’ seems to imply some kind of change in the character of the streptococcus making it especially toxic in some years, irrespective of any change in the susceptibility or resistance of the population at risk, and relatively non-toxic in other years. We have seen that there are numerous strains, differing in their M and T antigens. These antigens are thought to be related to pathogenicity. Increased virulence might appear to occur as a result of a relative increase in the proportion of highly virulent strains and a corresponding decrease in less-virulent strains. Higher mortality would follow as a result of what has been called the ‘dosage effect’ of toxic organisms. As Turk explains:

The number (‘dose’) of organisms taking part in an infection has an important influence on its outcome. In an epidemic, or in an outbreak of common-source illness such as food-poisoning after a party, it is often found that some of those exposed have symptomless infections, some are mildly ill and some are more severely affected. In general such variations are likely to be determined, in part at least, by differences in the dose of pathogens reaching individuals, though other factors such as host immunity can also be important10

What is suggested is that changes in the prevalence of the streptococcus within a given population can occur at two levels. First, changes in the prevalence of all strains of the streptococcus; secondly, changes in the relative proportions of virulent and less-virulent strains, independently of changes in total prevalence. This provides an attractive model for the wide and rapidly fluctuating variations in the incidence and death rate of streptococcal diseases which were such a feature of the nineteenth century.

To take a specific instance, deaths from puerperal fever and erysipelas swung from a low level in 1872 to a very high level in 1874, and back down to an even lower level in 1878 (the mortality rates of puerperal fever in these years were 17.0, 36.3, and 15.9 respectively). This huge swing up to a peak and down again could have been due to a rapid upsurge in the prevalence of a highly virulent strain, followed by its (temporary) diminution or disappearance.

Evidence on these points comes from modern work.11 It has been shown that (p.541) changes in the prevalence of different streptococcal strains or ‘serotypes’ do in fact occur independendy of each other over periods of a few years. It is also clear that the links between serotypes and specific disorders are highly complex. For example, in one study cases of erysipelas were strongly associated with M-type 1, scarlet fever with types 3 and 4, and the predominant strain in cases of puerperal fever was type 28. It was usual, however, to find many serotypes rather than one were associated with each particular streptococcal disorder. Finally, it is likely that variations in the virulence of certain serotypes can occur. Thus the authors of one study in 1987 wrote: ‘It seems clear that the strain, or strains, of M-type 1 that is currently being isolated is particularly virulent.’12

Modern methods of typing the streptococcus were not of course available in the pre-sulphonamide era. We will never know what serotypes predominated in the past. But it is reasonable to postulate that the trends and fluctuations in deaths from puerperal fever, erysipelas, and scarlet fever were associated with fluctuations in the prevalence of certain serotypes which varied widely in their toxicity or virulence; and some of them may now have become extinct.

There are still many unexplained features of streptococcal disease in the past. For example, the tendency for severe epidemics of erysipelas and puerperal fever to sweep through towns and villages in the eighteenth and nineteenth centuries, and the rarity of such epidemics by the early years of this century; the increasing mildness of scarlet fever in Western countries since the Second World War; the virtual disappearance of rheumatic fever and acute nephritis in Western countries. Whatever explanations are put forward to explain these phenomena or to elucidate the nature of bacterial virulence, it seems certain that before the introduction of specific therapy in the mid-19308, levels of puerperal fever in any given period—and thus levels of maternal mortality—were to a large extent determined by the bacterial changes we have discussed.


(1) J. M. Boissard and R. M. Fry, ‘Streptococcal School Outbreaks: A Method of Investigation and Control’, Journal of Hygiene of Cambridge, 64 (1966), 221–30.

(2) A. B. Christie, Infectious Diseases: Epidemiology and Clinical Practice, 2nd edn. (London, 1974), 1010–15.

(3) D. C. Turk, I. A. Porter, B. I. Duerden, and T. M. S. Reid, A Short Textbook of Medical Bacteriology, 5th edn. (London, 1983), 42. A vivid account of childhood deaths from mastoiditis during an epidemic of scarlet fever in Glasgow in 1935, which underlines how recently the streptococcus pyogenes was still a deadly organism in Britain, can be found in S. C. McEwan, ‘Personal View’, British Medical Journal (27 Oct. 1990), 993.

(4) W. T. Russell, ‘The Statistics of Erysipelas in England and Wales’, Journal of Hygiene, 33 (1933), 426.

(5) L. Hektoen, ‘The Specificness of Certain Haemolytic Streptococci’, Journal of the American Medical Association, 105 (1935), 1–2.

(6) Notifications of scarlet fever between 1911 and 1944 showed annual fluctuations with peaks in 1921 and 1934; but the general trend between these dates was level. In other words, the incidence (the number of new cases a year) of scarlet fever did not alter appreciably. Deaths from scarlet fever, however, fell sharply from the late 19th cent. In the 1870s the annual rate was about 14 deaths from scarlet fever per 100,000 total population. The number had fallen to 7.7 in 1914 and continued downwards to 1.2 by 1936. In the case of erysipelas, annual rates of notifications and deaths both rose from the beginning of the century to 1934. But the death rate increased more sharply than the rate of notifications so that the fatality rate rose from 3.9 per cent in 1911 to nearly 7 per cent by 1934.

(7) The link between erysipelas and puerperal fever was explored by Russell in 'The Statistics of erysipelas'. He made the fatal error of using cases of puerperal fever rather than deaths without apparendy realizing that the bacteriology of cases differed significandy from that of deaths, or that the case-notifications before 1927 were unreliable. This undermines his conclusions.

(8) Turk, Porter, Duerden, and Reid, A Short Textbook of Medical Bacteriology, 42.

(9) There is one possible objection, however, which must be mentioned to be dismissed. We have seen that the age-incidence of cases of erysipelas changed through the early decades of the 20th cent. The proportion of cases in infancy dwindled relative to those in late-middle and old age. This raises the question of whether the increasing fatality of erysipelas between 1911 and 1934 was an artefact associated with this change in age-incidence. That this was not the case is shown by calculating the age-specific fatality rates for erysipelas. They show that the fatality rate of erysipelas rose separately and similarly in the old and the young.

(10) Turk, Porter, Duerden, and Reid, A Short Textbook of Medical Bacteriology, 42.

(11) E. Gaworzewska and G. Colman, ‘Changes in the Pattern of Infection Caused by Streptococcus Pyogenes’, Epidemiology and Infection, 100 (1988), 257–69, and Colman, A. Efstratiou, and E. T. Gaworzewska, ‘The Pyogenic Streptococci’, PHLS Microbiology Digest 5/1 (1988), 5–7.

(12) Gaworzewska and Colman, ‘Changes in the Pattern of Infection’, 267.