Introduction

Encephalitis lethargica remains a puzzling chapter in the recent history of medicine. It was first described as a specific disorder in 1917 by the colorful Austrian neurologist Constantin von Economo (1876–1931); his papers on the small Viennese epidemic in the winter of 1916/1917 (von Economo 1917) were so complete that the disorder was frequently referred to as ‘von Economo’s disease’. At about the same time, the Frenchman René Cruchet reported a similar syndrome on the French battlefields, but without immediately recognizing it as a distinct disorder (Cruchet et al. 1917).

Encephalitis lethargica was a neuropsychiatric disease par excellence. The vast majority of those who survived the initial period of the disorder were plagued by both neurological complaints (particularly parkinsonism) as well as by psychiatric problems which ranged from nervousness and loss of concentration to a syndrome which was difficult to distinguish from schizophrenia. Most of the victims were younger than 40 years of age at the time of infection, so that these neuropsychiatric sequelae represented a long term problem for both patients and their families. Intensifying the disturbing nature of the disease was the fact that a lengthy period of apparent well-being might intervene between recovery from the initial attack and the presentation of sequelae, an interval typically ranging from a few months to several years, but not infrequently stretching to decades. It had quickly become clear that the consequences of encephalitis lethargica were to be a long term public health problem, even though the epidemic form of the disorder was not encountered in most countries after 1925.

The etiology of the epidemic has never been satisfactorily explained and remains a source of controversy. The most widely accepted hypothesis is that the H1N1 influenza A virus was involved. This is principally because a pandemic of this virus swept the world in the wake of the First World War, claiming more lives than were lost in the military catastrophe. But a causal relationship between the two disorders was regarded by most authorities with scepticism, and only became well established as the two epidemics receded into history.

An examination of the etiology of encephalitis lethargica must nevertheless address the influenza issue, at least in response to challenges such as that posed by Maurizi (1989): if influenza was not the virus of encephalitis lethargica, what was? As some of the phenomena associated with encephalitis lethargica had also been reported during the 1889–1892 influenza pandemic, my research has also investigated this earlier outbreak as well as the literature which appeared between 1889 and 1918 concerning the epidemiology, etiology and clinical features of this and earlier nineteenth century influenza epidemics. This research revealed, however, another problem which has not hitherto been expressly addressed: what is the precise relationship between the epidemic disease ‘influenza’, as observed in 1918/1919, and the ‘influenza virus’, discovered a decade after the end of the pandemic?

The nature of 1918/1919 influenza pandemic

The 1918/1919 influenza pandemic was regarded until recently by many as ‘the forgotten catastrophe’, but a flood of monographs, scientific journal articles and newspaper stories has significantly raised consciousness of the events of 1918/1919. In scientific, medical and public health circles, the 1918/1919 pandemic serves both as a warning and as a model of the threat posed by influenza in its most dramatic guise, and expectations concerning and preparations for the anticipated next ‘big one’ are based largely on analyses of this pandemic.

Community interest in influenza has been elevated over the past decade by two events: (1) the very public discussions of the threat to human health posed by avian H5N1 influenza; and (2) the reconstruction by Taubenberger et al. of the virus presumed to be responsible for the 1918/1919 influenza pandemic (Taubenberger 2005). Uneasiness regarding the possibility of another influenza pandemic of similar scale displaced for a few months even global warming from public discussions of potential worldwide catastrophes, with a combination of horrified fascination and the confidence that Western medicine could manage any medical emergency. Smil (2005) characterized influenza as the single most likely ‘transformational catastrophe’ in the foreseeable future, the event with the greatest capacity for having a major impact on history. But its recent absence from the front pages of newspapers certainly does not mean the threat has subsided, and the monitoring of human and animal influenza viruses throughout the world remains a priority for the World Health Organization and of national health authorities and institutions.

The mechanisms underlying the emergence of novel influenza strains, however, would not be elucidated until long after 1918. Despite major scientific advances achieved in the wake of the 1890s pandemic, the 1918/1919 pandemic represented a crisis for both bacteriology and scientific medicine in general, as, in the absence of antibiotics, effective therapy was largely limited to rest and support. Nevertheless, while the rising death toll was a source of consternation, most infected persons in the Western world, especially those not hospitalized, survived with minor if any after-effects, a feature of the pandemic which is often overlooked.

The first conclusion which emerged from my investigation of historical influenza epidemics was that the 1918 influenza pandemic was exceptional in both its epidemiology and its clinical characteristics. To be specific: the 1918 pandemic was not only quantitatively different from both earlier and more recent influenza epidemics, there were also qualitative differences which must be considered when attempting to address future pandemics. Four of the most significant features which are difficult to reconcile with what has since been established concerning viral influenza are:

  • The speed with which the disease spread around the world;

  • The rapid succession of three ‘waves’ of infection, commencing in the northern summer of 1918;

  • The higher than usual death rate in most countries during the second and deadliest wave, and the variability of the death rate around the globe (compare, for instance, the death rate in any developed nation with the horrendous toll in India);

  • The unusual W-shape of the deaths versus age distribution during the second wave; that is, in contrast to most epidemics, unusually high death rate in people between 10 and 40 years of age, a phenomenon that was observed in most countries investigated in detail.

These and other features of the 1918/1919 pandemic have long puzzled epidemiologists, and epidemiological solutions which incorporate all these critical phenomena have not emerged. Indeed, Morens and Fauci (2007) recently noted in a review that these features remained ‘mysteries’.

Knowledge of past epidemics is essential for addressing future crises, but it is crucial that this knowledge encompasses all relevant phenomena, not just those which are explained by existing models. As bluntly expressed by Davenport (1977): “Epidemiological hypotheses must provide satisfactory explanations for all the known findings––not just for a convenient subset of them”. Unusual or even unique features of the 1918 pandemic need to be understood in order to determine the validity of conclusions based upon this particularly epidemic. Close reading of the recent literature on the 1918 pandemic indicates that a large proportion of information is repeatedly recycled; the hypotheses regarding the difficult facets of the epidemic are similarly recycled, while generally concluding that many problems remain unresolved. My contention is that it is precisely these aspects of the epidemic which demand closer attention.

A medico-historical approach to influenza

An historical perspective might shed some light on these phenomena, essentially by exploring potential explanations regarded as less important or disregarded altogether by current influenza models. Recent authors have concentrated on the 1918 event in isolation, or with passing reference to the 1890s pandemic; English language papers, in particular, largely ignore many foreign works of repute which deal with influenza both prior to 1889, such as Ozanam’s outstanding chronicle of its history in his Histoire médicale générale et particulière des maladies épidémiques (Ozanam 1817–1823), as well as encyclopedic reviews of the 1890s pandemic, such as the outstanding investigation by the Berlin Association for Internal Medicine (Leyden and Guttmann 1892).

The first matter which must be addressed is the nature of ‘influenza’ in 1918. Since the discovery of the influenza virus in 1929, ‘influenza’ has been tautologically defined as the disease caused by this virus. This was clearly not the case in 1918. Most contemporary authorities regarded influenza as a protean disorder with a long history, with the first ‘definite’ outbreaks dated to the twelfth century, but it was less defined as an entity than, for example, cholera or tuberculosis. Prior to 1889, influenza (‘epidemic catarrh’, ‘grip’, ‘grippe’) was associated with a broad and variable range of symptoms, with the ‘influenza’ of particular years being ascribed, for example, to ‘nervous’, ‘gastrointestinal’ or ‘catarrhal’ forms of the disease: the most constant feature of influenza was its inconstancy. It generally visited communities for several years before disappearing, then re-appeared decades later, often to be described as a ‘new disease’––until the historically minded pointed to forgotten parallels (reviews: Forbes and Conolly 1839; Leyden and Guttmann 1892; Althaus 1892).

After the German bacteriologist Richard Pfeiffer isolated the so-called ‘influenza’ or ‘Pfeiffer bacillus’ (now: Haemophilus influenzae) in 1892 (detailed discussion: Pfeiffer 1892, 1893), influenza was provisionally re-defined as the disease elicited by this pathogen. Nevertheless, doubts were immediately expressed concerning the simplistic nature of this explanation, and by 1918 the etiology was again uncertain. As doubts increased regarding the essential involvement of the ‘influenza bacillus’, it was confidently asserted by many investigators that ‘influenza’ could be elicited by a number of pathogenic agents, all producing a similar clinical picture. The issue was further clouded by the fact that it was not clear whether pandemic influenza and interpandemic influenza were actually the same disorder (Leichtenstern 1912; Jordan 1922).

In 1931 Shope published the results of his detailed investigation of swine influenza, wherein he emphasized the etiological role of a filterable virus. Then in 1933 the British group led by Christopher Andrewes demonstrated that a similar virus could be isolated from the throat washings of influenza patients, a virus which could induce influenza in both humans and ferrets, and that antibodies to this virus were present in the blood of those exposed to the 1918 epidemic (Smith et al. 1933). The case was thereby metaphorically closed, and ‘influenza’ was henceforth defined specifically as the disorder caused by this virus.

The ascription of the 1918/1919 pandemic to the virus identified in 1933 was thus retrospective, and even at the time open to criticism. Andrewes, who played a major role in elucidating the role of influenza virus in the disorder, was nonetheless later troubled by doubts as to whether it was capable of explaining everything about influenza epidemics (reviewed by Hope-Simpson 1986).

Further research revealed the remarkable genetic mutability of the influenza virus. Influenza pandemics, according to current models, occur when ‘antigenic shift’ facilitates the sudden emergence of a novel influenza virus, providing an immunological challenge for the unprepared potential hosts. Antigenic shift is the process whereby an existing influenza strain acquires a new hemagglutinin (HA) gene, encoding a protein markedly different from that expressed previously by this virus, by exchange with another virus strain, with the most feared possibility currently being the introduction of HA gene of the highly pathogenic H5N1 avian influenza. This process is distinct from the incremental ‘antigenic drift’ in the amino acid constitution of the influenza virus HA and neuraminidase (N) proteins, which altering the antibody-binding properties of the virus without radically altering its other properties (Webster and Laver 1971; Hampson and Mackenzie 2006).

There is thus obvious interest in understanding the molecular biology of influenza viruses, and in this regard Taubenberger’s reconstruction of the 1918 virus has been justifiably very influential. It has been demonstrated, for instance, that this virus effectively killed 100% of treated mice, autopsies of which revealed massive pulmonary bleeding (Taubenberger 2005). It should, however, be noted that animal models of influenza in 1918 were far from successful. Laboratory techniques have certainly improved since the 1920s, but it was only with the serendipitous infection of ferrets by Andrewes’ group which succeeded in producing an animal model of human influenza––success which he explained in terms of the similar nasal flora (that is, bacteria) in humans and ferrets. Further, the death rate among humans in 1918 did not approach 100%, and in most settings did not exceed 2%, nor was dramatic pulmonary involvement a constant feature of the disorder.

To take pre-1933 descriptions of ‘influenza’ at face value, in any case, is fraught with danger. Not everything which was labeled ‘influenza’ would be regarded today as influenza, let alone the result of infection with the influenza virus. Historical analysis of early twentieth century and earlier texts concerning influenza indicates that there were major differences from current concepts regarding the use of this term. This is particularly the case as until 1918 ‘influenza’, although often interchangeably employed with terms such as ‘epidemic catarrh’, had been regarded as a ‘nervous disorder’: that is, the primary concern was not with pulmonary or catarrhal symptoms, but rather with the underlying nervous affliction caused by the infectious agent (see for example Althaus 1892). Indeed, the vagueness of the term ‘influenza’ was such that as respected (if unorthodox) an epidemiologist as Francis Crookshank (1920, 1922) could seriously assert that not only were influenza and encephalitis lethargica different presentations of a single disorder, but also that encephalomyelitis, poliomyelitis, ergotism and the ‘English sweats’ were all essentially further manifestations of the same disease. This must also be considered in the context of concerns that epidemic brain disease appeared to have increased significantly in the first quarter of the twentieth century, with major epidemics of poliomyelitis, meningococcal disease and encephalitis lethargica contributing to a disquieting era of epidemic neurologic disease (see MacNalty 1927).

There was much discussion at the time regarding the nature of the 1918/1919 epidemic, particularly concerning its relationship with the 1890s pandemic and with interpandemic influenza, and also of its apparently abrupt emergence from nowhere in 1918. But closer examination of medical journals and statistical data indicates that influenza had clearly been present at significant levels prior to the first wave of 1918, and had been since 1892; many authors commented that this was different to the period between the epidemics of 1853 and 1889, where influenza per se was virtually unknown, so much so that a large number of medical theses were devoted in the early 1890s to the ‘half-forgotten infection’ (see for example Eberling 1892). Contemporary writers also noted that ‘purulent bronchitis’ had caused concern, particularly in military hospitals, from as early as 1915, and that English hospitals had recorded increased isolation of the ‘influenza bacillus’ from patients in 1916/1917 (Hammond et al. 1917; Abrahams et al. 1917). One of the criticisms faced by German politicians in the winter of 1918 was that they had undertaken little to resolve the medical emergency, despite the fact that the number of respiratory deaths had exhibited a marked, continual increase since 1916. Contemporary authorities thus questioned whether the events of 1918 were in fact an autonomous pandemic, or rather the culmination of an evolving medical emergency which had evolved over the previous 3 years (Witte 2006, pp 273f).

It is important to note that it was the ‘second wave’ of the pandemic (autumn 1918) which was particularly virulent and widespread. Careful examination of statistics reveals that the first and/or third waves were not always apparent in some locations, and in others would not have attracted particular attention if not associated with the second wave. Without dismissing the potential significance of the ‘trailer waves’, the second wave is thus of primary importance with regard to mortality.

Apart from the magnitude of the death toll associated with this wave, the first feature which strikes the investigator is the shape of the age-related curve; specifically, the fact that mortality was highest among those in the “prime of life” (20–40 years) rather than among infants and the elderly, who normally constitute the worst affected groups in ‘normal’ influenza seasons. The W-shape of the mortality curve is not attributable to reduced mortality at either end (although for infants there was a reduction in many countries) but includes an indubitably marked increase in the middle age groups. The suggestion that normal U-shaped curve observed in most influenza epidemics was shifted to the left and combined with a certain degree of immunity in older age groups must thus be rejected.

Protection of older persons by prior exposure to the pandemic influenza virus––during the 1889–1892 pandemic, for instance––has been invoked to explain their sparing in 1918/19, but it is now generally reported that the virus responsible for the 1890s pandemic was probably a H3 type virus, not the H1N1 held responsible for 1918/19. More recently the acquisition of protection through exposure during the first wave to a less virulent form of the virus has been proposed (Barry et al. 2008), but this would not explain the age distribution of deaths in the second wave.

The increased mortality in younger (but not youngest) sufferers was attributed by the Australian immunologist Frank Macfarlane Burnet to an excessive immunological response to infection in this age-group (Burnet and Clark 1942), an hypothesis which receives some support from recent studies of Taubenberger’s resurrected virus in mice (Kash et al. 2006; Pappas et al. 2008) and macaques (Kobasa et al. 2007). But there is no direct evidence that this mechanism contributed significantly to the shape of the mortality curve, and it should be noted that in most cases contemporary physicians did not record any notable differences in clinical presentation from other cases of influenza/pneumonia. This conclusion was also reached most recently by Taubenberger and Morens (2008).

Three waves of influenza in rapid succession and the W-shaped mortality curve are not explained by current epidemiological models of influenza. It should, for a start, also be remembered that three waves were not observed in all parts of the world, and this schema is primarily the retrospective assessment of the course of the pandemic in the United States and parts of Europe.

An alternative explanation is that the mortality hump is attributable to the death rates of influenza and a second disease being superimposed: that is, the younger victims’ deaths are not primarily attributable to influenza itself, but rather to another infection. It is parenthetically interesting that the same age band was found for diagnoses of encephalitis lethargica. The ‘two infection’ hypothesis receives support from several directions:

  • ‘Influenza deaths’ actually represent the total of all respiratory and any other deaths which might conceivably be linked to influenza and its consequences. In particular, it was nigh impossible to reliably differentiate between viral and bacterial pneumonia in 1918. Curiously, reported pneumonia and bronchitis rates in England declined during the 1918 epidemic. It should also be noted that “pneumonia” was already a major killer in North America before 1918 (Collins and Lehmann 1953).

  • It proved impossible in 1918 to demonstrate transmission of influenza to animals by administration of either filtered or non-filtered fluids from human patients. Indeed it proved impossible, despite ethically questionable approaches, to demonstrate human-to-human transmission of the pathogen. It should, however, also be noted that it has since proved astonishingly difficult to demonstrate directly human-to-human transmission of the influenza virus (Earn et al. 2002). The techniques then available to investigate what we now term ‘viruses’ were sufficient for the identification of a filterable virus in body fluids, just as Landsteiner had succeeded in 1909 in demonstrating the viral etiology of poliomyelitis. Instead, the most common organisms associated with patients of the second wave were Streptococcus pyogenes, Streptococcus pneumoniae and Staphylococcus aureus (see detailed review: Levinthal et al. 1921).

  • Mortality during the 1918/1919 pandemic was markedly different across the globe. This can possibly be explained by differing degrees of viral ‘naivety’ of respective populations, but this still assumes the existence of a similar virus prior to 1918. There were also marked differences in mortality within countries, one of the common features associated with higher mortality being density of population (including hospitals, especially military hospitals).

  • Most importantly, contemporary researchers explored carefully the possible involvement of filterable viruses in influenza, and the consensus reached was that they were not. There were certainly dissenting opinions, but were regarded at the time as less probable (reviewed: Thomson et al. 1933).

Many recent accounts of the 1918 epidemic lend emphasis to the dramatic cyanosis and the extreme pulmonary flooding experienced in mortal cases. Close examination of the literature indicates, however, that (1) similar cases were also described before the 1918 ‘first wave’, and attributed variously to pneumonia, influenza and other respiratory disorders; and (2) such cases were atypical of the general course of the epidemic and were largely confined to particular settings, especially military hospitals (where death rates of up to 10% were reported in the United States; see review in Hall 1928). That 99% of European influenza sufferers recovered without notable long term problems is sometimes obscured by the attention paid to the more dramatic and tragic outcomes.

The multi-factorial etiology of influenza: the ‘complex virus’

Some leading authorities posited at the time of the 1918 epidemic (and also earlier) that influenza was not explicable through the actions of a single infectious agent. In particular, the highly respected Swiss physician and physiologist Hermann Sahli proposed the concept of a ‘complex virus’, whereby the synergistic action of two or more entities elicited influenza (Sahli 1919). Sahli’s reputation was based on his precise and controlled laboratory methods (see Sahli 1913–1920). The uncertainty regarding etiology and the possibility that more than one infectious agent was involved was also reflected in public health responses to the epidemic: vaccines generally included the ‘influenza bacillus’ together with ‘accompanying bacteria’, including pneumococci, streptococci, diplococci, staphylococci and Moraxella catarrhalis (for example, Paton et al. 1920). There were also reports which explicitly observed that influenza alone was rarely fatal during the epidemic (for example, Christian 1918). The multi-factorial etiology hypothesis for influenza has also been recently revived elsewhere (McCullers 2006).

The complex virus model differs in a crucial point from current concepts of ‘complications’ or ‘passenger infections’ associated with influenza, in that the infection with a second agent––such as a bacterium––is considered an essential feature of the disorder ‘influenza’. It replaces the ‘one germ-one disease’ concept with a multi-factorial approach. In veterinary medicine this is not unfamiliar: in swine the synergistic infection with Haemophilus suis and swine influenza has been recognized since the identification of the swine influenza virus in the 1930s, and outbreaks of ‘swine ‘flu’ often refer to the bacteria rather than the virus. The same relationship exists in cattle for Haemophilus somnus and respiratory syncytial virus, and in mice for Streptococcus pneumoniae and influenza viruses. Moreover, the synergistic nature of infection with influenza and, for instance, Streptococcus and Pneumococcus species has been demonstrated at the biochemical level (Okamoto et al. 2003; McCullers 2006). Finally, there have been several recent reports suggesting that pneumococcal vaccines are at least as effective in reducing influenza deaths in at-risk populations, but that the effects of the two vaccine types is not additive (Blay et al. 2007).

In calculating the number of deaths caused by the 1918 virus, reviewers explicitly include almost all respiratory deaths in the relevant period, most of which are directly bacterial. My argument is that this is not in itself incorrect, but that they should not be regarded simply as ‘complicated influenza’: the etiological significance of the bacterial component of the disorder termed ‘influenza’ should be upgraded.

The major working hypothesis to be drawn from this research recognizes that the 1918 epidemic was in many ways significantly different from inter-pandemic influenza and indeed from more recent influenza pandemics. What appears to have occurred is a fatal co-incidence of epidemics of influenza virus and another infection, probably bacterial in nature. It is further contended that the elevated death toll was primarily due not to the direct effects of influenza infection, but due to the combined effects of the two infections, neither of which may be regarded as the ‘primary’ and the other as the ‘secondary’ infection.

Conclusions: implications of the ‘complex virus’ hypothesis of influenza for addressing future influenza epidemics

The ‘complex virus’ hypothesis has several crucial consequences for addressing future influenza epidemics:

  1. 1.

    Expensive antiviral agents, the effectiveness and safety of which in a situation of mass administration remains untested, should be reserved for those involved in close contact with numbers of afflicted persons; that is, for medical and other emergency services personnel.

  2. 2.

    Amelioration of the worst effects of an influenza epidemic can probably be achieved with antibiotic agents. Indeed, it may be the widespread employment of such drugs which has prevented subsequent influenza epidemics assuming the proportions of the 1918/1919 epidemic.

  3. 3.

    Where alternatives are possible, influenza patients should not be concentrated in hospitals or elsewhere. Control of bacterial infection will be the major priority, and the isolation of infected persons will assist this aim. Isolation in the age of constant electronic connectivity is incomparably more practical and tolerable than in previous epidemics, although the economic impact of widespread implementation may be considerable in the short term.

  4. 4.

    Attention should be paid to changes in the incidence or virulence of bacterial respiratory diseases, both inside and outside of major hospitals. Nosocomial bacterial infections, in particular, may play a particularly decisive role in the course of an influenza epidemic. Forewarning of an impending epidemic may be provided by a rise in general respiratory disease incidence.

  5. 5.

    Another, indirect consequence is that, while alertness to changes in avian influenza is prudent, the threat to humans posed by avian influenza is probably more complex than first thought. Human infections with H5N1 influenza virus to date have involved without exception persons living in extreme proximity to infected birds and exposed to their bodily fluids, and the relatives of such persons. The presence of receptors for the avian virus in the lower bronchial regions of humans suffices to explain such infections (Shinya et al. 2006). The probability of the feared crossover event, leading to an influenza virus combining the virulence of the avian type with the capacity for human-to-human transmissibility, is much greater in East Asia in conditions of close human/bird/swine proximity than in Europe or North America following the arrival of an infected migratory bird.

It is not my intention to impugn the value or the validity of the invaluable research conducted over the past 20 years concerning the genetics and molecular biology of the influenza virus, nor that of mathematical modeling of influenza epidemics. It must be remembered, however, that such models describe transmission of the virus, not necessarily that of the disease. This is important when discussing the 1918 pandemic: if, as now believed, viral transmissibility was moderate (see for example Andreasen et al. 2008; Cliff et al. 2008) and mortality was quite low in most settings, it becomes valid to ask whether the virus alone could have wreaked the havoc it did.

The proposal here is that the course of an influenza pandemic is not determined by the characteristics of the influenza virus and the host population alone. It emphasizes rather the role of co-infecting agents, thereby providing a better explanation of some of the difficult features of the 1918 pandemic. Shortly before submitting the original version of this paper I became aware of the recent papers by Brundage and Shanks (2008) and Morens et al. (2008) regarding the role of bacterial pathogens in the mortality of the 1918/1919 pandemic. My views most closely concur with those of the first mentioned publication, in that I not only propose that the bacterial component of a dual infection was more critical in determining the dimensions of the 1918/1919 pandemic, but also that addressing this bacterial threat directly, rather than relying on virus containment, might be a more convenient means to address a future pandemic. I would perhaps proceed a step further, and suggest that without coincidence of viral and bacterial infection––and it is not clear in this respect whether we are dealing with bacterial co-infection as such or potentially with a specific bacterial pathogen with a genetic make-up which has also undergone an ominous shift with regard to virulence––an influenza pandemic might not pose as great a public health threat as anticipated.

A detailed discussion of the discussion of th e nature of and etiological factors in influenza between the 1890s and 1918/1919 pandemics will be published elsewhere; the intent of the present paper was simply to establish the conceptual basis for the discussion of the relationship between influenza and encephalitis lethargica which will be undertaken in part II of this publication.