key: cord-0013123-ytr5fsj7
authors: Chertow, Daniel S.; Cai, Rongman; Sun, Junfeng; Grantham, John; Taubenberger, Jeffery K.; Morens, David M.
title: Influenza Circulation in United States Army Training Camps Before and During the 1918 Influenza Pandemic: Clues to Early Detection of Pandemic Viral Emergence
date: 2015-04-02
journal: Open Forum Infect Dis
DOI: 10.1093/ofid/ofv021
sha: 74965b149fbcdb3f50817edaeafe5fd6f71011d4
doc_id: 13123
cord_uid: ytr5fsj7

Background. Surveillance for respiratory diseases in domestic National Army and National Guard training camps began after the United States’ entry into World War I, 17 months before the “Spanish influenza” pandemic appeared. Methods. Morbidity, mortality, and case-fatality data from 605 625 admissions and 18 258 deaths recorded for 7 diagnostic categories of respiratory diseases, including influenza and pneumonia, were examined over prepandemic and pandemic periods. Results. High pandemic influenza mortality was primarily due to increased incidence of, but not increased severity of, secondary bacterial pneumonias. Conclusions. Two prepandemic incidence peaks of probable influenza, in December 1917–January 1918 and in March–April 1918, differed markedly from the September–October 1918 pandemic onset peak in their clinical-epidemiologic features, and they may have been caused by seasonal or endemic viruses. Nevertheless, rising proportions of very low incidence postinfluenza bronchopneumonia (diagnosed at the time as influenza and bronchopneumonia) in early 1918 could have reflected circulation of the pandemic virus 5 months before it emerged in pandemic form. In this study, we discuss the possibility of detecting pandemic viruses before they emerge, by surveillance of special populations.

United States Army statistics on influenza and pneumonia have been recorded since the middle 1800s [1] . Surveillance for influenza and pneumonia in domestic National Army and National Guard training camps began after the United States' entry into World War I, in April 1917. Seventeen months later, the "Spanish influenza" pandemic appeared [2] . Although the vast majority of pandemic influenza cases in the general population were of a self-limited nature, similar to cases seen in influenza epidemics before and since, a small percentage (in the range of 1 percent) were fatal, and nearly all of these were complicated by severe secondary bacterial pneumonias [3] .

Human influenza A viruses were first identified in 1933 [4] , and the virus responsible for the 1918 pandemic was first identified in 1997 [5] , when its RNA genome was sequenced from preserved and frozen human tissues [6] . By studying the reconstructed 1918 influenza virus, important insights into its origin and pathogenesis were found [7] . Understanding the 1918 pandemic also requires understanding the epidemiology of its occurrence, complications, and mortality. It is of particular value to examine, using standardized approaches, clinical and epidemiologic aspects of influenza-like illnesses and complications in large, healthy, homogeneous populations under continuous and intense surveillance before and during the pandemic. In this study, we examine data on respiratory illnesses and deaths in US military training camps during World War I.

Data were abstracted from a 1925 report of the US Army Surgeon General's Office summarizing admissions and deaths attributable to influenza, pneumonia, and other respiratory diseases, in 39 US military training camps from April 1917 to December 1919 [8] . Data were tabulated by month for each individual camp. Troop strengths by month and by camp were also recorded. Because most camps were not yet fully operational between April and September 1917, and because camps were largely demobilized between April and December 1919, data from these periods, although available in the Surgeon General's report [8] , are not considered here.

Original 1917-1919 diagnostic categorization of soldiers' illnesses reflected military providers' clinical diagnoses and military epidemiologists' judgments about diagnostic categorization. There appear to have been no standardized criteria for diagnosis or diagnostic categorization, but because of rapid and broad Army-wide dissemination of medical information, including exchange of information between Army camps, diagnostic criteria may be reasonably inferred from clinical descriptions and Army reports of the time [9] . Because the cause of influenza was unknown in this era, diagnoses could not be confirmed by viral culture, serology, or molecular-based diagnostic tests, although complications of pneumonia were readily diagnosed by radiography and by bacterial culture of pleural fluid, blood, or tissue at autopsy. Disease misclassifications seem to have been uncommon with the possible exceptions, discussed below, of lobar versus bronchopneumonia in the prepandemic months, and of some of the specific diseases subsumed within the Common respiratory diseases category. Data abstracted from original diagnoses of respiratory tract diseases were placed by military epidemiologists in the 1917-1919 era into the following 7 diagnostic categories: (1) Influenza, uncomplicated, (2) Influenza and lobar pneumonia, (3) Influenza and bronchopneumonia, (4) Influenza and other complications, (5) Bronchopneumonia, (6) Lobar pneumonia, and (7) Common respiratory diseases.

"Influenza, uncomplicated" referred to fever and malaise with or without upper respiratory symptoms. Soldiers admitted with Influenza, uncomplicated who subsequently developed complications were to be recategorized according to the complication, but such recategorizations were apparently not always made [1] . "Bronchopneumonia" was distinguished from "Lobar pneumonia" by clinical signs of diffuse lower respiratory tract involvement without evidence of early lung consolidation [9] . "Influenza and other complications" referred to involvement of organs outside the respiratory tract including heart, brain, gastrointestinal tract, and kidneys, eg, complications resulting from secondary bacteremia. "Common respiratory diseases" were included in some of the analyses because of the possibility that, if not correctly diagnosed, sporadic cases of mild influenza, or cases seen in small influenza outbreaks, might be categorized under this rubric, which also included noninfluenza diagnoses of pharyngitis, tonsillitis, bronchitis, and presumed viral "colds" [1] . Postmeasles lobar pneumonia and bronchopneumonia, including those cases diagnosed during Army-wide measles epidemics in late 1917 and early 1918 [10] , were recorded in separate diagnostic categories and were not considered here. Of the 605 651 admissions and 18 354 deaths recorded in the 7 categories noted above, made between October 1917 and March 1919, 122 diagnoses (26 admissions and 96 deaths) appeared to be inaccurate and were omitted from analyses. In these instances, the number of monthly deaths in a category were greater than admissions in that category for the concurrent and preceding months.

Diagnostic category-specific attack rates (number of admissions per 1000 troops), mortality rates (number of deaths per 100 troops), and case-fatality rates (number of deaths per 100 admissions), with 95% confidence intervals (CIs), were calculated in each month for all camps combined. Confidence intervals were calculated using the binomial distribution [11] .

In some analyses, certain of the original diagnostic categories were combined in order to consider effects of their specificity or sensitivity. The original 7 respiratory diagnostic categories were combined as 1 diagnostic category referred to below and in the Figures and Tables as "All influenza illnesses", meaning all 7 of the diagnostic categories combined, all of which were believed to contain at least some influenza cases. The 4 original diagnostic categories thought to contain cases of influenza complicated by pneumonia (Influenza and bronchopneumonia, Influenza and lobar pneumonia, Bronchopneumonia, and Lobar pneumonia) were combined in some analyses and termed "Influenza complicated by pneumonia." The latter 2 of these categories (Bronchopneumonia, and Lobar pneumonia), in which a diagnosis of influenza had not been made, were included in such analyses because presenting postinfluenza bacterial pneumonias might not have been correctly associated with antecedent influenza illnesses. Diagnostic categories believed to contain cases of influenza that were not complicated by pneumonia or other complications (Influenza, uncomplicated, and Common respiratory diseases) were combined in some analyses and termed "Influenza not complicated by pneumonia."

The 4 original diagnostic categories believed to be most specific for influenza infection (Influenza, uncomplicated; Influenza and lobar pneumonia; Influenza and bronchopneumonia; and Influenza and other complications) were combined and termed "Specified influenza infection." The 2 original diagnostic categories thought to be most specific for postinfluenza pneumonia (Influenza and lobar pneumonia and Influenza . Given possible differences between camps due to geographical or other factors, a random effects model accounting for potential differences between camps was used. Log odds ratios (ORs) were converted to ORs with associated 95% CI.

Admission rates of All influenza illnesses during the 1917-1918 influenza season (December 1917-April 1918) peaked in January 1918 at 46 cases per 1000 troops and peaked again in April 1918 at 51 cases per 1000 troops ( Figure 1A) . In both the January 1918 peak and the April 1918 peak, Influenza complicated by pneumonia made up a relatively small proportion of All influenza illnesses (7% and 5.5%, respectively; Figure 1A ). Admission rates of All influenza illnesses peaked again during the pandemic in October 1918 at 164 cases per 1000 troops (Figure 1A ). At this time, Influenza complicated by pneumonia accounted for a significantly larger proportion of All influenza illnesses (19%; Figure 1A ), primarily due to Influenza and bronchopneumonia (53%; Figure 1B) .

Odds of admission to camp hospitals in September-October 1918 relative to December 1917-April 1918 were increased in 6 of the 7 original diagnostic categories, the exception being Common respiratory diseases ( Table 1) . Odds of admission for Influenza complicated by pneumonia were elevated approximately 10-fold in September-October 1918 relative to December 1917-April 1918 (OR, 9.45; 95% CI, 9.25-9.66) ( Table 1) , and relative to the March-April 1918 period (OR, 10.04; 95% CI, 9.72-10.37), with odds of admission for Influenza and bronchopneumonia being exceptionally high (OR, 784.98; 95% CI, 620.64-992.84).

Mortality rates of All influenza illnesses were 0.05% during the December 1917 and April 1918 mortality peaks (Figure 2A ). Most deaths during these times (98% in December 1917 and 99% in April 1918) were due to Influenza complicated by pneumonia ( Figure 2A ). Mortality rates of All influenza illnesses peaked again during the October 1918 pandemic at 0.85%, with nearly all deaths (98%) attributable to Influenza complicated by pneumonia ( Figure 2A) . Significantly, unlike the earlier periods, the largest proportion of these deaths (48%) was due to Influenza and bronchopneumonia ( Figure 2B ).

Odds of mortality in September-October 1918 relative to December 1917-April 1918 were increased in each of the 7 original diagnostic categories (Table 1) . Mortality odds for Influenza complicated by pneumonia in September-October 1918 relative It was not possible to comprehensively examine the "W-shaped" mortality curve [2] in this population because most of the soldiers were young adult men.

Case-fatality rates of All influenza illnesses remained approximately 1 percent throughout winter-spring 1917-1918, peaking in November 1917 at 1.4% and then again in the September-October 1918 pandemic onset at 5.2% (Figure 3 ). Increased case-fatality odds were observed in 6 of the 7 original diagnostic categories, the exception being Influenza and bronchopneumonia (Table 1) . Case-fatality odds were most elevated for the individual categories Common respiratory diseases (OR, 18.41; 95% CI, 11.25-30.13) and Influenza, uncomplicated (OR, 6.54; 95% CI, 2.99-14.31). Misclassification of 1918 pandemic influenza cases into the Common respiratory diseases category during September-October 1918 likely accounted for the observed increase in case-fatality odds in this category. Although mild cases of illness in the Common respiratory diseases category and in the Influenza, uncomplicated category were supposed to be reclassified according to subsequent complications per US Army guidelines [1] , this clearly this was not always accomplished. Increased case-fatality odds were most prominently observed in the combined category of Influenza not complicated by pneumonia (Table 1) given that this category is a combination of Common respiratory diseases and Influenza, uncomplicated, and All influenza illnesses for which the majority of cases derive from the Common respiratory diseases and Influenza, uncomplicated categories.

Relative to the prepandemic December 1917-April 1918 influenza season, the odds of death among soldiers admitted for All influenza illnesses during September-October 1918 were approximately 5-fold higher (OR, 4.90; 95% CI, 4.66-5.15) ( Table 1) , as were the odds of dying when comparing the March-April 1918 period alone (OR, 4.92; 95% CI, 4.58-5.28). However, it is noteworthy that soldiers admitted with Influenza and bronchopneumonia in September-October 1918 were no more likely to die than those admitted with the same diagnosis in December 1917-April 1918 (OR, 1.15; 95% CI, 0.65-2.04) ( Table 1) .

The fatal complication rate from Influenza and bronchopneumonia among individuals with Specified influenza infection (Influenza, uncomplicated; Influenza and lobar pneumonia; Influenza and bronchopneumonia; and Influenza and other complications, combined) peaked at 0.06% in January 1918 and dropped to 0.02% in March 1918 ( Figure 4 ). However, during the 5 months preceding the pandemic peak (April-August 1918), the fatal complication rate from Influenza and bronchopneumonia rose steadily despite low incidence of Specified influenza infection among the general troop population during late spring and summer (May-August 1918) (Figure 4) . The fatal complication rate from Influenza and bronchopneumonia peaked in October 1918 at 2.8% and remained elevated through February 1919 in the 2%-3% range, even as influenza admissions returned to very low levels. Among individuals with Specified influenza infection, the fatal complication rate from Specified influenza and pneumonia (Influenza and bronchopneumonia and Influenza and lobar pneumonia combined) followed a pattern similar to that of Influenza and bronchopneumonia alone, except during February-April 1918. During February-April 1918, elevated fatal complication rates from Influenza and lobar pneumonia contributed to high fatal complication rates from Specified influenza and pneumonia (Figure 4 ).

Analyses of respiratory disease admission incidence, mortality, case fatality, and complications in 39 domestic US Army training camps were performed to characterize patterns of respiratory illnesses and deaths among US soldiers during World War I. We examined 7 original respiratory diagnostic categories including 4 pneumonia subcategories, the latter singly and combined.

Separating and combining the 4 pneumonia categories in some analyses was considered necessary because of the possibility that physicians could have missed diagnosing antecedent influenza in soldiers presenting with Lobar pneumonia or with Bronchopneumonia, and because of potential misclassifications between Lobar pneumonia and Bronchopneumonia. The data showed 3 distinct peaks of respiratory disease admission incidence preceding and during the pandemic onset. Mortality rates from All influenza illnesses in December 1917 and in April 1918 were both 0.05%. By comparison, mortality rates from All influenza illnesses in September-October 1918 were significantly higher (0.85%), and they were consistent with age-specific pandemic mortality rates in 1918 US civilian populations [2] .

Excess mortality in September-October 1918 resulted from a combination of high influenza admission rates, more frequent case progression to secondary bacterial pneumonia, especially bronchopneumonia, and high but typical pneumonia casefatality rates. During the 1918 pandemic, many military and civilian observations made it clear that severe and fatal postinfluenza pneumonia was linked with high frequency to bronchopneumonia specifically [12] . During September-October 1918, the odds of being admitted with Influenza complicated by pneumonia were approximately 10-fold higher than in December 1917-April 1918 (OR, 9.45; 95% CI, 9.25-9.66), even though the odds of dying from Influenza complicated by pneumonia, once it occurred, were only modestly elevated (OR, 1.77; 95% CI, 1.68-1.88). These data support the notion, widely cited in the medical literature of the 1918 era [13] , that high pandemic influenza mortality rates during fall 1918 resulted primarily from increased frequency of postinfluenza bacterial pneumonia rather than increased lethality of such pneumonias, and further that the case fatality of postinfluenza pneumonia in 1918-1919 was not remarkably different from general pneumonia casefatality rates observed in the years before and after the pandemic [13, 14] .

Of the 4 pneumonia categories, the role of Influenza and bronchopneumonia in excess influenza mortality in September-October 1918 stands out. Soldiers were approximately 785 times more likely to develop Influenza and bronchopneumonia during September-October 1918 relative to December 1917-April 1918 (OR, 784.98; 95% CI, 620.64-992.84), even though they were no more likely to die from it (OR, 1.15; 95% CI, 0.65-2.04). Experience with highly fatal measles epidemics in the same US Army training camps in the months immediately preceding the 1918 influenza pandemic had shown that postviral bronchopneumonia was a fundamentally different disease than lobar pneumonia. The 1917-1918 Army measles and the 1918 influenza epidemics advanced understanding of the natural history and pathogenesis of postviral bronchopneumonia as a virally induced cytolytic process that begins in the upper airway, extends to the lower airway, becomes complicated by invasive bacterial infection, and in some cases manifests as a confluent infectious process resembling lobar pneumonia clinically or radiographically [15] .

The slow spring-summer 1918 increase in the percent of All influenza illnesses attributable to Influenza complicated by pneumonia ( Figure 1A) , and the concomitant rise in the proportion of cases and deaths of Influenza complicated by pneumonia attributable to Influenza and bronchopneumonia ( Figures 1B and 2B) , could indicate early emergence of the pandemic virus, which was later shown, during the fall 1918 pandemic peak, to be associated with a high rate of severe and fatal postinfluenza bronchopneumonia. During the spring months of 1918, military providers recognized and evaluated rare cases of severe pneumonia that retrospectively fit the clinical and epidemiologic pattern of early pandemic viral emergence on the background of outbreaks of mild influenza-like illnesses. These data seem consistent with the presence of at least 2 epidemiologically and clinically distinct respiratory diseases occurring between October 1917 and March 1919: (1) low-to intermediate-incidence influenza-like illnesses with few pneumonia complications and low mortality, peaking in January and April 1918, and (2) high-incidence pandemic influenza associated with a high rate of pneumonia complications, but with typical postinfluenza pneumonia case-fatality rates, peaking in September-October 1918. The low rates of postinfluenza complications and deaths during December 1917-April 1918 suggest that the predominant virus circulating in the prepandemic months was distinct from the September-October 1918 pandemic virus.

The initial recognized appearance of the Spanish influenza pandemic in Northern Europe in summer 1918 [16] suggests that the pandemic virus must have been circulating below the level of detection well before emergence. However, the data examined here do not strongly support the possibility that the pandemic virus was the cause of most influenza-like illnesses in US Army camps in spring 1918, which some have called "a" or "the" "spring wave" [17, 18] . Unlike the pandemic that emerged in summer-fall 1918, spring influenza cases tended to be mild, not complicated by pneumonia, and rarely fatal; the spring outbreaks were limited in progression, even in intensely crowded military camps during a season (winterspring) historically favorable to influenza spread [19] .

Alternative explanations for the limited spring outbreaks of mild influenza-like illnesses that occurred sporadically and disappeared quickly on the US East Coast, in parts of Europe, and in US Army camps include the possibility that the responsible viruses were noninfluenza respiratory viruses [9] ; that they were seasonal or endemic influenza viruses; that the responsible virus was the 1918 pandemic virus of stable virulence at the time, whose later high pathogenicity, in the 1918-1921 period, reflected unknown and transient nonviral cofactors; or that early in its emergence, at some time after spring 1918, the pandemic virus mutated to enhanced virulence. This latter possibility appears problematic in several respects, including early (May 1918) influenza viral hemagglutinin gene sequences that are little-changed from the sequences of viruses studied at the pandemic peak, and which were derived from cases with clinical courses and histopathologic characteristics indistinguishable from later pandemic cases; [20] and the seemingly spontaneous pandemic emergence within a narrow window of time in summer-fall 1918, in a pathogenic form widely dispersed geographically, consistent with global pre-seeding of the virus [21] .

Although none of the 4 emergence scenarios mentioned above can be conclusively ruled out, the data presented here appear most consistent with the possibility that the December 1917-February 1918 and the March-April 1918 influenza-like illness activities were both largely due to 1 or more unrelated influenza viruses, exhibiting properties of typical seasonal influenza viruses, their lowly pathogenic outbreak properties masking the very slow, simultaneous, emergence of a pathogenic but low-incidence pandemic virus ( Figure 4 ). The data also appear consistent with the notion that this latter virus remained almost undetectable in its mortality impact (Figure 2A ) while nevertheless causing rare, but increasingly more frequent, cases of severe and fatal bronchopneumonia in the epidemiologic background, consistent with an emerging pandemic virus (Figure 4) .

Others have observed an association between presumed respiratory viral exposure in spring 1918 and reduced rates of infection and death in fall but not winter 1918 [22] . This apparent protection might be attributable to inter-or intra-subtypic influenza humoral immunity, T-cell immunity, or short-term innate antiviral responses. Viral genetic and immunologic data derived from the 1918 era would be required to reliably address this issue as ecological studies, where inferences about the nature of individuals are deduced from inference for the group are subject to significant bias.

Attempts to understand these decades-old pandemic phenomena are of importance in our efforts to understand pandemic emergence today. Simple mathematics indicates that pandemic viruses must circulate in 1 or more human populations for months before emergence, allowing an opportunity for early detection. This finding is consistent with observations made since the 1957 pandemic. Detecting such pandemic viruses when they are still circulating at a low prepandemic level is a critical public health goal because it would potentially provide significantly more time to develop vaccines and formulate prevention and mitigation strategies.

Our interpretation of the 1917-1919 US Army data suggests the theoretical feasibility of detecting a modern pathogenic prepandemic virus by aggressive surveillance in selected human populations during nonpandemic periods. Viewed in retrospect, the 1917-1919 Army data seem to be consistent with a "signal" of an approaching pandemic 5 months before it appeared when, in spring 1918, rare cases of severe and fatal Influenza and bronchopneumonia began to be detected in the midst of outbreaks of mild influenza-like illnesses. Had this happened in the modern era, severe cases of respiratory viral illness detected in emergency rooms, inpatient hospital wards, or intensive-care units could be confirmed by viral culture or molecular diagnostic assay. Influenza A virus subtypes, other than those predominantly circulating, could be rapidly genetically sequenced, allowing detection of a novel influenza virus. Early detection would lead to enhanced virus-specific surveillance and to public health strategies to prevent and control pandemic emergence (ie, vaccine development and distribution).

The importance of detecting pandemic viruses as early as possible in the course of their months-long emergence cannot be overstated. Early detection shortens the lag time in vaccine development, policy and strategy development, and education of providers, the public, and other key groups. Understanding how pandemic viruses emerge and development of early detection strategies are key elements of effective prevention and control. The 1917-1919 data presented here should stimulate additional discussions aimed at optimizing pandemic detection and response.

Inflammatory diseases of the respiratory tract (bronchitis; influenza; bronchopneumonia; lobar pneumonia)

Influenza: the mother of all pandemics

Predominant role of bacterial pneumonia as a cause of death in pandemic influenza: implications for pandemic influenza preparedness

A virus obtained from influenza patients

Initial genetic characterization of the 1918 "Spanish" influenza virus

Origin and evolution of the 1918 "Spanish" influenza virus hemagglutinin gene

Reconstruction of the 1918 influenza virus: unexpected rewards from the past

US Government Printing Office

Pathology of the acute respiratory diseases. I. In camps in the United States

The Medical Department of the United States Army in the World War

Theory and Problems of Probability and Statistics

Pathological anatomy associated with influenza; 23 plates

Age and sex incidence of influenza and pneumonia morbidity and mortality in the epidemic of 1928-29 with comparative data for the epidemic of 1918-19

A study of the incidence of the types of pneumococci isolated from acute lobar pneumonia and other infections, and an analysis of the cases classified by types in regard to mortality, complications, associated diseases, bacteremia, and leukocytosis

The pathology of the pneumonia in the United States Army camps during the winter of 1917-18. Monograph of the Rockefeller Institute for

Epidemiologic characterization of the 1918 influenza pandemic summer wave in Copenhagen: implications for pandemic control strategies

Epidemic Influenza: A Survey

Mortality patterns associated with the 1918 influenza pandemic in Mexico: evidence for a spring herald wave and lack of preexisting immunity in older populations

Pandemic influenza: certain uncertainties

Autopsy series of 68 cases dying before and during the 1918 influenza pandemic peak

Influenza A pandemics of the 20th century with special reference to 1918: virology, pathology and epidemiology

Cross-protection between successive waves of the 1918-1919 influenza pandemic: epidemiological evidence from US Army camps and from Britain

Influenza Circulation in United States Army Training Camps Before and During the 1918 Influenza Pandemic • OFID • 9

Financial support. This work was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health.Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.