Influenza and tuberculosis co‐infection: A systematic review

Abstract Introduction There are limited data on risk of severe disease or outcomes in patients with influenza and pulmonary tuberculosis (PTB) co‐infection compared to those with single infection. Methods We conducted a systematic review of published literature on the interaction of influenza viruses and PTB. Studies were eligible for inclusion if they presented data on prevalence, disease association, presentation or severity of laboratory‐confirmed influenza among clinically diagnosed or laboratory‐confirmed PTB cases. We searched eight databases from inception until December 2018. Summary characteristics of each study were extracted, and a narrative summary was presented. Cohort or case‐control studies were assessed for potential bias using the Newcastle‐Ottawa scale. Results We assessed 5154 abstracts, reviewed 146 manuscripts and included 19 studies fulfilling selection criteria (13 human and six animal). Of seven studies reporting on the possible effect of the underlying PTB disease in patients with influenza, three of four analytical studies reported no association with disease severity of influenza infection in those with PTB, whilst one study reported PTB as a risk factor for influenza‐associated hospitalization. An association between influenza infection and PTB disease was found in three of five analytical studies; whereas the two other studies reported a high frequency of PTB disease progression and complications among patients with seasonal influenza co‐infection. Conclusion Human analytical studies of an association between co‐infection and severe influenza‐ or PTB‐associated disease or increased prevalence of influenza co‐infection in individuals' hospitalized for PTB were not conclusive. Data are limited from large, high‐quality, analytical epidemiological studies with laboratory‐confirmed endpoints.


| INTRODUC TI ON
Influenza virus infections cause substantial annual morbidity and mortality in humans worldwide. [1][2][3] Globally, it is estimated that annual influenza epidemics result in three to five million cases of severe illness and between 290 000 and 650 000 influenza-associated respiratory deaths. 4,5 In 2015, there were an estimated 10.4 million incident cases of tuberculosis and 1.8 million tuberculosis deaths globally. 6 In 2015, tuberculosis was the most common cause of infectious disease-related deaths worldwide, with the majority of cases reported in Asia and Africa. 6 Both influenza and tuberculosis impair host immune responses.
Specifically, influenza can impair T-cell immunity and weaken innate immune responses against secondary bacterial infections. [7][8][9][10][11][12] Lethal synergism associated with viral and bacterial infections can result in increased risk of influenza-associated mortality. 13 Furthermore, individuals with pulmonary tuberculosis (PTB) may be at increased risk for severe influenza disease and death due to chronic lung disease and immunosupression. Ecological studies and mathematical modelling of epidemiologic data suggest an increase in the frequency of influenza disease or severe influenza-associated disease in individuals with PTB during influenza pandemics [14][15][16][17][18] or during seasonal influenza epidemics 19 compared with otherwise healthy individuals.

Influenza infection may facilitate the progression of latent
Mycobacterium tuberculosis infection to tuberculosis disease and alter the clinical presentation of tuberculosis. 20 It is also possible that influenza infection may exacerbate PTB.
Whilst chronic lung diseases are a known risk factor for severe outcomes due to influenza infection and influenza vaccination is recommended in this group, PTB is not listed as a separate priority group. 21 Understanding the interaction between influenza and PTB may assist in determining whether individuals with PTB should be prioritized for influenza vaccination and treatment with antiviral medications. We conducted a systematic review of published literature on the association between laboratory-confirmed influenza and PTB, that is influenza in individuals with tuberculosis and tuberculosis in individuals with influenza infection, in order to summarize whether co-infection affects presentation, progression or disease outcome.

| ME THODS
We conducted a systematic review, which is reported in accordance with PRISMA guidelines, 22 to summarize whether individuals with co-infection present with severe influenza or PTB disease as compared to those with single infection or disease. Burden, transmission and severity of co-infection were included for completeness.

| Eligibility and inclusion criteria
This review was restricted to published abstracts and articles from inception to December 2018 that reported data on the association (burden of disease, transmission and severity) between laboratoryconfirmed influenza and clinically diagnosed or laboratory-confirmed PTB. Due to the scarcity of published data, descriptive studies, including studies without comparison groups, were included. Articles that included seasonal or pandemic influenza and animal experimental studies were also included. For human studies, inclusion was limited to studies in which influenza was laboratory-confirmed and tuberculosis included PTB. Animal studies were included as they may provide useful insights into possible underlying mechanisms of interactions in humans. Studies that modelled ecological data on the association between influenza and tuberculosis, individual case reports, vaccine studies and influenza antiviral therapy in patients with tuberculosis were not included. Study selection is summarized in Figure 1.

| Search strategy
We conducted a systematic review of the scientific literature identified through searches using online databases. For our search, we included terms for influenza ("influenza" or "flu" or "influenza virus" or "human influenza") and for tuberculosis ("tuberculosis" or "TB"). The Medline, Embase, PsycINFO, CINAHL, Web of Science, Cochrane, CAB Abstracts and Global Health databases were searched. The search strategy, which was completed in consultation with a research librarian, differed slightly by database (Appendix S1). In addition, bibliographies of papers that were reviewed were checked for further relevant publications. The search was restricted to articles published in English, French, Italian, German, Russian, Finish, Japanese or Portuguese.

| Study selection
Literature search results (titles and abstracts) were screened inde- Discrepancies in included articles were resolved by consensus between the two reviewers with involvement of a third reviewer (CC) where necessary. Animal experimental studies, descriptive and analytic studies in humans were included. Studies that reported data on the association between laboratory-confirmed influenza and clinically diagnosed or laboratory-confirmed PTB including the following were included: • Prevalence and risk for influenza-associated severe disease among patients with PTB disease; • Prevalence and risk for PTB-associated severe disease among patients with influenza infection; • Effect of influenza on PTB disease progression • Clinical presentation of influenza and PTB co-infection; and • Immune response to co-infection, presentation or outcome of influenza-tuberculosis co-infection in animal studies. Individual studies were independently assessed for potential bias or confounding. When studies used either cohort or case-control designs, we used the Newcastle-Ottawa Scale to rate the quality of the included papers. 23 Studies were considered high quality if the Newcastle-Ottawa Scale was ≥7 out of 9 and were considered of low quality if the score was ≤3 out of 9. Study methods differed; summary measures (odds ratios, relative risks), when reported, were abstracted. Data synthesis consisted of reporting the key findings of the different studies. Where possible the studies were classified according to whether they fall among the 22 high tuberculosis burden countries (HBC) that account for aproximately. 80% of world's tuberculosis cases. 3 Review protocol attached (Appendix S2).

| Ethics
Since this study used published data, it was exempt from human subjects ethics review.

| RE SULTS
The search identified 5752 records; 598 of these were duplicates and were removed. Seven additional records were identified through other sources ( Figure 1). The remaining 5154 titles and abstracts were screened. Of these, 146 articles were identified for full review, and 19 articles met the inclusion criteria. Of these, 13 were in humans and six were animal experimental studies.

| Human studies
Of the 13 human studies, 10 used real-time reverse transcription polymerase chain reaction (RT-PCR) and three used unpaired serology to test for influenza infection. A total of 27 566 individuals (range 19 24 -12 196 25 ) were included in human studies, 12 777 (range 31 26  Archer, Abbreviations: AFB, acid-fast bacilli; AOR, adjusted odds ratio; CPR: case-population ratio; lab, laboratory; NA, not applicable; OR, odds ratio; RT-PCR, real-time reverse transcriptase polymerase chain reaction; SARI, severe acute respiratory illness; TB, tuberculosis; USA, United States of America. a Association-Evidence of/or association (univariate/multivariable analysis) with increased severity of influenza disease in those with vs without tuberculosis; or prevalence of co-infection in those with Score out of a possible score of 9.
TA B L E 1 (Continued)

| The effect of PTB in patients with influenza
Of the seven studies that reported on PTB in individuals with influenza, six were from HBCs (Table 1)

| Descriptive studies
Three descriptive studies from HBC using data from the 2009 influenza pandemic reported the prevalence of PTB in individuals with influenza. Two of these studies reported a high frequency of tuberculosis (9% and 10%) in cases hospitalized with influenza and among influenza deaths relative to expected community prevalence.
However, no inferences could be made on the significance of the association as there were no comparison groups or data were not evaluated statistically. 34

| Analytical studies
Three of the four analytical studies were from HBC, including one of high quality that reported no association with severe disease among patients with influenza-PTB co-infection compared to patients with influenza only. 24,29,30 One analytical study reported PTB as a risk factor for influenza-associated severe acute respiratory illness (SARI) hospitalization. 27 In this case-population study from South Africa, tuberculosis was twice as prevalent among hospitalized influenza-associated SARI cases compared with the gen-

| The effect of influenza in patients with PTB
Of the seven papers that reported data on influenza in patients with PTB (Table 2), 26,28,[30][31][32][33]36 four were from tuberculosis HBCs and the other three were from Europe in a period with high tuberculosis prevalence. [30][31][32][33] Four of these papers were reported by the authors as analytical studies, [30][31][32]36 and a fifth 28 had data suitable for authors of this manuscript to review analytically.

| Descriptive studies
Two descriptive studies reported on influenza in cases with tuberculosis housed at a sanatorium. Of these, one study described the effect of seasonal influenza on tuberculosis disease progression and complications. 26

| Summary of quality of human studies
Of the eight analytical studies, three were high-quality studies as assessed by the Newcastle-Ottawa score, of which one showed an association between influenza-PTB co-infection and increased mortality compared with tuberculosis only 32 and two showed no association between co-infection and severe influenza disease 29 or correlation between influenza infection and tuberculosis. 31 Over a third of the studies about PTB and influenza virus co-infection were descriptive case series that included univariate analysis, and the causal relationship could not be demonstrated. Some of the studies used clinical criteria for PTB cases; however, the specifics of the criteria used were not always fully described. Among the studies that included laboratory-confirmed PTB, screening for tuberculosis was not done systematically.

| Summary of findings from experimental animal models
In murine models, five studies suggested that influenza and tuberculosis co-infection affected tuberculosis and influenza disease presentation or outcome, 20 and one study showed no effect (Table 3). 37 Five of the murine studies reported on the effect of influenza on tuberculosis, and one study reported on the effect of influenza on tuberculosis and the effect of tuberculosis on influenza. co-infected mice compared with the influenza only group). 38 Redford et al 39  and outcomes. Three of the descriptive studies, although not assessed for statistical significance, reported either a high prevalence of co-infection in cases with severe influenza disease 34,35 or increased severe disease or progression of disease in co-infected individuals. 26 Of the five studies reporting on pandemic influenza only, two descriptive studies from HBCs reported a high prevalence of tuberculosis in cases with severe influenza-associated disease. 34 Since the 1950s, authors have recommended influenza vaccination among patients with tuberculosis during influenza epidemics. 28 Influenza vaccination is the most effective way to prevent influenzaassociated disease. Influenza vaccine has been shown to generate antibody response in patients with tuberculosis that is similar to those without tuberculosis, although these studies were conducted in the 1950s and 1960s and did not include HIV-infected individuals. 36 Antiviral treatment for influenza improves outcomes for patients with severe influenza-associated disease. 46  Some of the studies reported on tuberculosis in patients from sanatorium. 26,28,36 It is possible that the high prevalence of influenza reported in these studies is due to increased risk of influenza transmission resulting in high transmission rates in these closed settings. In addition, the influenza transmission may not reflect community-acquired influenza and results from these studies cannot be generalizable to other settings. Due to a possibility of increased risk of high concentration of persons with co-morbidities resulting in poor outcomes, closed settings should be prioritized for influenza vaccination.

| Effect of influenza on tuberculosis
There were a number of limitations to this systematic review. Broad search terms were used to increase sensitivity to identify relevant articles, although this may have somewhat reduced search specificity. Over a third of the observational studies were descriptive, and due to the nature of these studies, an association could not be evaluated. The type of tuberculosis included differed among the studies, with some studies reporting newly diagnosed tuberculosis, some reporting on cases in a tuberculosis sanatorium for a number of months and some included cases who had completed tuberculosis treatment, thus making data less comparable. There were differences in the population tuberculosis incidences where studies were conducted which could affect the power to detect an association. However, the majority of studies were from tuberculosis high burden countries or were conducted during the period when tuberculosis burden was high. We included animal studies although these may not be generalizable to humans.
Many studies did not adequately assess the underlying conditions such as HIV and malnutrition. HIV infection is a risk factor for severe influenza disease as well as for PTB, and it is an important contributor to the overall burden of severe influenza in high HIVprevalence settings. 50,51 However, only a few papers reported data on HIV infection. 27,29,30,32 In one study, patients with co-infection of HIV and PTB were at high risk of being hospitalized with influenza; however, the number of co-infected individuals was low and the association was not statistically significant. 29 If the association with se-

| CON CLUS ION
Although the majority of experimental animal studies suggested increased severity of disease with co-infection of influenza and PTB, only half of the analytical studies on influenza and PTB in humans found the same. Descriptive studies, although they could not evaluate an association, reported an increased prevalence of co-infection among cases with severe influenza or PTB disease. Data are limited from large epidemiological studies, studies with laboratory-confirmed influenza and PTB, studies from high tuberculosis burden settings and studies that include data on HIV. In order to study the association between influenza and PTB and make inferences about causal associations, more epidemiological studies with systematic testing for influenza and tuberculosis are needed.

ACK N OWLED G EM ENTS
We thank Professor Anne von Gottberg for translating the papers published in German and Tatiana Baranovich for translating the abstracts published in Russian.

CO N FLI C T O F I NTE R E S T
No authors have any competing interests.