Influenza in patients with cancer after 2009 pandemic AH1N1: An 8‐year follow‐up study in Mexico

Abstract Background Immunosupressed patients are at high risk of influenza‐related complications. Influenza AH1N1 has been hypothesized to induce worse outcomes in patients with malignancies, but after the A(H1N1)pdm09 few publications have analyzed the presentation and complications related to influenza afterward. Objectives We aimed to describe the characteristics, risk factors, and outcomes of influenza in an oncologic center after the 2009 pandemic and to compare our case distribution to the National community acquired influenza databases in Mexico and the United States. Methods We reviewed the cases of confirmed influenza in patients with cancer from an oncological center in Mexico from April 2009 to April 2017. Data on severity and influenza type, malignancy, comorbidities, and outcomes were recorded. We correlated data between the Centers for Disease Control and Prevention (CDC) in the United States and SISVEFLU (Influenza Surveillance Program) in Mexico. Results One hundred eighty‐eight patients were included; 75 (39.9%) had a solid neoplasm and 113 (60.1%) had hematologic malignancies. AH1N1 was the most frequent influenza type (54.2%). Patients with hematologic malignancies had more pneumonia (55% vs 25%, P < .001), needed more hospitalizations (75% vs 39% P < .001), had higher all‐cause mortality at 30 days (20% vs 9% P = .048) and influenza‐associated mortality (17% vs 7% P = .041). Thirty (16%) patients died within 30 days, and 24 (12.7%) were related to influenza. Influenza type was not associated with worse outcomes. Yearly occurrence of influenza reported by the CDC and SISVEFLU showed a significant correlation (ρ = 0.823, P = .006). Conclusions AH1N1 was the dominant serotype. Patients with hematologic malignancies had more severe influenza and presented worse outcomes. Annual SISVEFLU and CDC surveillance information showed a similar distribution of cases along time but influenza serotypes did not match for all seasons.


| BACKG ROU N D
Influenza viruses can spread rapidly throughout the world during epidemics and yearly seasonal outbreaks in regions with temperate weather. During these outbreaks, 5%-10% of all adults may acquire the disease, causing serious illness in 3-5 million persons each year, and between 250 000 and 500 000 deaths all over the world. 1 In immunocompromised patients, influenza closely parallels the occurrence of the infection in the community, 2 with higher rates of pneumonia and infection-related mortality. [3][4][5][6][7][8][9] In 2009, a new pandemic influenza virus A(H1N1)pdm09 was identified in Mexico, which rapidly spread all over the world. 1,4,10 Among general population, infection with the AH1N1 virus is usually associated with mild respiratory disease with little hospitalization or death 7,8 ; however, it has been reported that influenza AH1N1 infection can cause a more severe disease than the seasonal influenza virus. 2 In Mexico, the AH1N1 virus has been accompanied by higher mortality rates and a cyclic biannual behavior-alternating with the AH3N2 virus. 1 Globally, during the 2009 A(H1N1)pdm09 pandemic, this virus caused 84.1% of all influenza infections, approximately 18 000 deaths and an overall case-fatality rate lower than 0.5%. 3,7,10,11 Patients with cancer and those with immunosuppression are at high risk of influenza-related complications. 1,6,12 These patients were particularly prone to develop serious complications and adverse outcomes at the beginning of the A(H1N1)pdm09 pandemic.
The AH1N1 strain has been hypothesized to induce worse outcomes in patients with malignancies. Early studies in patients with hematologic malignancies (HM) and hematopoietic cell transplant demonstrated higher rates of hospital admissions, pneumonia, intensive care unit admissions, and mortality 2,3,7,8,10 ; nonetheless, there is insufficient evidence on whether this strain induces worse outcomes. 8,13 Few studies have reported no difference in outcomes when comparing AH1N1 and other types of influenza in patients with cancer. 2,4,10 During the peak of influenza season, the rate of pneumonia in patients with cancer varies between 21% and 80%, and fatality rates range from 11% to 38%. 4,8,9,[14][15][16] Patients with HM and bone marrow transplant remain at high risk of severe influenza infections. 8,13,15,17,18 The underlying disease, chemotherapy related toxicity, immunosuppression caused by medications and poor vaccination rates are all contributing factors for infection and severe forms of the disease in this subset of patients.
After the A(H1N1)pdm09 pandemic, limited studies on cancer patients were published 5,8,15,16,19 and very few publications have reported the presentation and complications related to influenza afterward. The aim of this study was to describe the epidemiology, characteristics, and outcomes of influenza of eight consecutive years in a cohort of patients with cancer and to compare the yearly trends for the same time lapse with community acquired influenza data published by the United States' Centers for Disease Control and Prevention (CDC) and Mexico's Influenza Surveillance (SISVEFLU).

| Setting and design
We conducted a retrospective analysis of patients with cancer and confirmed influenza at the Instituto Nacional de Cancerologia (INCan) during the study period. INCan is a 136-bed referral, teaching hospital for adolescents and adult patients with cancer, in Mexico City. Since the A(H1N1)pdm09 pandemic a prospective surveillance program for patients with influenza-like illness was established, all hospitalized patients and those with a high suspicion of influenza are evaluated by an Infectious Diseases specialist. We searched the influenza database for all patients with cancer and confirmed diagnosis of influenza, along with the electronical medical record of each patient. We collected the following data: demographics (age and gender), comorbidities, underlying malignancy, oncological treatment, clinical presentation, radiologic characteristics, laboratory data, co-infections, antiviral treatment, and outcomes.

| Definitions
A case of influenza was defined as a patient with either a solid malignancy or an HM (leukemia, lymphoma, multiple myeloma, and others) having acute respiratory illness and a positive real-time polymerase chain reaction (PCR) assay. The infection was classified as community acquired if the onset of symptoms occurred before or within the first 2 days after hospital admission, or nosocomial if the symptoms developed any time after that during hospitalization. Upper respiratory infection (URI) was characterized by onset of rhinorrhea, nasal/ sinus congestion, pharyngitis, or cough with or without expectoration, and its diagnosis was confirmed by PCR. Lower respiratory infection (LRI) or pneumonia was characterized by new or changing pulmonary infiltrates suggestive of viral etiology on chest imaging and confirmed via examination of respiratory specimen, including nasal washes, endotracheal tube aspirates, sputum specimens, and bronchoalveolar lavage fluid specimens. A co-infection was considered when another organism was isolated from a patient at the time of influenza diagnosis. A concurrent infection was defined when another organism was isolated from a patient within 3 days after the diagnosis of influenza. Non-infectious complications were also considered. Acute kidney injury was defined as specified by the AKIN classification (ie, sudden decrease [in 48 hours] of renal function, expressed by an increase in absolute serum creatinine of at least 0.3 mg/dL). Hypoxia was defined as ambient air blood oxygen saturation <90% by pulse oximetry at diagnosis. Severe neutropenia was considered as an absolute neutrophil count <500 cells/mL and severe lymphocytopenia as an absolute lymphocyte count <200 cells/ mL. Elevated creatinine was defined as creatinine serum levels over 1.2 mg/dL and hypoalbuminemia as albumin serum levels <3.5 g/dL.

| Statistical analysis
Statistical analysis was performed with Stata version 12.0 (StataCorp LP). Categorical variables were compared using the chi-square test or Fisher's exact test. Continuous variables were compared using the Mann-Whitney U test or the Student's t test. Logistical regressions were conducted for all outcome categorical variables including known risk factors to obtain Odds Ratios on both, unadjusted and adjusted models.
Multiple Kaplan-Meier survival analysis was performed, and Log-rank test results were included when group comparison was needed. Finally, we used Pearson's correlations to measure associations between the CDC's, SISVELU's, and our own yearly influenza reports. All reported P values were two-sided and set with significance levels at <.05. From the 114 patients that required hospitalization (  (31% vs 10% P = .023) and neutropenia (24% vs 7% P = .046) ( Table 2).

| Outcome analysis
Risk factors associated with pneumonia after adjusting for dyspnea, lymphocytopenia, and creatinine >1.2 mg/dL are shown in Table 3.

| Survival analysis
We performed Kaplan-Meier curves on survival at 30 days. Patients with solid malignancies had a greater survival probability than patients with HM, except for patients with leukemia (P < .001) (Figure 2).

Influenza type
Survival probability of patients with AH1N1 did not differ from those with other types of influenza (P = .14) (Figure 3). Treatment with oseltamivir <48 hours after initiation of symptoms did not change the survival probability (P = .85) as shown in Figure S1.

| Influenza trends
Yearly occurrence of influenza cases reported by the CDC followed a similar trend to that reported by the SISVEFLU in Mexico, with a significant correlation between both data sets (ρ = 0.823, P = .006) ( Figure 4). Similarly, cases of influenza occurring in our hospital

| D ISCUSS I ON
In this study, we showed that oncologic patients, particularly, patients with HM had more severe forms of influenza, as has previously been described. 15 They also showed lower probability of survival compared with patients with solid neoplasm. No significant differences between years or subtype of influenza were observed, and influenza A H1N1 was not associated with worse outcomes. The season-to-season variability of influenza is well described, but only a few studies analyzed various consecutive years and outcomes related to influenza in patients with HM and solid tumors.
Symptoms presented by our patients were similar to those described in non-immunosuppressed patients, [3][4][5]7,11,13,15,20  In this cohort, patients with HM, lower albumin levels, and hypoxia at diagnosis were at an increased risk for LRI in the multivariable analysis. Pneumonia was developed in 43%, with a higher proportion in patients with HM (55% vs 25%) (P < .001). HM patients presented more frequently with severe lymphopenia (HM 31% vs solid malignancies 10%, P = .023) and severe neutropenia (HM 24% vs solid malignancies 7%, P = .046), consistent with data from other series 3,4,8,10-13 ; however, this association was not significant in the multivariable logistic analysis.
Decreased albumin level has been associated with LRI. 5,10 In this series, hypoalbuminemia was associated with pneumonia, which in part reflects the chronic and more advanced stage of disease presented in patients with cancer at our Institution. We also found that hypoxia was as an independent risk factor associated with pneumonia, a characteristic that has previously been described in patients with HM. 5 Hypoxia has also been linked with greater mortality 5,8 ; but in our study, it was not a predictor of death by multivariate analysis. It is difficult to explain the lack of association between hypoxia and mortality but it is probably related to the fact that other variables such as the underlying malignancy and its advanced stage in most patients remained the most important predictors of mortality. Hypoxemia is probably a marker of severe disease, but not a predictor of adverse influenza-related outcomes.
Obesity has been described as a risk factor for influenza, along with severe complications and death. Also, obese patients shed influenza virus for longer periods of time compared with non-obese individuals. 21 In our cohort, although obesity was prevalent, it did not increase the risk of pneumonia or death, but was correlated with dyspnea and hypoxemia, which is consistent with ventilatory dysfunction and increased risk of mechanical ventilation and death.

F I G U R E 2
Thirty-days survival probability in patients with influenza and solid or hematologic malignancies Delayed initiation of antiviral therapy has been correlated with pneumonia and mortality. 2,6,7,10,16 In this analysis, patients that received oseltamivir after 3 days had a greater influenza-related mortality only on the univariate model, but the same variable had no effect on the development of pneumonia. Likewise, in the survival analysis, probability at 30 days was unaffected by the delayed initiation of antiviral therapy; 58.5% of our patients received oseltamivir during the first 3 days of symptoms onset.
Overall, influenza mortality in patients with cancer varies between 22% and 40%. 11,16 In our cohort, 30-day mortality was 16% (n = 30) and influenza-related mortality was 12.7% (n = 24 As an original result from our analysis, we investigated a possible correlation between yearly national cases of influenza (SISVEFLU) and our data, and the cases occurring in the United States, in agreement to the proposed relationship between circulation of influenza in immunocompromised patients and general population from the northern hemisphere. 18 As shown in Figure 4, there was a correlation between the number of influenza cases occurring in the United States and Mexico; however, the magnitude of the association was mostly determined by the A(H1N1)pdm09 pandemic. Also, the CDC's reports show a yearly and continuous larger proportion of AH3N2 in the United States in comparison with Mexico's influenza subtypes, which alternate biannually between AH1N1 and AH3N2. 1 The less frequent occurrence of H3N2 and an almost perfect yearly alternating behavior between AH1N1 clearly deserves a further and more detailed analysis between the association of subtypes found in Mexico and the United States.
Our study has several limitations since it is a single center observational study, and although most data were collected prospectively, the analysis was conducted retrospectively. Despite the fair number of patients with influenza and cancer, and information from 8 years, we still included more patients with influenza AH1N1 than H3N2 or influenza B, which probably decreased the power to find differences between the outcomes caused by the different types of influenza.
Our study might also have some selection bias, as ambulatory patients with mild influenza were probably treated symptomatically with no further testing for influenza. There is also the possibility that patients with mild symptoms requested medical attention in a different health-care facility. Despite these limitations, this series shows the spectrum of influenza in patients with cancer across a long period of time in Mexico since 2009 when A(H1N1)pdm09 showed up in our country; we also compared our data with that of our national and the CDCs' surveillance system, a correlation that shows the annual trends of influenza seen in hospitals from Mexico City, and much likely, those located on the central and northern of Mexico, with temperate climates.
In conclusion, influenza AH1N1 was the dominant serotype of influenza in our patients, as it has been in the country since the A(H1N1) pdm09 pandemic with biannual peaks. SISVEFLU surveillance information was correlated with that of the CDC. Patients with hematologic malignancies had more severe influenza and presented worse outcomes, including a high-associated mortality. Influenza vaccination was low in this cohort despite being recommended in many international and national guidelines. The biannual presentation of influenza AH1N1 in Mexico and the less frequent occurrence of H3N2 compared with other countries deserves further investigation.

ACK N OWLED G EM ENTS
General support: Dirección General de Epidemiología del Sistema Nacional de Vigilancia Epidemiológica for the SISVEFLU data.

CO N FLI C T O F I NTE R E S T
Nothing to disclose.