Immunogenicity and clinical effectiveness of the trivalent inactivated influenza vaccine in immunocompromised children undergoing treatment for cancer

Abstract Influenza is associated with significant morbidity and mortality in children receiving therapy for cancer, yet recommendation for, and uptake of the seasonal vaccine remains poor. One hundred children undergoing treatment for cancer were vaccinated with the trivalent inactivated influenza vaccine according to national guidelines in 2010 and 2011. Influenza‐specific hemagglutinin inhibition antibody titers were performed on blood samples taken prior to each vaccination and 4 weeks following the final vaccination. A nasopharyngeal aspirate for influenza was performed on all children who developed an influenza‐like illness. Following vaccination, seroprotection and seroconversion rates were 55 and 43% for H3N2, 61 and 43% for H1N1, and 41 and 33% for B strain, respectively. Overall, there was a significant geometric mean fold increase to H3N2 (GMFI 4.56, 95% CI 3.19–6.52, P < 0.01) and H1N1 (GMFI 4.44, 95% CI 3.19–6.19, P < 0.01) strains. Seroconversion was significantly more likely in children with solid compared with hematological malignancies and in children <10 years of age who received a two‐dose schedule compared to one. Influenza infection occurred in 2% of the vaccinated study population, compared with 6.8% in unvaccinated controls, providing an adjusted estimated vaccine effectiveness of 72% (95% CI −26–94%). There were no serious adverse events and a low reactogenicity rate of 3%. The trivalent inactivated influenza vaccine is safe, immunogenic, provides clinical protection and should be administered annually to immunosuppressed children receiving treatment for cancer. All children <10 years of age should receive a two‐dose schedule.


Introduction
The last 40 years have seen marked improvements in the treatment for childhood cancer, with a significant reduction in mortality in developed countries [1][2][3][4]. For a number of childhood malignancies, increased cure rates have been achieved by increasing the intensity and duration of chemotherapy. However, this leads to prolonged periods of immunosuppression and vulnerability to infectious and toxic complications. In particular, influenza infection remains a significant cause of morbidity, mortality, and health expenditure among children undergoing treatment and within 6 months following the completion of therapy for cancer [5][6][7][8][9][10].

ORIGINAL RESEARCH
Immunogenicity and clinical effectiveness of the trivalent inactivated influenza vaccine in immunocompromised children undergoing treatment for cancer Abstract Influenza is associated with significant morbidity and mortality in children receiving therapy for cancer, yet recommendation for, and uptake of the seasonal vaccine remains poor. One hundred children undergoing treatment for cancer were vaccinated with the trivalent inactivated influenza vaccine according to national guidelines in 2010 and 2011. Influenza-specific hemagglutinin inhibition antibody titers were performed on blood samples taken prior to each vaccination and 4 weeks following the final vaccination. A nasopharyngeal aspirate for influenza was performed on all children who developed an influenza-like illness. Following vaccination, seroprotection and seroconversion rates were 55 and 43% for H3N2, 61 and 43% for H1N1, and 41 and 33% for B strain, respectively. Overall, there was a significant geometric mean fold increase to H3N2 (GMFI 4.56, 95% CI 3.19-6.52, P < 0.01) and H1N1 (GMFI 4.44, 95% CI 3.19-6.19, P < 0.01) strains. Seroconversion was significantly more likely in children with solid compared with hematological malignancies and in children <10 years of age who received a two-dose schedule compared to one. Influenza infection occurred in 2% of the vaccinated study population, compared with 6.8% in unvaccinated controls, providing an adjusted estimated vaccine effectiveness of 72% (95% CI −26-94%). There were no serious adverse events and a low reactogenicity rate of 3%. The trivalent inactivated influenza vaccine is safe, immunogenic, provides clinical protection and should be administered annually to immunosuppressed children receiving treatment for cancer. All children <10 years of age should receive a two-dose schedule.
Currently, annual vaccination with the inactivated influenza vaccine is recommended for all children undergoing treatment for cancer [11]. However, up to one-third of pediatric oncologists do not recommend yearly influenza vaccination [12,13], with poor uptake (27-55%) identified in children with cancer [14][15][16][17]. The lack of knowledge regarding benefit of the influenza vaccine in this population has been identified as one of the reasons for poor compliance [14,17]. The absence of literature correlating clinical outcome with immune response following influenza vaccination in children with cancer may reflect this lack of knowledge [18]. Given such findings, we undertook a prospective study to evaluate the immunogenicity and clinical effectiveness of the seasonal trivalent inactivated influenza vaccine in immunocompromised children receiving therapy for cancer, with the aim of providing evidence for annual influenza vaccination in this population and identifying risk factors that predict response.

Patient selection
Children between the ages of 6 months and 18 years who were receiving, or within 4 weeks from completion of immunosuppressive therapy for cancer were eligible. Recruitment was undertaken during the active influenza seasons of 2010 and 2011 (March-September) from the Department of Clinical Haematology and Oncology, Princess Margaret Hospital for Children (PMH) in Perth. PMH is the sole treatment center for all children with cancer in the state of Western Australia, which has a population of 2.6 million. Exclusion criteria included anaphylaxis to previous doses of any influenza vaccine, a history of egg anaphylaxis, receipt of intravenous immunoglobulin within the last 3 months, a neutrophil count of ≤0.5 × 10 9 /L, a history of Guillain-Barré syndrome and children having undergone autologous stem cell rescue or allogeneic hematopoietic stem cell transplant. Informed consent was obtained from the parents of each child prior to recruitment.

Study design
Patients were vaccinated according to national Australian standards [19]. Children <10 years of age, receiving influenza vaccine for the first time, were given two doses of the trivalent inactivated influenza vaccine 1 month apart. Children <10 years of age who had previously received influenza vaccine and children who were ten or older, were given a single dose of the vaccine. A 0.25 mL dose was administered to children <3 years of age, whereas those three or older were given 0.5 mL. The strains included in the vaccine for the 2010 and 2011 seasons were A/Perth/16/2009 (H3N2), A/California/7/2009 (H1N1), and B/Brisbane/60/2008 (B). Children were observed for 20 min after each vaccination for immediate vaccine-related adverse events. Parents were given a daily diary for documentation of vaccine-related adverse events occurring in the 7 days following vaccination and this information was collected from the parents at each visit.
Blood was taken prior to each vaccination and 4 weeks following the final vaccination to assess influenza-specific immune responses. Following collection, blood samples were centrifuged and sera stored at −20°C. At the end of each season, the samples were sent to the World Health Organization (WHO) Collaborating Centre for Reference and Research on Influenza, Victorian Infectious Diseases Reference Laboratory (VIDRL) where influenza-specific hemagglutinin inhibition (HI) antibody titers were performed against each vaccine strain by standardized assay [20].
Susceptibility to a vaccine-like strain was defined as a prevaccination HI titer of <40. Seroprotection in an individual was defined as a postvaccination HI titer of ≥40. Seroconversion was defined as either a fourfold increase in HI antibody titer if the prevaccination titer was ≥10 or a rise in HI titer from <10 to ≥40 following vaccination [21]. The percentage (95% confidence interval [CI]) of patients who individually met these criteria for seroprotection and seroconversion to each strain of the vaccine was calculated.
Criteria as established by the Committee for Proprietary Medicinal Products (CPMP) [22], were used to determine whether the vaccine was considered to elicit an effective overall immunogenic response in our immunocompromised population. According to these criteria, the influenza vaccine is considered effective if it meets one of the following three criteria: seroprotection in >70% of patients; seroconversion in >40% of patients; or a geometric mean fold increase (GMFI) >2.5. These criteria were used to calculate one-sided P values in relation to the null hypotheses for overall seroprotection and seroconversion to each strain of the vaccine. GMFI was calculated for each strain as the geometric mean of the fold increase in antibody level after vaccination, with 95% CI and one-sided P values estimated using a lognormal approximation for the distribution of antibody levels pre and postvaccination and the CPMP defined threshold of GMFI >2.5.
Multivariate logistic regression models were used to assess the influence of clinically relevant variables to predict seroconversion to each strain and complete seroconversion to all three strains. The variables analyzed within the model were sex (male, female), age groups Influenza Vaccination in Children with Cancer R. S. Kotecha et al. as determined by the stratification according to the vaccination schedule (<3 years, 3-<10 years, 10-<18 years), tumor type (solid, hematological), treatment intensity in the 4 weeks preceding vaccination (low, high; classified according to anticipated extent and duration of immunosuppression (see Table S1) and lymphocyte count for age (normal range, less than lower limit of normal). Lower normal limits for absolute lymphocyte counts according to age were defined as 1.7 × 10 9 /L for children <5 years of age, 1.1 × 10 9 /L for 5-<10 years, and 1.0 × 10 9 /L for ≥10 years [23]. The analyses were repeated on the subgroup of patients <10 years of age with the addition of vaccine doses received (one, two) as a variable in the model.
The standard procedure for all children with cancer in Western Australia who develop influenza-like illness, is to present for a clinical review and a nasopharyngeal aspirate is taken. Influenza-like illness is defined as an elevated temperature (≥37.5°C) or a clear history of fever (e.g., chills, rigors); the presence of at least one constitutional symptom from irritability, myalgia, headache, vomiting, diarrhea, or malaise; and the presence of at least one respiratory symptom from cough, sore throat, or rhinorrhea; with onset of symptoms occurring greater than 72 h after vaccine administration. The nasopharyngeal aspirates of all enrolled patients that were polymerase chain reaction positive for influenza were sent to the WHO Collaborating Centre for Reference and Research on Influenza, VIDRL, for culture and specific strain typing. Clinical effectiveness was assessed by comparing the proportion of laboratory confirmed influenza infection of vaccinated children on study with unvaccinated immunocompromised children receiving treatment for cancer who did not partake in the study at the time it was undertaken. Relative risk of infection comparing vaccinated and unvaccinated groups was estimated using log binomial regression, adjusting for age group and tumor type, with vaccine effectiveness calculated as 100 × (1 -relative risk). Clinical features of all children with cancer and laboratory proven influenza infection were documented and a qualitative assessment of clinical severity between vaccinated and unvaccinated patients was performed.
The multivariate analysis of predictive variables revealed that children with solid tumors were significantly more likely to serorespond to each vaccine strain (H3N2: OR 7.39, 95% CI 2.42-22.53, P < 0.01; H1N1: OR 2.90, 95% CI 1.02-8.23, P = 0.045; B: OR 3.75, 95% CI 1.25-11.24, P = 0.02) and to undergo complete seroconversion to all three strains (OR 6.03, 95% CI 1.56-23.29, P < 0.01) compared to patients with hematological malignancies. Children with a lymphocyte count in the normal range for age were significantly more likely to serorespond to B strain (OR 2.97, 95% CI 1.07-8.28, P = 0.04) than those with a low lymphocyte count at vaccination. Children who received low intensity therapy in the 4 weeks preceding vaccination were significantly more likely to serorespond to the B (OR 3.16, 95% CI 1.09-9.18, P = 0.03) and H3N2 (OR 2.81, 95% CI 1.00-7.89, P = 0.049) strains compared with those who received high intensity therapy, with a trend for seroconversion demonstrated for H1N1 (OR 2.36, 95% CI 0.91-6.09, P = 0.08). Table 3 shows the multivariate analysis of factors predicting seroconversion. Multivariate analysis for the subgroup of patients <10 years of age revealed that vaccine naïve children who received two doses of the vaccine were significantly more likely to serorespond to each strain (H3N2: OR 6.08, 95% CI 1.56-23.63, P < 0.01; H1N1: OR 6.03, 95% CI 1.74-20.90, P < 0.01; B: OR 14.72, 95% CI 2.80-77.36, P < 0.01), and to serorespond to all three strains (OR 14.71, 95% CI 1.27-170.2, P = 0.03), than children who had been vaccinated in previous seasons and only received one dose of the vaccine.
The incidence of laboratory proven influenza infection in the vaccinated study population was 2% (n = 2/100), whereas in the unvaccinated control population there was an incidence of 6.8% (n = 11/161), giving an adjusted estimated vaccine effectiveness of 72% (95% CI −26-94%). Of the children with laboratory proven influenza, there were 10 with acute lymphoblastic leukemia (ALL), including both vaccinated study patients, and individual patients with acute myeloid leukemia, Langerhans cell histiocytosis, and osteosarcoma. The two patients in the study had laboratory confirmed influenza B infection. Both patients failed to mount an immunological response to this strain with postvaccination HI titers <40. Unvaccinated children with influenza had an increased length of hospital admission (5.1 vs. 4 days, mean) and delay in the delivery of scheduled chemotherapy (4.5 vs. 0.5 days, mean) compared to vaccinated children with influenza. One unvaccinated control required supplemental oxygen for 3 days, however, there were no severe complications of influenza illness, such as admission to intensive care or death.
There were no vaccine-related serious adverse events. Reactogenicity, that was considered attributable to the vaccine, occurred in four children, who all developed fever within 24 h of receiving the vaccine, with no other cause identified. All four children required brief inpatient admissions for empiric antibiotic therapy until the fever subsided. The parent reported reactogenicity rate was 3% (n = 4/133 vaccinations).

Discussion
Influenza is a common respiratory pathogen associated with significant morbidity and mortality in children undergoing therapy for cancer [5][6][7][8][9][10], yet recommendation for and uptake of the seasonal influenza vaccine remains suboptimal [12][13][14][15][16][17]. Lack of knowledge regarding the benefit of the influenza vaccine in this population has been cited as a reason for poor uptake [14,17]. Our study provides evidence that the trivalent inactivated influenza vaccine is safe, immunogenic and provides clinical protection for immunosuppressed children receiving treatment for cancer.
Our study confirms that immunocompromised children receiving treatment for cancer are able to mount a clinically significant immune response to the trivalent inactivated influenza vaccine. A wide range for seroprotection and seroconversion following administration of the trivalent influenza vaccine has been reported in this setting (summarized in Table 4) [24][25][26][27][28][29][30][31][32][33][34][35]. In our cohort, immunogenicity was toward the middle of these ranges: seroprotection H3N2 55% (range from published literature: 25-92%), H1N1 61% (5-96%), and B 41% (15-87%); seroconversion H3N2 43% (25-78%), H1N1 43% (16-84%), and B 33% (12-60%). Although this is reassuring, comparison between studies should be interpreted with caution due to the differences in methodology, season, and vaccine composition. In particular, different vaccination schedules, differing definitions of seroprotection and seroconversion, inclusion restricted to variables such as tumor type, and a broad range of methods to conduct statistical analyses have been used. The studies have been conducted over a large chronological time span, thus variations in susceptibility, differing circulating influenza strains, and vaccine composition need to be considered. Our study is only the second performed in the Southern Hemisphere using Southern Hemisphere influenza vaccine formulation, with the other study undertaken more than 30 years ago [35].
The CPMP have defined criteria to assess whether influenza vaccines are effective within a population. The vaccine can be considered immunologically effective in our cohort as it satisfied criteria for GMFI. Although the vaccine did not meet the numerical threshold for seroprotection and statistical significance was not achieved for seroconversion, this should be taken in the context of how the criteria have been defined and not interpreted as vaccine failure. Although the CPMP criteria have been used to assess influenza vaccines in children, they are more specific to adults. In addition, the CPMP criteria have not been defined according to an immunocompromised population. This highlights the need for revised definitions, which take into account age and immune competence, to determine whether influenza vaccines are effective within specific populations.
Our study is the first to identify that children with solid tumors mount a significantly higher immune response to each strain individually and all three strains collectively compared to those with hematological malignancies. A previous study identified superior seroconversion to the H1N1 strain in children with solid tumors compared to those with ALL [25], and the study reporting the highest seroprotection and seroconversion rates in children with cancer was undertaken in children with solid tumors [29] ( Table 4). This difference can be explained by the direct adverse effect of hematological malignancies on the immune system, as well as therapy directed toward continuous myelosuppression for leukemia, whereas treatment for solid tumors is generally shorter and cyclical in nature. The influence of treatment on effector cells of the immune system has been examined previously, demonstrating prolonged suppression on the B-cell compartment in children receiving therapy for ALL compared with solid tumors [25].
Additional factors identified as predicting seroconversion from prior studies include higher white cell count, lymphocyte count, or IgG levels; increasing age; induction phase of therapy in ALL; and following completion of therapy (Table 4) [24,25,[29][30][31]34], which can all be considered correlates of underlying immune function. This is further emphasized by our results identifying a significantly superior seroconversion toward the B strain in children with a lymphocyte count in the normal range for age at vaccination and toward the B and H3N2 strains in children who received low compared with high intensity chemotherapy in the 4 weeks preceding vaccination. These correlates can be used as a guide to tailor the timing of vaccination on an individual basis, with optimal timing occurring prior to high intensity therapy, with normal lymphocyte counts and IgG levels in the patient. However, waiting for optimal conditions in an individual patient should not be at the expense of timely vaccination, given the risk of influenza to an unvaccinated immunocompromised child, the low chance of satisfying all predictive criteria at any one time in this population and that the vaccine remains safe and effective in children undergoing therapy for cancer despite immunosuppression.
Multivariate analysis of the subgroup of children <10 years of age has shown that vaccine naïve patients who received two doses of the vaccine were significantly more likely to seroconvert to each strain individually and all three strains collectively, than those vaccinated in  Seroprotection and seroconversion given for up to 18 patients receiving chemotherapy; analysis of predictive variables undertaken on 44 patients of which 26 were up to 60 months following the completion of therapy. 3 Includes one patient with acute myeloid leukemia and three with solid tumors. 4 Results shown for patients who received trivalent inactivated influenza vaccine containing H1N1, H3N2, and B strains.
Influenza Vaccination in Children with Cancer R. S. Kotecha et al. previous seasons receiving one dose. This finding has previously been limited to seroconversion to the H1N1 strain in a small group of children with ALL [32]. Our findings provide conclusive evidence to recommend that all immunocompromised children <10 years of age undergoing therapy for cancer should receive two age-appropriate doses of the vaccine, regardless of prior vaccination history.
We have demonstrated that the trivalent inactivated influenza vaccine is clinically effective in immunocompromised children receiving treatment for cancer, with an adjusted estimated vaccine effectiveness of 72% (95% CI −26-94%). During the same influenza seasons, the estimated vaccine effectiveness in healthy children <5 years of age, recruited onto the Western Australian Influenza Vaccine Effectiveness study at PMH [36], was 60% (95% CI−70-91%); and for children <18 years of age presenting to general practitioners in Western Australia, as part of the Western Australian sentinel medical practice surveillance system for influenza [37], was 82% (95% CI 17-96%). The clinical effectiveness of the trivalent inactivated influenza vaccine in immunocompromised children receiving therapy for cancer was therefore comparable to geographically matched children during the same influenza seasons.
The average length of hospital admission and delay in the delivery of scheduled chemotherapy was noted to be greater in unvaccinated compared with vaccinated children with laboratory proven influenza infection. Although this study was not adequately powered for clinical features of influenza infection as an endpoint, these observations concur with outcomes of previous studies [6,7,38]. There were no serious adverse events and a parent reported reactogenicity rate of 3%, which is comparable to the low rates in the literature [24,27,[29][30][31], further supporting the safety of the trivalent inactivated influenza vaccine in immunocompromised children receiving therapy for cancer.
There are several limitations upon which this study is based. Despite recruiting a large number of children in comparison to contemporary published studies, the statistical analyses were limited by patient number and may explain the lack of significance for some of the analyses. The relative rarity of childhood cancer and the low incidence of laboratory proven influenza contribute to this limitation. Future studies should focus on recruiting larger numbers of patients, which may be facilitated through the conduct of large collaborative multicenter trials. To provide additional strength to the data, future studies should also consider prospective comparison of immunogenicity and clinical effectiveness to healthy age-matched controls. It has been shown that children on maintenance therapy for ALL have an inferior but acceptable immune response when compared to healthy controls [27,32], however, these findings have not been extended to other tumor types or other phases of ALL therapy. Finally, there is potential for variable exposure to influenza between the study and control population. This remains one of the generic limitations of testing clinical effectiveness in all influenza studies. It is also possible for test practices to vary between study and control populations, however, we do not expect this to influence the results of our study as children receiving therapy for cancer on our unit all routinely present for an assessment and nasopharyngeal aspirate if they develop influenza-like illness.
In conclusion, our data demonstrate that the trivalent inactivated influenza vaccine is safe, immunogenic, and provides clinical protection in immunosuppressed children receiving therapy for cancer. On the basis of these outcomes, we recommend administration of the inactivated influenza vaccination on an annual basis for children undergoing treatment for cancer. An age-appropriate twodose schedule should be administered to all children <10 years of age, regardless of prior vaccination history. The optimal time for immunization is prior to high intensity therapy, with response more likely in individuals with normal lymphocyte counts and IgG levels. These clinical correlates are reflective of underlying immune function and can be used as a guide to tailor the timing of vaccination on an individual basis, although should not result in vaccination delays. Larger studies are required to further validate the variables predicting seroconversion and to determine the benefit of a two-dose schedule in all immunocompromised children receiving therapy for cancer regardless of age and prior vaccination history.