Chronic obstructive pulmonary disease (COPD) prevalence is increasing over time. It is the fourth leading cause of death in the world and it will be the third by 2021. COPD exacerbations are episodes of temporary symptom worsening (e.g., breathlessness, cough, sputum production) that carry significant consequences for patients [1
] Additionally, exacerbations are associated with an accelerated rate of lung function decline, reduced quality of life, and an increased mortality risk [3
]. COPD exacerbations are heterogeneous and dependent on multiple risk factors [4
]; however, generally, their frequency and severity increase as the disease worsens [5
The fact that COPD exacerbations can be associated with influenza is of great public health interest. The most frequent causes of exacerbations are respiratory infections, most of them viral. In particular, influenza can be a causative virus for exacerbations, of which the more severe ones result in hospital admissions and even death [6
]. Two systematic reviews recently analyzed the effectiveness of seasonal influenza vaccination in patients with COPD, concluding that the evidence supports a positive benefit–risk ratio for seasonal influenza vaccination in these patients, although most of the included trials were more than a decade old [8
]. Considering the role of influenza in contributing to COPD exacerbations, the associated complications, and their related healthcare costs, immunization against influenza is recommended for all patients with COPD by major agencies and guidelines [1
Despite influenza vaccination being recommended, vaccination rates are highly variable among different countries, including Spain [12
], with further room for improvement [18
]. On the one hand, those COPD patients who are likely to exacerbate could have more motivation to accept influenza vaccination [16
]. On the other hand, vaccination rejection among COPD patients has been attributed to concerns about increased exacerbations or adverse reactions caused by the vaccine itself [13
]. There are also many knowledge gaps regarding the impact of COPD severity and co-morbidity on influenza vaccine effectiveness. Therefore, additional studies are required [6
Our primary aim is to determine the prevalence of influenza vaccination in COPD patients in a real-life population cohort in the Balearic Islands (Spain) and the association between history of influenza vaccination and COPD exacerbations.
2. Materials and Methods
Study design: Retrospective cohort study analyzing real-life data from the COPD population of the MAJOrca Real-world Investigation in COPD and Asthma cohort (MAJORICA-cohort). We reported our findings in line with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies using routinely collected health data [19
Data source: The MAJORICA-cohort contains data from all patients (≥18 years) with a primary care diagnosis of asthma and/or COPD in 2012 (N
= 68,578), irrespective of health insurance, with at least two years follow-up available. Most recent follow-up data are from 2015. The cohort data are anonymously collected from the unique primary care system (eSIAP), all hospital systems, as well as the electronic prescription system (RELE) in the Balearic Islands, Spain [20
Study population: Among the overall MAJORICA-cohort, patients who were using respiratory medication (ATC (Anatomical Therapeutic Chemical) code: R03) in 2012, 2013, and 2014 were firstly selected. From the included sample, all patients who had a physician diagnosis of COPD (ICD-9 codes: 491, 492, and/or 496) and were over 40 years of age were identified as patients with actively treated COPD.
Exposure: Patients were classified in two cohorts (exposed versus non-exposed) according to whether or not they had received influenza vaccination in the 2012–2013 campaign (between 18 October and 30 November 2012).
Outcome: The outcome of interest was COPD exacerbation. COPD exacerbation was defined as any episode involving an increase in patient’s baseline COPD symptoms (cough, pleghms, and/or dyspnea), requiring the prescription of an antibiotic and/or systemic corticosteroid (moderate exacerbation), or hospital admission for more than 24 h (severe exacerbation) [23
]. The total frequency of exacerbations (moderate and severe) was quantified in 2012 and 2013. In a sensitivity analysis, exacerbations that occurred during the 2013 influenza epidemic seasonal period in our setting (7 January 2013 to 31 March 2013) [25
] were also identified and quantified as a specific variable. In order to differentiate a new exacerbation from a previous treatment failure, each exacerbation had to be separated for at least 6 weeks from previous exacerbation initial treatment or 4 weeks from hospital discharge.
Co-variates: At baseline, for each patient, sociodemographic and clinical characteristics were collected, including sex, age, smoking habit, COPD obstruction severity (by GOLD grades 1–4), relevant comorbidities, respiratory treatments, and number and severity of COPD exacerbations in the previous year in order to identify the exacerbation “phenotype”. “Exacerbator phenotype” was defined as a patient who suffered from at least two moderate exacerbations or one severe exacerbation in a one-year period according to the main national and international guidelines definition [1
]. “Non-exacerbator phenotype” patients were those having ≤1 moderate exacerbation and no severe exacerbations in a one-year period.
Statistical analyses: For categorical and discrete variables, proportions were estimated using Pearson’s Chi-squared tests for comparisons or Fisher’s exact tests when necessary. Quantitative variables were expressed as mean and standard deviation (SD) using the student’s t-test for comparisons after normal distribution was confirmed using the Shapiro–Wilk test. Dichotomous values (“at least one admission for COPD exacerbation”, “at least one moderate COPD exacerbation”, and “exacerbator phenotype or not” during the following year) were treated as dependent (effect) variables. To assess the crude univariate association between influenza vaccination and the risk of COPD exacerbations, crude odds ratios (OR) were estimated using unconditional logistic regression with 95% confidence intervals (95% CI). Non-vaccinated COPD patients were the reference category. Thus, an OR greater than one indicates that the vaccine is a risk factor for exacerbations. An OR less than one indicates that vaccination is protective, and an OR equal to one indicates a null effect on the risk of exacerbations the following year. To control for confounding bias, we predefined confounders related with the exposure (vaccination) and/or the outcome (exacerbations) in a stepwise regression analysis. The co-variables finally included were age (continuous variable), gender, concomitant asthma diagnosis, COPD severity (GOLD grades 1–4), smoking status (ordinal variable: non-smoker, former smoker, current smoker), number of moderate exacerbations in the previous year, number of severe exacerbations in the previous year, and the following comorbidities: heart failure, atrial fibrillation, cor pulmonale, anxiety, osteoporosis, allergic rhinitis, gastroesophageal reflux disease, and diabetes. Associations were stratified according to COPD severity based on forced expiratory volume in one second (FEV1%) predicted (GOLD grades 1–4) and restricted to patients with spirometry confirmed COPD (FEV1/FVC (Forced Vital Capacity) < 0.7) in a sensitivity analysis. We set the alpha error at 0.05, and all p-values were bilateral. All statistical analyses were performed using IBM SPSS Statistics, version 22.0 (IBM, New York, NY, USA).
Ethics: The study protocol PI17-07 was approved by the Primary Care Research Committee of Mallorca. All data came from real-life clinical files, but they were anonymized for analysis, making it impossible to identify individual patients.
Our results suggest that influenza vaccination in this concrete campaign (2012/2013) did not have a protective effect on the risk of severe and/or moderate exacerbations in the following year, not even if restricting the analysis to the epidemic influenza period only.
The absence of a preventive effect of seasonal influenza vaccination on moderate exacerbation risk is also supported by previous studies. However, in contrast to our results, no previous studies have shown an increased risk for severe exacerbations in very severe COPD patients after vaccination. In the recently published Spanish [16
] or international studies [9
], a protective effect on risk for severe exacerbations (admission for COPD exacerbation) or all-cause admissions was observed in those receiving seasonal vaccination. The higher protective effect being previously shown for the most severe COPD patients (GOLD 4) [17
] is not supported by our results, given we found the opposite effect.
The influenza strains contributing to epidemics vary annually, and they influence the severity and the length of the epidemic seasonal period. Vaccination impact is supposed to be greater in seasons in which circulating strains match those used in vaccination schemes and lower in poorly matched seasons or so-called mismatched seasons. Influenza activity in Spain in the 2012–2013 season was moderate and was associated with a majority circulation of seasonal influenza B virus with a lower contribution of A(H1N1) virus throughout the pandemic wave [25
]. The circulating viruses were consistent with the strains included in the 2012–2013 season vaccine for the northern hemisphere, except for the B viruses of the Victoria lineage not included in the vaccine recommended for the 2012–2013 season. However, the B viruses of the Victoria lineage circulated in minority, thus a lack of efficacy or mismatched vaccine would not explain the lack of preventive results in our study [25
]. Since the 1996–1997 season, the influenza B virus has circulated predominantly in Spain in just two seasons, 2002–2003 and ten years later in 2012–2013. On the other hand, the 2012–2013 season had a late presentation compared to the previous ones, with an epidemic peak in mid-February. Although the epidemic wave was similar to the previous year, the 2012–2013 season was characterized by a prolonged period of intense flu virus circulation, with a percentage of flu-positive samples remaining above 50% for eleven consecutive weeks [25
]. Regarding the immunogenicity of seasonal influenza vaccines in COPD patients, the seroconversion rates ranged from 34.4 to 61.3% for influenza B in published studies [29
], being lower than the seroconversion rates for A/H1N1 (43 to 80.0%) or A/H3N2 (53.1 to 84.1%). Sero-protection rates for influenza B are also the lowest [9
]. To what extent these peculiarities with respect to previous seasons could explain our results may deserve further attention.
The overall influenza vaccination coverage in the 2012–2013 campaign was 59.6%. This sub-optimal coverage is similar to those generally reported in developed countries, including Spain [12
]. In our study, in line with other authors [16
], we found that patients with several co-morbidities had greater vaccination coverage. It should also be mentioned that vaccination coverage among smokers was lower than among non-smokers, as also reported in other studies [16
Vaccination rejection among COPD patients has been attributed to concerns about increased exacerbations or adverse reactions caused by the vaccine itself, which could explain the rejection, even among those with the more severe disease [13
]. However, a greater incidence of exacerbation in the early weeks after vaccination was previously ruled out [35
], and evidence for the safety of the vaccine seems to be conclusive [9
Our study has several limitations.
Only a single influenza season (2012–2013) was studied. Looking at other seasons, especially those where recommended vaccine formulations more closely matched circulating virus strains, is needed to fully understand our observations. Similarly, data on serious influenza-related complications in the general population (vaccinated vs. unvaccinated) need to be taken into consideration.
Our study population was selected under the criterion that they should have used respiratory medication during the years of follow-up. This has advantages because it represents a population with active and relevant COPD. On the other hand, it makes it impossible, for example, to study mortality as a dependent variable in relation to influenza vaccination. A differential survival bias based on vaccination status could be a source of selection bias. If unvaccinated people had higher mortality and exacerbated more, the bias would underestimate the preventive effect of the vaccine. If, on the other hand, vaccinated patients, because they have more comorbidities or are older, die and exacerbate more, the bias would be in favor of the protective effect of the influenza vaccine.
The process for classifying record information into the outcome (i.e., moderate or severe exacerbations) was carried out by two physicians who were blinded to the influenza vaccination status of the study subjects. As the classification of the outcome was blinded to the exposed/non-exposed status of the study subjects, any misclassification of the outcome would have been non-differential towards the null association. It could be also a source of explanation for our results.
External validity is one of the main classical limitations of clinical trials, and this problem can also affect observational studies based on small samples [37
]. A major strength of the present study is the inclusion of the entire population, that is, all patients diagnosed with COPD in the Balearic Islands, Spain, (+/− 1.1 million subjects) receiving primary health care and current treatment. This provides a useful COPD cohort representative for real-life care. It is improbable that bias arose through lack of blindness among patients’ care providers (they treated the patients blinded retrospectively before the development of the study). Another advantage is that we attempted to obtain the independent effect of flu vaccination in our epidemiological and statistical approach by controlling for potential confounding.
On the other hand, as with many registers and database studies, our data were collected from day-to-day clinical practice registers not specifically designed for the present study, and data collection was not under the researchers’ control. Therefore, some unmeasured variables could have confounded the results. Additionally, COPD misdiagnosis could promote a high risk of selection bias. To minimize selection bias, we performed a sensitivity analysis by restricting to only spirometry confirmed COPD cases, and we obtained similar results in this analysis.