Test-Negative Design: The Importance of Laboratory-Confirmed Illness in Estimating the Effectiveness of Influenza Vaccine in Older Adults

(See the major article by Andrew et al, on pages 405–14.) Older adults are among the groups at highest risk for serious complications of influenza and, in many countries, are thus included among the high-priority groups for annual influenza vaccination [1, 2]. Observational studies of influenza outcomes in the United States have shown that approximately 60% of the nearly 300000 excess respiratory and circulatory hospitalizations [3] and 85%–90% or more of the 31000–51000 excess deaths from all causes [4] occur in the population aged ≥65 years. These same observational methods have been used to estimate the effectiveness of influenza vaccine by comparing rates of influenza-like illness (ILI) in influenza-vaccinated older adults as compared to those in unvaccinated older adults. However, these observational designs to estimate the VE have several limitations. Most notably, the criteria that are used to define ILI lack specificity for the diagnosis of influenza, particularly in older adults [5]. Furthermore, observational designs in community-dwelling adults have been shown to be subject to so-called healthy-vaccinee bias: healthier older adults, who are at lower risk of severe outcomes due to influenza, are more likely to be vaccinated than their frailer counterparts, introducing a major confounder that leads to an overestimate of the effectiveness of influenza vaccine [6, 7]. One meta-analysis of available observational studies, using Cochrane methods, found no evidence of a benefit of influenza vaccination in older adults, owing largely to the bias and confounding evident in observational trials that were felt to make conclusions about VE impossible to draw [8]. Although the Cochrane methods for this analysis have been challenged and a reanalysis of the same data by use of a more biologically relevant framework demonstrated the effectiveness of influenza vaccine against several adverse outcomes in older adults [9], questions regarding the benefits of influenza vaccination remain and potentially undermine policies for influenza vaccination of adults aged ≥65 years. More-recent observational studies use a test-negative case-control design, in which influenza vaccination status is compared between cases with laboratory-confirmed influenza and controls presenting with ILI who test negative for influenza virus. This test-negative design has a core assumption that influenza vaccination has no effect on the etiology of viruses other than influenza virus and has been shown to provide valid estimates of the effectiveness of influenza vaccine when attention is paid to controlling for potential confounders [10]. The study reported by van Beek et al in this issue of The Journal of Infectious Diseases used a prospective test-negative case-control design involving community-dwelling older adults who reported acute respiratory illness to their general practitioner during the winter months. By comparing rates of laboratory-confirmed influenza to rates of ILI due to other respiratory viruses, van Beek et al demonstrated that the benefit of influenza vaccination extended only to those with laboratory-confirmed influenza and had no effect on rates of ILI. Influenza vaccination was shown to reduce laboratory-confirmed influenza rates by 73% (95% confidence interval, 26%–90%) and 51% (95% confidence interval, 7%–74%) in the 2011–2012 and 2012–2013 influenza seasons, respectively [11]. In contrast, there was no benefit of influenza vaccination in ILI cases that were caused by a respiratory virus other than influenza virus. This was a prospective study conducted during seasons in which influenza A(H3N2) was the dominant circulating strain. It has been well established that A(H3N2) strains having the greatest impact in older adults, with hospitalization and mortality rates that far exceed those related to influenza B and influenza A(H1N1) strains [3, 4]. This pattern has persisted since the 2009 influenza A(H1N1) pandemic, with the highest hospitalization rates among older adults observed in the years when the predominant circulating strains were the A(H3N2) subtype. In contrast, the effectiveness of influenza vaccine in older adults varies across influenza virus types and subtypes, with protection against A(H1N1) strains being the highest and protection against B strains the lowest, in studies using a test-negative case-control design [12]. Vaccine effectiveness estimates also depend on the severity of the influenza season, with an increased ability to detect a laboratory-confirmed influenza signal among all ILI cases during A(H3N2) epidemic seasons and, thus, higher estimates of vaccine effectiveness especially in observational studies of ILI [9, 13], and offer no benefit in nonepidemic years [12]. Earlier estimates of the effectiveness of influenza vaccine showing significant benefit in the older population [14] were heavily criticized because of the presence of a healthy vaccinee bias. This bias led to overestimates of the effectiveness of influenza vaccine for preventing hospitalization and death in seniors [6, 7]. More-recent, test-negative case-control design studies in community-dwelling elderly individuals, including a large meta-analysis, have demonstrated a benefit in some subsets of this population, including those with cardiovascular or lung diseases and those who are <75 years old [12]. In older adults hospitalized with influenza, a US study with a test-negative case-control design showed a benefit for preventing influenza-related hospitalization in seniors, with no additional confounding related to frailty (assessed retrospectively through chart review), thus addressing the healthy vaccinee bias [15]. In a more recent analysis of the effectiveness of influenza vaccine following the 2009 influenza pandemic, the level of frailty, measured prospectively by the frailty index [16], among older adults hospitalized with ILI was higher among those who were vaccinated, compared with the level among those who were unvaccinated [17]. This finding is the reverse of the healthy vaccinee bias and demonstrates that frailty is a significant confounder in the analysis, leading to underestimates of vaccine efficacy, even in studies of hospitalized older adults that use the test-negative case-control design. van Beek et al acknowledge that their study included a relatively healthy older adult cohort with underrepresentation of the population aged >80 years, may have led to overestimates of vaccine effectiveness against laboratory-confirmed influenza, and did not include estimates of vaccine effectiveness against hospitalization and death. Taken together, these studies highlight the importance of test-negative case-control study designs with prospective measures of frailty in older adults and support including influenza seasons in which A(H3N2) is the predominant circulating strain to assess the effectiveness of influenza vaccine in older adults. Potential conflicts of interest. J. E. M. reports receiving honoraria from GSK, Sanofi, Pfizer, and Merck outside the submitted work. S. A. M. reports receiving research grant funding from GSK, Pfizer, and Sanofi; receiving honoraria from GSK, Merck, and Pfizer; and having served as a consultant to Pfizer. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.