CONSISE statement on the reporting of Seroepidemiologic Studies for influenza (ROSES‐I statement): an extension of the STROBE statement

Background Population‐based serologic studies are a vital tool for understanding the epidemiology of influenza and other respiratory viruses, including the early assessment of the transmissibility and severity of the 2009 influenza pandemic, and Middle East respiratory syndrome coronavirus. However, interpretation of the results of serologic studies has been hampered by the diversity of approaches and the lack of standardized methods and reporting. Objective The objective of the CONSISE ROSES ‐I statement was to improve the quality and transparency of reporting of influenza seroepidemiologic studies and facilitate the assessment of the validity and generalizability of published results. Methods The ROSES‐I statement was developed as an expert consensus of the CONSISE epidemiology and laboratory working groups. The recommendations are presented in the familiar format of a reporting guideline. Because seroepidemiologic studies are a specific type of observational epidemiology study, the ROSES‐I statement is built upon the STROBE guidelines. As such, the ROSES‐I statement should be seen as an extension of the STROBE guidelines. Results The ROSES ‐I statement presents 42 items that can be used as a checklist of the information that should be included in the results of published seroepidemiologic studies, and which can also serve as a guide to the items that need to be considered during study design and implementation. Conclusions We hope that the ROSES‐I statement will contribute to improving the quality of reporting of seroepidemiologic studies.

of humans, given uncertainties about assay performance and antibody kinetics in exposed and unexposed populations. [9][10][11] In addition, new immunoassays and modifications of well-established assays are increasingly being used for the detection of influenza virus strainspecific antibodies. [12][13][14][15][16][17][18] These issues led to the formation in 2010 of the Consortium for the Standardization of Influenza Seroepidemiology (CONSISE). 19 CONSISE is comprised of international scientists experienced in conducting seroepidemiologic studies of influenza and other emerging respiratory viruses; two working groups on epidemiology and laboratory matters were formed to provide tools to help standardize protocols and laboratory methods used (see https://consise.tghn.org/ about/working-group-projects/). The overarching goal of CONSISE is to improve the quality of data arising from influenza seroepidemiologic studies, harmonize methods used in such studies, and thereby provide better evidence for policy makers that guides rational implementation of intervention and control measures. 19 Guidelines for the reporting of the design, conduct, and results of research have been an effective tool for improving the quality and interpretability of published data. Examples include the Consolidated Standards of Reporting Trials (CONSORT) and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE). 20,21 These guidelines have become, in some instances, widely accepted standards for reporting of research studies, and the expectation that publications should meet these standards has helped to improve the design and conduct of studies. CONSISE has prepared the following statement, Reporting Of Sero-Epidemiologic Studies for Influenza (ROSES-I), which distills the experience of the working groups into a set of recommendations on the optimal reporting of influenza seroepidemiologic studies.

| Objective
The aim of the CONSISE ROSES-I statement was to improve the quality and transparency of reporting of seasonal, avian, and pandemic influenza seroepidemiologic studies in order for the validity and generalizability of the results to be better assessed. This statement also aims to improve the design and conduct of influenza seroepidemiologic studies by proposing reporting standards that investigators should consider when designing studies. CONSISE has developed a number of protocols as guides to the design and implementation of seroepidemiologic studies, and these protocols (available at https://consise. tghn.org/articles/available-consise-influenza-protocols/) are a valuable resource that should be consulted in addition to the ROSES-I statement ( Table 1).
The components of the ROSES-I statement can be used as a checklist to help guide what key information should be included in the results of published seroepidemiologic studies, and can also serve as a guide to the items that need to be considered during study design and implementation. As with other reporting guidelines, this statement is not intended as a required framework that must be followed in content and format. It is also not designed as an instrument for assessing study quality, for which other instruments exist 22,23 .

| Title and abstract
To facilitate the clear identification of studies that quantitatively measure antibodies concentration in members of a defined population in order to make inferences about exposure of that population to emerging respiratory viruses, transmission, and severity, one of the terms "seroepidemiologic," "seroepidemiology," "seroprevalence," or "seroincidence" should be used in the title and/or abstract of the study, and the MeSH term "Seroepidemiologic Studies" should be used as a keyword [ROSES-I 1.1] ( Table 2).

| Introduction
The   The term "seroepidemiologic," "seroepidemiology," "seroprevalence," or "seroincidence" should be applied to the study in the title or abstract, and the medical subject heading "Seroepidemiologic Studies" be used when the report is of a population-based serological survey Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen, and why ROSES-I 11.1: Describe the serological assay's limit of detection and how this limit is defined or calculated. Describe how samples with a result below or on the borderline of the limit were handled in the analysis ROSES-I 11.2: Describe and justify the titer or other result used to define "seropositivity," or the antibody titer change or change in other assay result used to define "seroconversion." Avoid the term "seroconversion" unless referring to change from undetectable to detectable antibody level. Otherwise report the fold-rise in titer. Avoid the term "infection" but report "seroprevalence at a titer of …."

Study design and setting
A number of different seroepidemiologic study designs can be used to estimate various measures of virus infection risk, and different designs have different strengths and weaknesses depending on the objectives of the investigation ( Table 1). The methods section should begin by describing the study design, the study population, sampling procedures (e.g., random or convenience), the source of serum or plasma that was analyzed (e.g., frozen stored vs recently collected and tested), the rationale for choosing the study design, and the gen-

Participants
For studies where close contacts of confirmed influenza cases (e.g., household contacts of confirmed cases;

Study subjects and sample size
In order to permit the generalizability of the study results, it is preferable that the study population be as close as possible to the general population under study. However, this is not always feasible, and for example, many serologic studies detecting antibodies to A(H1N1) pdm09 virus infection were conducted using residual sera from blood donors 4,28 or hospitalized patients, 29 who may not be representative of the broader populations in those locations. The potential to introduce selection bias into the study should be addressed in the discussion. Any method used to infer cumulative incidence of infection among the population based on results from the study sample, for example, weighting or standardization, should be reported in sufficient detail to permit reproducibility (ROSES-I 16.2).
In addition, the confidence in the results and conclusions of any seroepidemiologic study depends, among other things, on whether the planned study sample size was sufficient to provide estimates of prevalence or incidence of infection with sufficient precision and certainty. 30 To assess whether the planned sample size was adequate, the How these titers are analyzed and interpreted can affect the results, especially if they constitute a large proportion of the results. A common convention, which is acceptable, is to consider a result below the limit of detection as a serial step below that limit; that is, if the starting antibody dilution is 10, then a value of <10 can be reported as a five for the purposes of data analysis rather than a zero or a "not

Statistical methods and presentation of results
In cross-sectional studies, seroprevalence can be estimated by the proportion of specimens with antibody titers at or above a specific threshold, with 95% confidence intervals typically obtained using the binomial formula or the normal approximation to the binomial. If a number of additional assumptions are met, including that seroprevalence before an epidemic is very low, and almost all infected individuals have rises in convalescent antibody titers above the chosen threshold, the post-epidemic seroprevalence can provide an approximate estimate of the cumulative incidence of infection. 34 Note that seroprevalence is a proportion and not a rate.
In studies with paired sera, the cumulative incidence of virus infection can be estimated by the proportion of persons with a rise in antibody titer, traditionally a fourfold or greater rise. 31 In most studies, 95% confidence intervals are typically estimated using the binomial formula or the normal approximation to the binomial, implicitly assuming that each person can experience no more than one virus infection during the period considered. It is noteworthy to point out also that cumulative incidence of virus infection is sometimes referred to as an "attack rate," although a proportion of infections may be asymptomatic (and therefore not "attacks"), and the quantity measured is a proportion and not a rate. The term "cumulative incidence of infection" should therefore be preferred to "attack rate" in the context of serological studies.
In either case, the methods used to account for the probability of seropositivity or seroconversion if infected, and any method used to account for decay in antibody titer over time, should be reported (ROSES-I 12.2). To increase transparency of cumulative incidence of infection estimates, it is often helpful to report unadjusted estimates of the distribution of antibody titers by age group (ROSES-I 16.1).
In some studies, particularly those with more complex designs in terms of timing of serologic measurements, improved estimates of the seroprevalence at a certain point in time, or the cumulative incidence of infection over a specified time period, may be obtained by fitting observed data to a mechanistic model of transmission dynamics. 4,35 This can account for non-independence in the data (ROSES-I 12.1).

Sample type and handling
Although serum samples are more commonly used for serologic studies, convenience sampling may only enable access to plasma. The use of anticoagulants to separate plasma has been shown to reduce the antibody titer to some influenza viruses. 36  This antigen should be antigenically equivalent to the specific virus strain to which the study population was exposed. To enable the comparison between laboratories and also aid in the development of Specifically, as the WHO recommendation for A(H5N1) viruses recommends the use of confirmatory serologic assays upon the detection of single serum positive by MN assay, any confirmatory assays used and the criteria for positivity also need to be described in the same details as above (ROSES-I 12a.13). 38,39 Inclusion of available international standards 40,41 facilitates the comparability of serological data.
Inclusion of the actual titers obtained from the international standards and indication whether the data are reported as raw values or international standard-adjusted values should therefore be described (ROSES-I 12a.14).

| CONCLUSIONS
The direct comparability of influenza seroepidemiologic studies is currently limited by a lack of standardization across such studies. 8 The