Treatment of seasonal allergic rhinitis

screen. Subsequently, 4,059 references were excluded for not meeting predefined criteria, and 285 were selected for full-text review. These were combined with the 4 articles identified through gray literature and hand search. After removing the references that failed tomeet the inclusion criteria, 59 unique trials were identified of which 13 reference articles were used to address the 3 questions in the current systematic review. The updated Rhinitis Workgroup literature search initially cast a large net for all articles published in regard to rhinitis and treatment with the therapies under consideration. This yielded the following total number of articles: PubMed MEDLINE, 6,536 records; PubMed EMBASE, 140,379; Ovid MEDLINE, 1,316; and Cochrane Trials Registry, 220; for a total of 148,451 articles. After the search terms were combined, the number of possibly relevant references for question 1 was 56, for question 2 was 20, and for question 3 was 40. A summary of the literature search is found in in Appendix A, Tables 4, 5, and 6. The details of the literature search are available in Appendix C. (MEDLINE and Cochrane database printed search with review notes.) Two workgroup members reviewed all abstracts and selected full-text articles. None of the articles met the inclusion criteria that had been established. Although the extended literature search conducted in 2016 by the JTFPP Rhinitis Workgroup did not uncover any new articles that met the inclusion criteria, based on additional selected reviews by workgroup members, including references identified in other recent rhinitis GRADE analyses, the Rhinitis Workgroup selected 3 additional articles,62-64 all pertaining to question 1, for review by themethods group. However, these studies were excluded from the final analysis because required data were incomplete because of data reporting issues (see Appendix A, Table 7 for details). Description of Studies Thirteen studies are reported as single trials.65-77 One metaanalysis reported study findings from 3 trials, one of which was also included as a single trial76 already included in this analysis and therefore was not repeated. Twelve of the studies were randomized, double-blind, placebo-controlled, parallel-group trials,65,67-78 and one study used a double-blind, placebo-controlled, crossover study design.66 The measures used in the studies are found in Appendix B, Table 1. Five studies65,71,72,74,78 disclosed and met the needed sample size to determine significant findings, whereas the remaining studies did not report this value or did not obtain the needed study participants. One study66 was funded by a grant from the Asthma and Allergy Research Group, whereas the remaining studies received funding from pharmaceutical companies or the members of the study teams were or had been a consultant or speaker for a pharmaceutical company or employees of a pharmaceutical company. Efficacy and Safety Outcome Assessment: Forest Plots We chose all variants of nasal with ocular symptom scores, rescue medication score, and Rhinoconjunctivitis Quality of Life Questionnaire (RQLQ) as outcome variables of efficacy. Continuous variables, such as nasal symptom scores, were analyzed in forest plots, and, where possible, the results of several trials were grouped. We chose local and systemic symptoms generally linked to AR medication (eg, somnolence for oral antihistamines and nasal bleeding for INCSs) as outcome variables of safety. Effect Size and Standardized Mean Difference Often when combining data from a large number of studies, which have outcome variables that are not uniform among the trials (eg, some score nasal symptoms scores of 0e12, others of 0e24), the standardized mean different (SMD) is used to determine effect size. The SMD (Hedges g) is the difference between the 2 means divided by the pooled SD, with a correction for small sample bias. In general, when evaluating SMD, Cohen criteria are used to interpret SMD results, in which 0.2 is considered a small effect, 0.5 a moderate, and 0.8 or higher a large effect. The methods group made a decision to combine the data for all studies that used uniformly reported outcomes, such as total nasal symptom score (TNSS). However, for studies for which outcome variables were not uniform, these studies were evaluated separately; thus, SMD was not used. Quality Assessment of the Included Studies: Risk of Bias Using GRADE Analysis An assessment of risk of bias factors (random sequence generation, allocation concealment, blinding adequacy, completeness of data, reporting, and other potential biases) that may contribute to risk of bias was initially conducted independently by 3 reviewers (2 Children’s Mercy, Kansas City, evidence-based practice scholars and J.A.B.) based on the Review Manager software criteria. One nonclinician reviewer (J.A.B.) conducted a draft evaluation on the methodologic quality of the evidence based on the GRADE criteria independently. The workgroup and ultimately the Joint Task Force reviewed these draft assessments, applied their assessments of clinical importance for each patient-important outcome, and determined an overall quality of evidence across outcomes. For studies in which there had been incomplete reporting of information that might affect bias assessment, an attempt was made to contact authors to provide additional information. On the basis of additional information received from authors (Appendix B) and the workgroup and JTFPP’s assessment of the risk of bias using each end point, a final bias assessment was determined by the JTFPP using the modified Delphi process. The level of methodologic quality for the identified literature is summarized after each clinical question. Certainty of the Body of Evidence Using GRADE Analysis79 For GRADE analysis of the certainty of the evidence, 3 areas were evaluated: inconsistency, indirectness, and imprecision. Inconsistency: studies are reviewed in terms of populations, interventions, and outcomes for similarity, or consistency, among the compared studies. Indirectness: analysis occurs around comparisons, populations, and outcomes among intervention studies. Indirectness in comparisons occurs when one drug is compared with placebo and another drug is compared with placebo, but the researchers do not compare the first drug and the second drug in a head-to-head comparison. Indirectness in populations means that the population inwhich the drug was studied doe not reflect the population in which the study drug would be used. Indirectness of outcome refers to a primary or secondary outcome that does not exactly measure the intended outcome (eg, improved quality of life related to rhinitis measured with the generic quality-of-life tool SP27 instead of the specific RQLQ) and thus is not powered for the outcome of choice. Imprecision: when too few study participants were enrolled or too few events occurred in the study, imprecision is detected. The GRADE quality analysis defines the certainty of the evidence. There are 4 levels of evidence: High: The team is very confident that the true effect lies close to the estimate of the effect. Moderate: The team is moderately confident in the effect estimate. The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low: The team confidence in the effect estimate is limited. The true effect may be substantially different from the estimate of the effect. M.S. Dykewicz et al. / Ann Allergy Asthma Immunol xxx (2017) 1e23 6 Very low: The team has very little confidence in the effect estimate. The true effect is likely to be substantially different from the estimate of effect. The GRADE system for evaluating the quality of evidence (http:// gdt.guidelinedevelopment.org/app) defines the elements that guideline writing groups need to consider when evaluating the quality of references that address a specific outcome (ie, change in TNSS). These elements include the risk of bias, described above, as well as the article design (eg, RCT, inconsistency, indirectness, imprecision, and other considerations). Articles are not individually graded for these components but are reviewed overall by the guideline writing group and assigned an overall quality rating. Although some guideline writing groups have tried to develop a point system for grading of individual articles,80 this is not part of the formal GRADE system and was not used in this systematic review. The methods group used by the JTFPP designed a rating of individual references to assist them in their analysis, focusing on the lowest-quality grade assigned to any individual reference as the grade for all of the references used to answer any single question (Appendix B). However, the JTFPP chose to follow the GRADE handbook and reviewed all articles together to determine the overall quality of the articles for each outcome. Each JTFPP member individually determined the quality rating and using the Delphi method, the JTFPP decided the overall quality assessment for each outcome of interest. This difference in approach to the quality assessment is reflected in the discussion within the Clinical Statement Profile for each of the 3 questions. As the final step, the JTFPP rated each outcome across all studies (ie, for a body of evidence) followed by determining an overall quality of evidence across outcomes, again using the Delphi method. The separate quality assessment tables for each of the 3 questions are included within this document. GRADE: From Quality of Evidence (Bias, Certainty) to Recommendations After the quality of evidence is evaluated, the GRADE analysis continues to take into account 3 other factors to finally recommend or suggest in favor or against a certain treatment or action: safety of the intervention, cost, and patient’s preference. As such, the GRADE analysis is not only a system focused on grading the level of evidence but also a much more complete system aimed at formulating recommendations, as its acronym indicates. Throughout the development of this practice parameter, we used the GRADE approach. In formulating the replies to the 3 key questions, we took into account the quality of evidence for treatment efficacy, combining this with patients’ safety, achieving adherence, and cost. Individual subgroups drafted the recommendations and justifications based on the GRADE analysis. Subsequently, all recommendations were reviewed by the workgroup and JTFPP. Both groups were provided the opportunity to comment, propose changes, and approve or disapprove each statement. Consensus was sought and reached for each recommendation’s direction and strength. Actual or potential conflicts of interest were disclosed semiannually, and transparency of discussion was maintained. External peer review was through appointed official reviewers and membership at large of the AAAAI and the ACAAI. All comments were discussed by the JTFPP, and revisions made when the workgroup and JTFPP believed this to be appropriate. Reaching Workgroup Consensus on Statements and Conclusions The workgroup used a modified Delphi process for the determination of the strength of the recommendation and the statement profile for each question. The Delphi method is a structured, interactive, decision-making process used by a panel of experts to arrive at a consensus when there are differing views and perspectives.81-83 For any statement or conclusion for which there was a difference of opinion, a modified Delphi method was used. Workgroup members provided anonymous answers via email to the JTFPP administrative director to the questions being considered. The administrative director provided via teleconference an anonymous summary of the experts’ answers and reasons they provided for their responses. The workgroup members discussed all the answers and then were encouraged to modify their answers on the next round(s) of email voting and teleconferences until a consensus was reached.