Browse the corpus

Periodic fever, aphthous stomatitis, pharyngitis, and cervical adenitis (PFAPA) syndrome is the most common periodic fever syndrome of childhood (–). Its clinical tetrad defines it, regularly recurring fevers accompanied by aphthous stomatitis, pharyngitis or tonsillitis, and/or cervical adenitis, with complete recovery between episodes and normal growth and development (, ). Onset is typically before five years of age, the reported incidence is approximately 2.3 per 10,000 children under five with a slight male predominance, and the disorder usually resolves spontaneously by early adolescence (, ). Although benign and self-limiting, it imposes a considerable burden on affected children and their families through repeated febrile episodes, missed school, and recurrent, often unnecessary, medical care (, ). PFAPA has no confirmatory laboratory or genetic test; the diagnosis remains clinical and is made by exclusion of infectious, immunodeficient, neoplastic, and monogenic autoinflammatory causes of recurrent fever (, ). Several partly overlapping criteria sets are in use. The original description by Marshall et al. (1987), the widely applied criteria of Thomas et al. (1999), the modified Padeh criteria, and the Eurofever/PRINTO classification, with further criteria proposed for adult-onset disease (, , –). The first PFAPA classification, described by Marshall, is presented in .

Because none of these definitions constitutes a reference standard, their coexistence contributes to diagnostic uncertainty and to the well-recognized problems of delayed and incorrect diagnosis (, ). Mechanistically, PFAPA is regarded as a polygenic autoinflammatory disorder driven by dysregulated innate immunity rather than by autoantibodies or antigen-specific adaptive responses (–). Febrile episodes are characterized by inflammasome activation and overproduction of interleukin-1β by monocytes, with downstream neutrophil recruitment, complement activation, and a T-helper-1/interferon- γ signature. Familial clustering consistent with incomplete-penetrance inheritance is also well documented (–). This innate immune activation provides the rationale for measuring circulating inflammatory mediators during flares and underlies the expectation that these mediators might assist in diagnosis. Corticosteroids reliably abort individual attacks, and tonsillectomy can be curative, but these therapeutic responses do not resolve the diagnostic question (, , ).

The central diagnostic challenge is differentiation. During a flare, PFAPA closely mimics common childhood infections. Most often, PFAPA simulates streptococcal pharyngitis or tonsillitis, and clinically overlaps with other autoinflammatory diseases, including the monogenic periodic fevers (familial Mediterranean fever, mevalonate kinase deficiency, tumor-necrosis-factor-receptor-associated periodic syndrome, and cryopyrin-associated periodic syndromes) and the increasingly recognized syndrome of undifferentiated recurrent fever (SURF), which itself retains an interleukin-1 signature (–). Failure to distinguish PFAPA from infection leads to unnecessary antibiotic courses, whereas failure to distinguish it from monogenic disease can delay disease-specific treatment; both consequences make an objective discriminating marker clinically valuable (, , ).

A range of inflammatory biomarkers rises during PFAPA flares, including the acute-phase reactants C-reactive protein and serum amyloid A, serum calprotectin, and several pro-inflammatory cytokines (–). The recurring difficulty is that these markers are sensitive indicators of inflammation but not disease-specific: comparable elevations occur in bacterial infections and other autoinflammatory syndromes, so an isolated elevated value rarely identifies PFAPA. Reported studies have also differed widely in design, comparator group, sampling time point (flare vs. interval), and assay units, which has hampered direct comparison. Whether any single biomarker, ratio, or composite can reliably distinguish PFAPA from its infectious and autoinflammatory mimics in children, therefore, remains unclear and has not, to our knowledge, been systematically appraised.

We therefore conducted a systematic review of inflammatory biomarkers in children with PFAPA syndrome. Framed by a PECO question: population, children with PFAPA; exposure, circulating inflammatory biomarkers; comparators, infection, other autoinflammatory diseases, or the asymptomatic interval; outcome, discriminating performance, the review aimed to identify which biomarkers have been studied. We would also like to synthesize reported biomarker values at flare and between episodes, and to evaluate how well they separate PFAPA from infection and from other autoinflammatory diseases, thereby clarifying whether any biomarker currently approaches clinical usefulness and where the principal evidence gaps lie. No systematic review published since 2018 has comprehensively synthesized quantitative biomarker data (WBC, neutrophils, lymphocytes, CRP, SAA, serum calprotectin, IL-6, IL-1β, IL-18, TNF-α, IFN- γ ) across the full spectrum of available pediatric PFAPA studies. This gap is clinically relevant: accurate biomarker characterization could accelerate diagnosis, reduce antibiotic overuse, and stratify patients for targeted anti-cytokine therapy.

This systematic review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) reporting guidelines. A completed PRISMA checklist is included in the (), while the PRISMA flow diagram is presented in . The review was not prospectively registered in PROSPERO. We acknowledge this as a limitation. The protocol (eligibility, extraction, and analysis plan) was nonetheless finalized before data extraction and is provided in .

Original clinical studies (prospective or retrospective cohorts, cross-sectional studies, case-control designs, or randomized controlled trials) Pediatric population (age 0–18 years) with confirmed PFAPA syndrome according to any validated diagnostic criteria (Thomas/Marshall 1999, Gattorno/Eurofever/PRINTO 2019, Takeuchi 2019, or equivalent) Reporting quantitative or semi-quantitative data for at least one of the following biomarkers: WBC, neutrophils, lymphocytes, CRP, SAA, serum calprotectin, IL-6, IL-1β, IL-18, TNF-α, or IFN- γ Published in English between 1 January 2018 and 07 June 2026

Review articles, meta-analyses, editorials, commentaries, and letters Case reports and case series with fewer than 5 PFAPA patients Studies with exclusively adult populations (age >18 years) Studies in which PFAPA served only as a reference/comparison group, with no PFAPA-specific biomarker values reported. Conference abstracts without corresponding full-text publication Studies without extractable biomarker data for PFAPA patients

A systematic literature search was conducted in two bibliographic databases, PubMed and Scopus, each interrogated with three complementary query blocks combining the disease term (PFAPA and its expansions) with biomarker, laboratory, and differential-diagnosis terms; the complete search strategies for both databases are provided in the accompanying search-strategy file (). These two databases were selected because together they index the vast majority of pediatric rheumatology and immunology literature in which PFAPA biomarker studies are published. This restriction is acknowledged as a limitation of the search strategy (Section ): studies indexed only in Embase, Web of Science, or the Cochrane Library, or reported in grey literature, may not have been identified. Searches were performed using English-language terms and restricted to articles published between 1 January 2018 and 07 June 2026. No automated database filters were applied during the search process. The final search was performed on 07 June 2026. Two independent reviewers screened titles and abstracts against predefined eligibility criteria; the remaining reports underwent full-text assessment, including relevant [as in Guan et al. ()]. No additional searches of reference lists, citation tracking in Google Scholar or Web of Science, or grey literature sources were conducted.

Three complementary search strings executed in PubMed/MEDLINE (National Library of Medicine, USA) and Scopus on 07 June 2026 yielded 299 records (S1 = 37, S2 = 18, S3 = 23 in PubMed; S1 = 95, S2 = 64, S3 = 62 in Scopus). S1–S3 refer to predefined keyword-based search strategies detailed in . De-duplication was performed using a normalized Digital Object Identifier (lower-cased, prefix-stripped), with a fallback to a normalized title key (lower-cased, punctuation removed, whitespace collapsed) for the small number of records lacking a DOI. Duplicates were removed in two passes. First, within each database across its three queries, then across the two databases, yielding a single transparent count of records removed before screening. The deduplication process, as well as the subsequent screening and analysis, was conducted manually and automatically using the Rayyan platform (a web-based software for systematic review screening) by two independent reviewers (). Disagreements were resolved by consensus.

After removing 129 duplicates, 170 unique records were obtained from database searches. Following title/abstract screening, 137 records were excluded (review articles, editorials, chapters, non-original n = 61; population not PFAPA n = 47; adults-only n = 10; no biomarker outcome n = 10; case reports/small case series n = 9). Thirty-three reports were sought for retrieval, of which 5 were excluded because of the unavailability of full text or non-English regional journals. Twenty-eight full-texts were analyzed for eligibility, and twenty-one reports were excluded (no extractable predefined-panel data n = 8; population not PFAPA-specific n = 6; outcome outside predefined biomarker panel n = 5; duplicate/overlapping cohort or publication n = 2). Seven studies from database searches were included. Two independent reviewers screened and analyzed titles and abstracts against eligibility criteria. Full texts were retrieved and assessed for all records that passed the initial screening. Disagreements were resolved by consensus by a third reviewer. The selection process is documented in the PRISMA 2020 flow diagram ().

Nakano et al. () contributed only three PFAPA patients, but met the review's eligibility criteria: the pre-specified small-sample exclusion applied to case reports and case series, not to controlled designs. Nakano et al. is a retrospective comparative study with a defined comparator (recurrent tonsillitis) and formal between-group testing, so it falls outside that exclusion as written rather than by post-hoc exception. In line with GRADE and Cochrane practice, its small sample was handled through transparent downweighting rather than exclusion. The high risk of bias () and a hypothesis-generating rating contributed to the GRADE downgrading (). No conclusion depends on it: its key signal, elevated interferon- γ in PFAPA vs. infection, is independently reproduced in two larger cohorts [Zhu et al. () and Takeuchi et al. ()]. Studies measuring tissue compartments (e.g., tonsillar) rather than circulating biomarkers were excluded as outside the predefined exposure [e.g., Luu et al. ()]. Non-peer-reviewed preprints and records not indexed in PubMed or Scopus were excluded from the search. They were not eligible [e.g., Gołuchowska et al. (), a Research Square preprint reporting only single-group descriptive laboratory values without a flare-vs.-interval, infection, or autoinflammatory comparator].

A standardized extraction form was applied to each included study. For every study the following were extracted by two reviewers independently: first author and year; the digital object identifier and, where available, the PubMed identifier; country of origin; study design; the number of patients with PFAPA and in the comparator groups; the diagnostic criteria applied; the mean or median patient age; the disease phase assessed (febrile attack, asymptomatic interval, or both), together with the comparator group where one was used (i.e., acute infection, SURF or other); and the numerical values, mean ± standard deviation or median with interquartile range, as originally published, for each of the eleven predefined target biomarkers (WBC, NEUT, LYMPH, CRP, SAA, calprotectin, IL-6, IL-1β, IL-18, TNF-α and IFN- γ ). Where a value was not available from the abstract, the full text was accessed via PMC or the publisher record and, where relevant, the [as for the immune indices of Guan et al.’s ()]. When a specific numerical value remained unavailable despite full-text and supplementary access, the entry was coded as “Not reported” (NR).

The methodological quality of the included studies was appraised using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Analytical Cross-Sectional Studies, applied uniformly to all seven cohorts so that heterogeneous observational designs could be assessed on a common set of domains; for the two studies built on paired within-patient flare-vs.-interval sampling, the equivalent JBI cohort items were additionally considered when interpreting the confounding domains. Two reviewers (P.T.; S.K.) independently assessed each study for potential selection, diagnostic, measurement, and reporting bias, and any disagreements were resolved through discussion and consensus or, where necessary, by a third reviewer (J.S.).

The JBI checklist was applied uniformly across all seven cohorts to appraise heterogeneous observational designs using a common set of domains. Because the JBI cross-sectional tool does not capture the patient-selection, index-test, reference-standard, and flow-and-timing domains specific to diagnostic test-accuracy research, the three studies that additionally reported diagnostic-accuracy estimates (Onur & Onur, Zhu et al., and Takeuchi et al.) were also appraised with the QUADAS-2 tool () across its four domains, each rated for risk of bias and, for the first three, for applicability concerns. QUADAS-2 was not extended to the remaining four studies, which reported only within-group (flare-vs.-interval) or descriptive between-group comparisons without accuracy estimates. On QUADAS-2, all three accuracy studies were judged at high risk of bias in the patient-selection and index-test domains, reflecting two-gate (case–control–type) sampling and in-sample derivation of discriminating thresholds. Their ROC/AUC estimates should be regarded as derivation-stage rather than externally validated performance. The full QUADAS-2 appraisal is provided in . The full item-level JBI appraisal, the overall risk-of-bias judgments, and the JBI traffic-light summary are presented in and ; the QUADAS-2 appraisal of the three diagnostic-accuracy studies, with its traffic-light summary, is presented in .

Because of substantial heterogeneity in comparator groups, assay platforms, units of measurement, and reporting formats, a quantitative meta-analysis was judged inappropriate, and a structured qualitative (narrative) synthesis was undertaken instead. Data were extracted into structured summary tables (study characteristics, hematological parameters, inflammatory markers, and cytokines, and additional acute-phase, cellular, and chemokine markers), with each study treated as an independent unit and no aggregation of values across studies. Continuous data are reported as originally published (mean ± standard deviation, median with interquartile or full range), and missing parameters are marked “Not reported”; where a study's biomarker values resided in [i.e., in Guan et al.’s ()], the was obtained and its data incorporated into the analysis.

We performed a structured GRADE appraisal across the principal questions of this review: the change in each biomarker during the febrile attack vs. the asymptomatic interval (disease-activity monitoring), the discrimination of PFAPA from acute infection, and the discrimination of PFAPA from other autoinflammatory diseases. The certainty of evidence was rated for each of the eleven predefined biomarkers using the standard GRADE domains (risk of bias, inconsistency, indirectness, imprecision, and publication bias), starting at low certainty given the observational nature of the included studies and allowing possible upgrading for large and consistent effects. Certainty was formally rated for the activity and infection-discrimination questions, whereas differentiation of PFAPA from other autoinflammatory diseases was assessed narratively because too few studies addressed it. The summary is presented in . For the discrimination outcomes (Question B), the risk-of-bias domain was additionally informed by the QUADAS-2 appraisal of the three diagnostic-accuracy studies (); the high risk of bias identified there was carried through as a risk-of-bias downgrade for the interferon- γ discrimination outcome.

This systematic review included seven studies examining circulating biomarkers in children with PFAPA syndrome. The study-selection process is summarized in the PRISMA 2020 flow diagram (). In total, 299 records were identified across PubMed ( n = 78) and Scopus ( n = 221), and 170 remained after 129 duplicates were removed. Most title-and-abstract exclusions ( n = 137) concerned non-original publications (reviews, editorials, and book chapters), studies of other autoinflammatory or unrelated conditions, adult-only populations, and papers without an extractable biomarker outcome. Of the 33 reports sought, 5 could not be obtained. Twenty-one of the 28 full texts assessed were excluded, most often because the analytes measured fell outside the predefined panel, the population was not PFAPA-specific, or the predefined biomarkers could not be extracted in a comparative format.

Ultimately, seven studies fulfilled the inclusion criteria. These comprised two prospective studies and five retrospective cohorts, with comparisons spanning the febrile flare vs. the asymptomatic interval, PFAPA vs. acute infection, and PFAPA vs. other auto inflammatory disease (SURF, familial Mediterranean fever) or healthy controls, reflecting both activity-monitoring and differential-diagnostic uses of circulating biomarkers. Data from all included studies were compiled into structured summary tables to facilitate comparison of study characteristics, methodologies, and key findings: study characteristics (), hematological parameters (), acute-phase reactants (), cytokines and interferons (), and additional or derived markers (). The certainty of evidence for each biomarker is summarized in the GRADE evidence profile (), and the per-study risk-of-bias appraisal is presented in and . On QUADAS-2, the three studies reporting diagnostic-accuracy estimates (Onur & Onur, Zhu et al., Takeuchi et al.) were each at high risk of bias in the patient-selection and index-test domains, with the reference-standard and flow-and-timing domains rated high or unclear (); their accuracy estimates are therefore interpreted as derivation-stage rather than externally validated.

Seven original studies published between 2019 and 2026 were included, encompassing 710 pediatric PFAPA patients from 5 countries [China [2 studies], Japan [2 studies], Romania, Slovakia, and Turkey]. The study designs comprised two prospective studies (one clinical and one observational) and five retrospective studies (including cohort, observational, and chart review designs). Diagnostic criteria varied: the Thomas 1999 criteria were used alone or in combination in five studies, the Marshall criteria in three studies, and the Eurofever/PRINTO classification in two studies, with additional guidelines (such as Padeh, Gattorno, or CARRA) supplementing the diagnosis in selected papers. This heterogeneity in diagnostic standards is a limitation affecting between-study comparability. A summary of the included studies is provided in .

Each study in – occupies a single row; within each marker cell, the value is shown as flare value/comparator value (comparator = asymptomatic interval or comparator disease, as indicated in the Contrast column). These are abridged versions; full between-group comparisons for every group reported in each study are provided in ().

Units: White blood cells (WBC), neutrophils (NEUT), and lymphocytes (LYMPH) are reported in the original units; ×10⁹/L and ×10 3 /mm 3 are numerically equivalent, and/mm 3 = ×10⁶/L. Takeuchi et al. () reported neutrophils only as a percentage (segmented + band), not as an absolute count. “Not reported” denotes a parameter not measured or not provided in the source. Values for Guan et al. () are taken from the study's (immune indices by subgroup and disease phase).

CRP is reported in mg/L in Kapustova et al. () and Zhu et al. (), and in mg/dL in Lazea et al. (), Nakano et al. (), Onur & Onur (), and Takeuchi et al. () (1 mg/dL = 10 mg/L); values are presented as published. SAA in mg/L equals μg/mL. IL-6 in pg/mL equals ng/L. IFN- γ was reported in IU/mL by Nakano et al. (), which is not directly convertible to the pg/mL used in Takeuchi et al. () and Zhu et al. (). Calprotectin was measured in serum only, not in stool (, ).

Several markers outside the predefined eleven were measured in individual cohorts and are clinically relevant, particularly for differentiating PFAPA from infection. They are summarized here because their omission would understate the available evidence.

3.3.1 Zhu et al. provided the most rigorous diagnostic-accuracy data in the series, although QUADAS-2 rated the study at high risk of bias in the patient-selection and index-test domains (). The IFN- γ /IL-6 ratio >0.43 differentiated PFAPA from bacterial infection with an AUC of 0.79, and a composite model combining this ratio with IL-10 and platelet count achieved an AUC of 0.95 (90.3% sensitivity, 89.6% specificity). Both thresholds were derived and tested in the same cohort without external validation. These estimates are best regarded as derivation-stage and require prospective confirmation. IL-6 itself was not statistically different between PFAPA and infection (25.1 vs. 27.6 pg/mL, p = 0.722), suggesting that IL-6 has no standalone discriminating value. At the same time, ferritin, fibrinogen, platelets, and ESR were all lower in PFAPA than in bacterial infection ().