The Importance of Sex Stratification in Autoimmune Disease Biomarker Research: A Systematic Review

The immune system is highly dynamic and regulated by many baseline characteristic factors. As such, significant variability may exist among different patient groups suffering from the same autoimmune disease (AD). However, contemporary research practices tend to take the reductionist aggregate approach: they do not segment AD patients before embarking on biomarker discovery. This approach has been productive: many novel AD biomarkers have recently been discovered. Yet, subsequent validation studies of these biomarkers tend to suffer from a lack of specificity, sensitivity, and reproducibility which hamper their translation for clinical use. To enhance reproducibility in validation studies, an optimal discovery-phase study design is paramount: one which takes into account different parameters affecting the immune system biology. In this systematic review, we highlight need for stratification in one such parameter, i.e., sex stratification. We will first explore sex differences in immune system biology and AD prevalence, followed by reported sex-bias in the clinical phenotypes of two ADs—one which more commonly affects females: systemic lupus erythematosus, and one which more commonly affects males: ankylosing spondylitis. The practice of sex stratification in biomarker research may not only advance the discovery of sex-specific AD biomarkers but more importantly, promote reproducibility in subsequent validation studies, thus easing the translation of these novel biomarkers from bench to bedside to improve AD diagnosis. In addition, such practice will also promote deeper understanding for differential AD pathophysiology in males and females, which will be useful for the development of more effective interventions for each sex type.

The immune system is highly dynamic and regulated by many baseline characteristic factors. As such, significant variability may exist among different patient groups suffering from the same autoimmune disease (AD). However, contemporary research practices tend to take the reductionist aggregate approach: they do not segment AD patients before embarking on biomarker discovery. This approach has been productive: many novel AD biomarkers have recently been discovered. Yet, subsequent validation studies of these biomarkers tend to suffer from a lack of specificity, sensitivity, and reproducibility which hamper their translation for clinical use. To enhance reproducibility in validation studies, an optimal discovery-phase study design is paramount: one which takes into account different parameters affecting the immune system biology. In this systematic review, we highlight need for stratification in one such parameter, i.e., sex stratification. We will first explore sex differences in immune system biology and AD prevalence, followed by reported sex-bias in the clinical phenotypes of two ADs-one which more commonly affects females: systemic lupus erythematosus, and one which more commonly affects males: ankylosing spondylitis. The practice of sex stratification in biomarker research may not only advance the discovery of sex-specific AD biomarkers but more importantly, promote reproducibility in subsequent validation studies, thus easing the translation of these novel biomarkers from bench to bedside to improve AD diagnosis. In addition, such practice will also promote deeper understanding for differential AD pathophysiology in males and females, which will be useful for the development of more effective interventions for each sex type.
Keywords: autoimmune diseases, sex differences, gender, sex stratification, biomarkers iNtRODUctiON Autoimmune diseases (ADs) are a complex class of diseases resulting from the immune system failure to differentiate between self and foreign antigens (1). This misrecognition directs the immune system to attack self-antigens, which consequently modifies the biological functions of the affected tissues. Ultimately, tissue damage and dysfunctions ensue and present as clinical symptoms. However, the onset of clinical symptoms is often delayed and occurs following irreversible damage to the affected tissues or organ. There is a global urgency for the discovery of specific and sensitive biomarkers for an early detection of ADs. Additionally, the ideal AD biomarker(s) should also be surrogate for disease severity, progression to disability and response to therapy (2).
The discovery of such biomarkers is not straightforward. Currently, validated biomarkers do not yet fulfill this tall order (we list validated biomarkers from select ADs in Table S1 in Supplementary Material). The immune system biology is dynamic-it varies with genetic background, age, sex, and the environment (3)(4)(5)(6)(7)(8)(9)(10)(11). Thus, patients with varying characteristics may present with different clinical phenotype and biomarkers despite suffering from the same AD. Conventional biomarker research strategy has been reductionist and aggregated: they compare all AD patients and controls of mixed baseline characteristics. This approach, although suboptimal, has been productive, leading to the discovery of many novel AD biomarkers (2). Yet, validation and clinical translation of such novel biomarkers have proven to be challenging, possibly due to: (1) lack of control and patient group stratification and matching; (2) inappropriate biomarker validation strategy; and/or (3) techniques used in clinical trials (12,13). In this review, we highlight the importance of sex stratification in AD biomarker research prior to the discovery-phase, by drawing attention to the fact that: (1) ADs display stronger female bias and (2) present with different disease trajectories in males and females. There are apparent sex differences in AD pathophysiology. These need to be recognized and hopefully over time, incorporated into AD research efforts, clinical diagnosis, and management for a better patient outcome.

metHODS
We assembled a comprehensive list of disease prevalence and associated biomarkers in females and males from different geographical locations. Extensive literature review for the study variables of interest for each of the diseases was mainly carried out using PubMed, although some books and online resources were also consulted. Standard search strategies were used, including medical subject headings such as "disease name" and "biomarker, " or "disease name" and other parameters of interest (e.g., "prevalence, " "in Japan, " "diagnosis, " "autoantibodies, " etc.). Of note, we searched in futility for "sex/gender differences, " "disease name, " and "biomarkers, " perhaps highlighting the paucity of such studies. In total, we reviewed >1,000 abstracts and >400 full papers and included papers which fulfilled the following criteria: 1.

ReSULtS males and Females Have Different Biological Landscapes
A report released by the Institute of Medicine in 2001, "Exploring the Biological Contributions to Human Health: Does Sex Matter?" describes in great detail the different factors that contribute to biological differences in males and females, and how these differences affect health and diseases in these two sexes (11). Beyond the overt differences in reproductive biology, males and females show differences in immune functions, brain organization, pain perception, gene dosing (for genes that escape X-chromosome inactivation) as well as metabolism, lifestyle, and physical performance, all of which may alter pharmacokinetic and pharmaco-dynamic variables in the two sexes (11). Males and females operate from different biological landscapes, be it in healthy, diseased, or recovering states. To explain autoimmune sex dimorphism, it becomes necessary to first describe the cellular and hormonal interactions found in normal immune regulation and thereafter extrapolate these to autoimmune phenomena. In comparison to the innate immunity, the adaptive immune system is known to be significantly affected by sex. Adult females in general, show stronger immune responses than others, and these responses are partially modulated by sex hormones (3,4,6,11). Other contributing parameters include genetic and environmental factors; we refer the readers to a string of excellent reviews for in-depth discussions of these factors (3)(4)(5)(6)(7)(8)(9)(10). We summarize the sex dichotomy in normal immune system biology in Table 1 (3). All components of the innate immunity, such as the toll-like receptor pathway, antigenpresenting cells, dendritic cells, macrophages, granulocytes, and natural killer cells, show stronger activity in females. Despite the lower CD8+ T-cell count, the cytotoxicity of each of these cells is higher in females. As estrogen and progesterone levels wax and wane during the menstrual cycle, the balance between TH1 and TH2 system fluctuates. This balance and its interactions with other systems such as the TH17, Treg, and B-cells dictate the overall immune response. Disruptions to the equilibrium of these different systems lead to different AD pathologies and disease onsets in males and females (6).  Interleukins: IL2, IL4, IL5, IL6,  IL9, IL10, IL13, IL16, IL24, IL32,  oncostatin  TH1 cells: IL20  TH2 cells: IL21, IL25, IL31  TH17 (212). 71% of these common ADs (Figure 1 below, in bold) are more prevalent in females than males (>50% female prevalence), suggestive of a stronger female bias (213). However, these data were an aggregated one from "world, " and "USA" (213). In Table 2, we stratify some of these ADs by geographical location (as a proxy for ethnicity), in order to gain a better insight of each AD's prevalence in different parts of the world. In addition, we include three ADs which have been reported to be more prevalent in males: idiopathic pulmonary fibrosis (IPF), ankylosing spondylitis (AS), and Guillain-Barré syndrome (GBS). Table 2 shows that female predominance prevails in all three systemic ADs. Regardless of the geography, females are more than thrice as likely as males to suffer from systemic ADs. For instance, Spanish women are 18.6 times more likely than men to suffer from Sjögren's syndrome (SS). A similar trend is also observed with endocrine ADs: Grave's disease and Hashimoto's thyroiditis (HT). In India, the female to male ratio for HT is an astonishing 21.7 to 1. Another strong female bias (at least a 2:1 female to male ratio worldwide) is observed with some gastrointestinal and hepatic ADs (primary biliary cholangitis   (17, 18, 27, 174, 176, 177, 185, 187-191, 195, 196) ( 16,181,183,186) Neurological and psychiatric disease (182,185), Seizure (197), peripheral neuropathy (17) Psychosis (177) (17), erythrocyte sedimentation rate, antinuclear antibody, anti-SSA, anti-SSB Mucocutaneous (16,193), vasculitis (181,186), low C3, anti-dsDNA, anti-Sm, anti-rRNP (181,193) Adapted with modifications from Ref. (15)(16)(17)(18).
The results depict studies where significant (P < 0.05) differences were detected. and autoimmune hepatitis), as well as a musculoskeletal AD (rheumatoid arthritis). Although a strong female preponderance was observed with other ADs, these tend to vary with geographical locations. For example, antiphospholipid syndrome shows a female to male ratio of at least 3:1 worldwide, except in Spain and Italy, where it shows lower ratios (2.1:1 and 1.7:1, respectively). Immune thrombocytopenic purpura (ITP) was reported to have a 70% female prevalence in Denmark (213); however, the sex ratio is much lower in our dataset, ranging from 1.1:1 in the USA to 1.7:1 in France. Asian females seem to be more likely to get ITP than other ethnicities, with a female to male ratios ranging from 1.9:1 in India to 4.3:1 in Singapore. Other ADs reported include celiac disease (CoD) and ulcerative colitis (UC), which affected 57 and 65% of females, respectively (213). The female-to-male ratios for CoD and UC vary in our dataset, favoring males or females   In SLE, clinical phenotypes show sex-bias. Alopecia, photosensitivity, malar rash, Raynaud's phenomenon, musculoskeletal complications, oral ulcers as well as psychiatric disorders are female-biased ( Table 3). In addition, females are more likely to suffer from relapses and a concurrent diagnosis for SS. Male-specific clinical phenotypes include discoid lesions, serositis, renal disease, seizure, and peripheral neuropathy, as well as cardiovascular complications. Males also tend to display constitutional symptoms and higher score in Systemic Lupus Erythematosus Disease Activity Index, indicative of a higher disease activity. Indeed, higher mortality has been reported in male patients vs female patients. Serology in SLE also shows some sex-bias: females SLE patients more frequently present with higher erythrocyte sedimentation rate (ESR) and test positive for anti-SSA/Ro while the male SLE patients more frequently test positive for lupus anticoagulant, anti-Sm, anti-dsDNA, and hypocomplementemia.
In AS, clinical phenotypes also show some degree of sex-bias. Male AS patients tend to have disease onset at younger age and present with higher CRP, more back pain, knee involvement, higher scores for BASRI (including BASRI-spinal and BASRIhip), radiographic sacroilitis, higher modified Stoke Ankylosing Spondylitis Spinal Score, lower functional indices (occiput-towall and finger-to-floor distances), but higher SF-36 mental and physical scores. Female AS patients, on the other hand, present more with AS family history, higher Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) score, enthesitis, more numbers of swollen, tender joints and peripheral arthritis but higher ASQoL score. Notably, in a multivariate model, Lee et al. found that for a given level of radiographic damage, female AS patients have more functional limitations than their male counterparts (200).

DiScUSSiON
In this systematic review, we summarize sex differences in immune system biology, AD prevalence, as well as clinical phenotypes of SLE and AS. Data accrued highlight female predominance in common ADs, although there exist geographical differences in some cases. These observed geographical differences are suggestive of potential contributions of genetics and environmental factors toward AD pathology.
In SLE, disease complications and serology seem to show sexbias. Alopecia, for example, is exclusively observed in females just as serositis is exclusively seen in males. It is interesting how the serology in these males and females is reflective of sex-bias in clinical phenotypes. In The Genetic Profile Predicting the Phenotype (PROFILE) multiethnic cohort of 2,322 SLE patients, anti-Sm were significantly associated with antinuclear antibody, anti-double-stranded DNA (dsDNA), and clinical phenotypes, such as serositis, renal involvement, psychosis, vasculitis, Raynaud's phenomenon, hemolytic anemia, leukopenia, lymphopenia, and arterial hypertension (214). Furthermore, double positive serology for anti-Sm and anti-dsDNA has been strongly associated with renal involvement (215)(216)(217)(218) and higher disease activity (219). Most of these clinical phenotypes are male-specific SLE complications (Tables 3 and 4). Similarly, anti-SSA/Ro antibodies have been reported to be strongly associated with low C3 (hypocomplementemia) and clinical phenotypes such depending on geography. In Italy, UK, and India, males are more likely to suffer from CoD than females, whereas the reverse is true in Denmark and Sweden. In Asia, males are more likely to suffer from UC than in the Western world. AS and GBS show greater bias toward males, regardless of the geographical locations. In IPF, however, male predominance was observed all around the world except for the North Americas.

aDs Have Different trajectories in Different Sexes
Autoimmune diseases are quintessentially immune system disorders. As the biology of the immune system varies between the two sexes, one would expect ADs to have different disease trajectories in males and females. Here, we utilize a female-biased AD: systemic lupus erythematosus (SLE), as well as a male-biased AD: ankylosing spondylitis (AS) to exemplify sex-bias in disease progression, complications, and mortality. SLE and AS data are stratified by clinical data, geography, and period of study, since older studies may suffer from delayed diagnoses in male SLE and female AS patients (Tables 3-6).
as photosensitivity, subacute cutaneous lupus erythematosus, cutaneous vasculitis (palpable purpura), hematological disorder (anemia, leukopenia, and thrombocytopenia) (220)(221)(222)(223)(224)(225)(226)(227), as well as Jaccoud's arthropathy (a type of arthritis) (228,229). These phenotypes show female bias in our dataset (Tables 3 and 4). There are some clinical phenotypes such as mucocutaneous and hematological involvements, vasculitis, and association of anti-SSA/Ro with low C3 that differ between these correlation studies and our dataset. This may arise either from ethnic or age differences in the different study groups, or the size of the study groups. The trend for renal involvement persists in all of the studies we have analyzed; however, this clinical phenotype may or may not show a statistical difference for sex-bias owing to the low number of male SLE patients in some studies.
We also observed sex-bias in our dataset for AS clinical phenotypes: female AS patients present with enthesitis and higher BASDAI scores, while male patients present with higher BASRI scores. AS is clinically tested with HLA-B27, ESR, and CRP. While some studies suggested that high CRP is more significantly seen in male patients and high ESR with female patients, many other studies have not come to similar conclusions.
The findings from SLE and AS suggest that disease phenotypes differ between males and females. In some cases, these diseases arguably have higher activity in the sex having lower prevalence. Awareness of sex-bias in disease presentation is crucial for early diagnosis, as well as treatment strategies for ADs in different sexes. More importantly, such awareness may guide the development of improved study design strategies for biomarker discovery.

FUtURe DiRectiON aND cONcLUSiON
Timely diagnosis and treatment can be very effective for AD patients (230,231) and biomarkers have great potential to enable it. Although AD biomarkers discovery is thriving, the same cannot be said of their clinical translation. Many biomarker projects fail at validation/replication stage (13) due to suboptimal sensitivity and specificity, as well as reproducibility in different studies (12). A few potential contributing factors to this observed failure include suboptimal infrastructure, study design, and execution in discovery-phase (12). Suboptimal study design includes small sample numbers, lack of patient history and subject matching (in terms of age, race, and sex) (12). Here we highlight the importance of sex stratification in biomarker discovery studies to promote reproducibility in replication/validation stage. Drawing example from SLE and AS, we note that differential clinical phenotypes exist in male and female patients. Different sexes may require different biomarkers for proper diagnosis of the same disease. From SLE serology we learn that some biomarkers are more frequently detected in one specific sex, and they show strong associations with sex-biased clinical phenotypes. Such specific associations may be missed when data from both sexes are aggregated.
In addition to enhancing sex-specific biomarker discovery and promoting reproducibility, a thorough understanding of sex differences in autoimmune milieu may guide disease prevention, diagnosis, and management. Our findings in Table 2 clearly demonstrate a higher prevalence ADs among females. Breast cancer screening mammography among women at average risk aged 50-74 has been shown to reduce breast cancer mortality by 30-40% (232). These findings suggest potential benefits of AD screenings specifically for women, for early AD detection and reduction of mortality rates through early intervention. Another plausible area of further study is to sex-stratify serological benchmarks for males and females, in light of varying cytokine levels and activity in different sexes as observed in Table 1. We have limited our scope in this review to SLE, AS and sex stratification. Further stratifications for improved patient segmentation and more specific biomarker discovery may include stratifications by age, ethnicities and disease stages.