Pregnancy Outcomes in Hidradenitis Suppurativa Patients

Abstract

Hidradenitis suppurativa is an autoinflammatory condition resulting in painful cysts, nodules, and sinus tracts in areas of high skin on skin contact. The microenvironment of affected tissues is high in pro-inflammatory cytokines and T-helper 17 cells. Other auto-inflammatory diseases, like psoriasis, have an enhanced risk of systemic inflammation and an elevated risk of spontaneous abortion. A cohort of pregnant patients from Cerner Health Facts® was identified using a Python adaptation of a validated pregnancy identification and classification algorithm. The HS population was identified among the pregnant population and was shown to be statistically significantly associated with outcome type by Chi square. A multinomial logistic regression also indicated a statistically significant increase in the odds of a pregnant patient having a spontaneous abortion over a live birth when controlling for thyroid disease, polycystic ovarian syndrome, antiphospholipid syndrome, other inflammatory diseases, and advanced maternal age.

Introduction

Hidradenitis suppurativa (HS), is not a disease of the apocrine glands as the name suggests but is more akin to acne¹. This arose due to the appearance that the disease is triggered by the obstruction of hair follicles and the subsequent dilation and rupture of the hair follicle space, or infundibulum, followed by an inflammatory response ¹. Overall, HS affects females at a rate of 3:1 over males ² and impacts 0.4% of the population ³. Those afflicted are characterized by a marked reduction in quality of life due to the extreme pain, localization to more sensitive areas, and associated depression ⁴.

The pathology of this disease is not entirely understood. HS lesional skin has been shown to contain increased pro-inflammatory cytokines such as Interleukin-17 (IL-17) and IL-23 ^5,6 as well as a downregulation of IL-37, an anti-inflammatory cytokine ⁷ implicated in enhancing the activity of regulatory T cells (Treg) ⁸. IL-17A has been shown to promote the proliferation of keratinocytes ⁹, which, in turn, produce IL-6 and, in the presence of IL-1, results in the differentiation of T helper 17 (Th17) cells which are characterized by the production of IL-17, IL-22, and Tumor Necrosis Factor α (TNF-α) ¹⁰. This creates a positive feedback loop and is consistent with the inflammation and pain associated with lesions. How this feedback loop gets started is unknown, however, the involvement of commensal microorganisms in the follicular infundibulum seems to be a likely candidate ¹¹. The inflammatory microenvironment, perhaps in conjunction with uncontrolled microbial growth, eventually leads to rupture of the follicle, exacerbation of inflammation, and the formation of tunnels within the dermis ¹¹.

Pregnancy and Inflammation

Pregnancy is considered a semi-allograft in that it contains genetic components from the same organism and from another individual of the same species. The embryo must invade the uterine lining and connect itself to the maternal blood supply. Thus, the maternal immune system must be appropriately modulated to allow for a successful pregnancy ¹². As with other bodily microenvironments, Th1, Th2, Th17, cytotoxic T lymphocytes, natural killer cells, dendritic cells, Treg cells, B cells, and others not mentioned here and those yet to be fully described all play a role in defending against pathogens and in modifying the behavior of each other ¹³. A careful balance must be maintained.

It has been demonstrated that excess Th17 cells and associated inflammatory cytokines are present in recurrent pregnancy loss ¹⁴. IL-17 is associated with angiogenesis ^15,16 and would be an expected component of early pregnancy but is also associated with embryo resorption in mice during early pregnancy when unregulated ^17,16. It has also been shown that an imbalance between Treg and Th17 cells can result in recurrent pregnancy loss and preterm birth ¹³.

HS and Pregnancy Outcomes

The association of HS with increased IL-17 and decreased IL-37 suggests an imbalance between Th17 and Treg activity. Given this pro-inflammatory state in HS patients, it is possible that this imbalance, if systemic, could affect reproduction. However, pregnancy outcomes are not well established with respect to HS. A single center study at Henry Ford Hospital concluded that, in 202 pregnancies examined, there were no statistically significant differences in the rates of miscarriage, cesarean section, preterm birth, or stillbirth between their HS cohort and the general US population although they did not control confounders ¹⁸. A dual-center study conducted at the University of Pennsylvania and the University of California Los Angeles polled 59 female patients with HS and found a 15% rate of miscarriage ¹⁹, which is consistent with the national average ²⁰. A retrospective analysis using the IBM Explorys platform found 1,600 females with HS over 1,862 pregnancies in which the odds ratios (OR) for miscarriage, and cesarean section were increased compared to controls when adjusting for comorbidities ²¹. A retrospective analysis of the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (HCUP) database found 1,021 pregnant women with HS had a greater likelihood of cesarean section when compared with non-HS pregnancies ²². Another recent retrospective study used IBM MarketScan to identify 998 pregnant women with HS and found a decreased OR for live birth and increased OR for elective termination and unspecified abortion, and cesarean section ²³.

This study seeks to understand the impact of HS on pregnancy outcomes while controlling for confounding variables. The covariates chosen are conditions that tend to be comorbid with HS, like inflammatory bowel disease ^4,24 or polycystic ovarian syndrome (PCOS) ²⁵, or may confound negative pregnancy outcomes due to associations with Th17/Treg imbalance like antiphospholipid syndrome (APS) ²⁶, thyroid disease ^27,28, systemic lupus erythematosus ^29,30, rheumatoid arthritis ^31,32, and psoriasis ³³.

Methods

A pregnancy datamart was constructed from Cerner Health Facts® (HF) using pypreg ³⁴, a Python package developed by the author that adapts a validated algorithm ³⁵ to accept both ICD9 and ICD10 diagnosis and procedure codes. The resulting datamart was cleaned for analysis by including only those patients indicated to be female without also being recorded as male.

Patients receiving a diagnosis of HS or PCOS at any point in their record in HF were considered to be positive for those conditions to account for delays in diagnosis as well as limitations with retrospective data. Thyroid disease, APS, and other inflammatory diseases (ID) were considered longitudinally with respect to each pregnancy. Presence during or prior to a pregnancy was recorded as positive for the condition for that pregnancy. Advanced maternal age was derived from the patient’s age at the time of the outcome and dichotomized as ≤35 and <35. The timespan of a pregnancy was estimated using the admitted date of the outcome encounter and the beginning of the pregnancy start window as determined by the Moll algorithm ³⁵ as not all pregnancies in the datamart were able to have the date of the last menstrual periods determined. The independent variables described above are defined in Table 1.

Table 1.

Description of independent variables used to study pregnancy outcomes.

Univariate statistics were conducted to assess the association of HS with pregnancy outcome type. A multinomial logistic regression was also performed to control for covariates as described in Table 1. The multinomial regression was calculated using the nnet ³⁶ R package. All statistics were performed in R version 4.3.1 ³⁷. P-values less than or equal to 0.05 were considered significant.

Results

Of the seven pregnancy outcome types available in the datamart (live birth, stillbirth, delivery of unknown type, ectopic, trophoblastic, therapeutic abortion, and spontaneous abortion), the “delivery of unknown type” outcome was excluded from analysis. This was due to containing misclassifications and representing a lack of specificity in the source data rather than being a true pregnancy outcome type. This left a total population of 1,572,279 pregnancies with 3,475(0.22%) associated with an HS patient.

A Chi square test of independence indicated significance (X²=73.359, df=5, p<0.001) in the relationship between HS and pregnancy outcome. However, a negligible effect size was also indicated (Cramer’s V=0.007). A residual plot indicated a negative contribution for live birth to the Chi square statistic and positive contributions for each other negative outcome type (Figure 1).

Figure 1.

Residual plot showing the strength of association with HS and pregnancy outcome. Larger and darker circles indicate a stronger contribution to the Chi square statistic, red is a negative association, and blue is a positive association.

A multinomial logistic regression was run using the pregnancy outcome as the dependent variable and the variables listed in Table 1 as independent variables. The model statistics for each pregnancy outcome (ectopic, spontaneous abortion, therapeutic abortion, stillbirth, and trophoblastic) are shown in Tables 2-6 respectively as compared against live birth.

Table 2.

Multinomial logistic regression statistics when comparing Ectopic to Live birth. ORs are recovered by exponentiating the coefficient estimates which are in terms of log odds.

Table 6.

Multinomial logistic regression statistics when comparing Trophoblastic to Live birth. ORs are recovered by exponentiating the coefficient estimates which are in terms of log odds.

Ectopic

When considering ectopic pregnancies, all covariates except APS were statistically significant. The odds of a pregnancy ending as ectopic rather than live birth were increased for patients identified with HS by 73.2% (46.1%105.4%) or PCOS by 87.5% (73.9%-102.1%) and a patient with a concurrent or previous diagnosis of ID by 41.1% (25.4%-58.9%) or advanced maternal age by 40.2% (36.7%-43.8%). Patients with a concurrent or previous diagnosis of thyroid disease had increased odds of live birth over ectopic pregnancy by 6.8% (0.2%-13.9%). The reported p-value for the APS term is the result of rounding, the 95% confidence interval confirms non-significance. (Table 2).

Spontaneous abortion

When considering spontaneous abortion pregnancies, HS, PCOS, ID, and advanced maternal age were all statistically significant. The odds of a pregnancy ending in spontaneous abortion rather than live birth were increased for patients identified with HS by 14% (2.6%-26.7%) or PCOS by 68.2% (61.6%-75.2%) and a patient with a concurrent or previous diagnosis of ID by 24.9% (17.3%-33%) or advanced maternal age by 80.9% (78.7%-83.1%) (Table 3).

Table 3.

Multinomial logistic regression statistics when comparing Spontaneous abortion to Live birth. ORs are recovered by exponentiating the coefficient estimates which are in terms of log odds.

Therapeutic abortion

When considering therapeutic abortion pregnancies, HS, thyroid disease, APS, ID, and advanced maternal age were all statistically significant. The odds of a pregnancy ending with a therapeutic abortion rather than live birth were increased for patients identified with HS by 52.9% (27.6%-83.3%) and a patient with a concurrent or previous diagnosis of ID by 22.7% (8.1%-39.3%) and 47.8% (44.2%-51.5%) for patients of advanced maternal age. However, the odds were reduced for patients with a concurrent or previous diagnosis of thyroid disease by 40% (30.2%-50.5%) and APS by 31.4% (4%-66.2%) (Table 4).

Table 4.

Multinomial logistic regression statistics when comparing Therapeutic abortion to Live birth. ORs are recovered by exponentiating the coefficient estimates which are in terms of log odds.

Stillbirth

When considering stillbirth pregnancies, HS, PCOS, ID, and advanced maternal age were all statistically significant. The odds of a pregnancy ending with a stillbirth rather than live birth were increased for patients identified with HS by 76.2% (30.3%-138.3%) or PCOS by 67.3% (45.3%-92.6%) and a patient with a concurrent or previous diagnosis of ID by 60.8% (32.1%-95.9%) or advanced maternal age by 53.1% (46.4%-60%) (Table 5).

Table 5.

Multinomial logistic regression statistics when comparing Stillbirth to Live birth. ORs are recovered by exponentiating the coefficient estimates which are in terms of log odds.

Trophoblastic

When considering trophoblastic pregnancies, thyroid disease, PCOS, ID, and advanced maternal age were all statistically significant. The odds of a pregnancy ending as trophoblastic rather than a live birth were increased for patients identified with PCOS by 106.4% (79.9%-136.9%) and a patient with a concurrent or previous diagnosis of thyroid disease by 31.7% (18.4%-46.4%), ID by 45.2% (16.2%-81.3%), or advanced maternal age by 79% (70.9%87.6%) (Table 6).

Discussion and Conclusions

In this study, pregnancy outcomes were examined to learn how they were impacted by HS. A Chi square test of independence demonstrated significance, thus the null hypothesis that HS and pregnancy outcome are not associated was rejected. However, the negligible effect size observed indicates that this observation could be spurious due to being overpowered or rare.

To control other covariates and remove the multiple comparisons problem, a multinomial logistic regression was employed. When considering the effect of HS on pregnancy outcomes, while holding all other covariates constant, HS raised the odds ratio of having a spontaneous abortion by 14% (2.6%-26.7%), an ectopic pregnancy by 73.2% (46.1%-105.4%), a therapeutic abortion by 52.9% (27.6%-83.3%), and a stillbirth by 76.2% (30.3%-138.3%) as compared against live birth. With respect to the increase in spontaneous abortion odds, this is consistent with expectations where a pro-inflammatory state could result in a deleterious microenvironment for the developing embryo. Similarly, the increased odds in therapeutic abortion are consistent with expectations as this outcome type is associated with a variety of motivating issues including depressive disorders ^38,39, with which HS is also associated ⁴. However, the precise underlying cause behind these therapeutic abortions is not known. The finding of an increase in the odds ratio for stillbirth and ectopic pregnancies over live births appear to be novel as previous literature has reported no difference in the HS population ^18,23. This finding could be due to a larger sample size and accounting for covariates, although it may be possible that the stillbirths have an association with preterm birth and thus may also be susceptible to imbalances in the Th17 microenvironment ¹³.

This study has several limitations. Retrospective data can contain inaccuracies and may not provide a full picture of a patients’ diagnoses. It is possible that some independent variables may have been incorrectly labeled due to the lack of codes present in HF. While HF is a large secondary database, it only contains information from Cerner EHR facilities that elected to contribute data to HF. Patient care for these conditions could be managed elsewhere and not recorded during the pregnancy. Many spontaneous abortions occur early in pregnancy outside of a medical facility and thus are never recorded. The pregnancy datamart, while consistent with national figures, has not been formally validated against chart review. The model treats each pregnancy as an independent event and could suffer from missing or incorrectly represented terms. HS flare state and severity could not be assessed as notes data is not available in HF nor were a sufficient number of HS patients with C-reactive protein lab results present in the data. Other conditions that could be confounded with HS and pregnancy outcome like BMI and smoking status were not considered. BMI is not always a reliable measure in HF due to missingness, longitudinal inconsistency, incorrect units, or being recorded in imperial units. Smoking status similarly has recording inconsistencies. This study also has several strengths. This study has the largest population of pregnant HS patients investigated to my knowledge, makes use of multinomial logistic regression to attempt to control for covariates without multiple testing, draws from a large secondary database with contributions from across the US, and extensive efforts were made to clean the source database before building the pregnancy datamart.

Many people struggle with their reproductive goals. Having a greater understanding of how a person’s existing diagnoses can impact these goals is important in empowering them to make informed decisions. This study indicates that HS increases the odds of negative pregnancy outcomes over live birth. As a result, it is imperative that HS patients seeking to become pregnant receive appropriate care to align with their goals. Addressing the Th17/Treg imbalance could lead to better outcomes in patients with HS and other conditions associated with a Th17/Treg imbalance. This is generally achieved with systemic biologic treatment, but guidelines for IL-17 antagonists recommend against their use during pregnancy ⁴⁰. Although, there is some evidence reporting no adverse effects of secukinumab exposure during pregnancy and no difference in the rate of spontaneous abortion from the general population in psoriasis patients ^41,42. More study is needed to determine safety and if the observed non-difference in negative outcomes from the general population is an improvement for HS patients and others with a Th17/Treg imbalance condition.