Bullous Pemphigoid and Diabetes medications: A disproportionality analysis based on the FDA Adverse Event Reporting System

Background: The world's first Diabetes Medications (Insulin) was marketed in October 1923. Some studies suggested the association of diabetes medications with Bullous Pemphigoid (BP), especially the Dipeptidyl Peptidase 4 (DPP-4) inhibitors. The study aims to detect an association between diabetes medications (focusing on DPP-4 inhibitors) and bullous pemphigoid based on FDA Adverse Event Reporting System (FAERS). Methods: All spontaneous reports of diabetes medications inhibitors-related BP recorded in the FAERS between March 2004 and August 2020 were included in the present study. Disproportionality analysis was performed to find the signal between diabetes medications and BP. The Chi-Squared with Yates' correction (χ2Yates), proportional reporting ratio (PRR) and the lower limit of the 95% confidence interval of the Reporting Odds Ratio (ROR025) were calculated as a measure. A signal was detected when ROR025 > 1, PRR > 2, χ2Yates > 4 and at least 3 cases. Results: There were 3770 reports for BP in FAERS. The strongest signal for diabetes medications-BP association were DDP-4 inhibitors (ROR025: 13.700, PRR: 15.408), followed by Meglitinides (ROR025: 12.708, PRR: 16.777), Non-sulfonylureas (ROR025: 6.434, PRR: 7.016), Alpha-glucosidase inhibitors (ROR025: 6.105, PRR: 10.738), Sulfonylureas (ROR025:2.655, PRR: 3.200). Conclusions: This study detected a strong signal between BP and DDP-4 inhibitors, alpha-glucosidase inhibitors, meglitinides, non-sulfonylureas, and sulfonylureas in FAERS. The signal was significantly higher with alogliptin than with the other DPP-4 inhibitors. The study doesn't suggest the association between the incretin mimetics, insulin, SGLT-2 inhibitors, thiazolidinediones and BP in FAERS.


Introduction
Bullous pemphigoid (BP) is a rare acquired autoimmune skin condition. It usually develops on areas of skin that often flex, such as the lower abdomen, upper thighs, or armpits. The clinical manifestations of BP include tense bullae, urticarial skin lesions and pruritus, oral mucous membrane erosions that may be present in 10-20% of patients [1][2][3]. In some patients, eczema-like erythema may proceed for months or even for many years as a prodromal phase before BP develops [1]. A retrospective monocentric cohort study confirmed that BP was associated with high mortality [4]. BP is most common in older adults, the incidence of BP appears to be equal in men and women and no known ethnic or racial predilection is detected for developing bullous pemphigoid [5]. BP is caused by an autoimmune reaction against bullous pemphigoid antigen 180 (BP180) and/or bullous pemphigoid antigen 230 (BP230), both BP180 and BP230 are a major structural component of hemidesmosomes [6,7]. BP230 localizes intracellularly and associates with the hemidesmosomal plaque, BP180 is a transmembrane glycoprotein with an extracellular domain [6]. Antibodies against both BP180 and BP230, as measured by ELISA, are used for the diagnosis of bullous pemphigoid [3]. But the exact reason for this abnormal immune response is unknown, although it sometimes can be triggered by taking certain medications, trauma, burns, radiotherapy, ultraviolet irradiation, the phenomenon of epitope spreading or genetic factor [8,9]. There are more than 50 medications have been associated with BP development [10].
The world's first Diabetes Medications (Insulin) was marketed in October 1923. A 6 study suggested the association of diabetes mellitus with BP [11]. Meanwhile, DPP-4 inhibitors (also known as "gliptins") and tolbutamide were associated with BP in the literature [10]. So, it was necessary to analyze the association between diabetes medications and BP. For individual diabetes medications, non-sulfonylureas (including the metformin) and DPP-4 inhibitors should be focused on. Metformin was a classic antihyperglycemic drug and the top treatment choice for type 2 diabetes. Metformin always was used in combination with DPP-4 inhibitors. DPP-4 inhibitors are a class of diabetes medications that are used with diet and exercise to control high blood sugar in adults with type 2 diabetes. DPP-4 inhibitors lower blood sugar by helping the body increase the level of the hormone insulin after meals. Insulin helps move sugar from the blood into the tissues, so the body can use the sugar to produce energy and keep blood sugar levels stable. The DPP-4 inhibitors may induce anti-basement membrane zone antibodies or other structurally close antibodies [12], leading to BP. Inhibition of DPP-4 has been shown to enhance the recruitment of eosinophils into the dermis, which may contribute to the blister formation and tissue damage observed in BP [13]. The inhibition of gliptins may cause the activation of eosinophils by a CCL11/eotaxinmediated mechanism. The activation of eosinophils and lymphocyte infiltration substantially contributes to the appearance of blisters and tissue damage in bullous pemphigoid. On the other hand, DPP-4 inhibitors may alter the antigenic properties of the epidermal basement membrane [14]. Even though an increasing number of cases of BP induced by DPP-4 inhibitor was reported in the literature, the exact mechanism underlying this association remains unclear and needs to be elucidated [8].
Previously, some case reports supported the hypothesis that there is a risk of BP in patients exposed to DPP-4 inhibitors [12,[14][15][16][17][18][19][20]. Some retrospective studies suggested that the use of DPP-4 inhibitors is associated with the development of BP in patients with diabetes [21,22]. A meta-analysis suggested that DPP-4 inhibitor exposure is associated with a significantly increased risk for BP [23]. And the warnings and precautions of DPP-4 inhibitors' latest label in the FDA showed that there have been reports of bullous pemphigoid requiring hospitalization. But other types of diabetes medications' labels in the FDA didn't include the warnings about bullous pemphigoid.
Data mining algorithms (DMAs) are currently and routinely used by pharmacovigilance experts for quantitative signal detection [24]. The accuracy of data mining techniques has been already tested retrospectively to determine if already known safety issues would have been detected 'earlier' [25]. Some scholars conducted disproportionality analyses based on DMAs for all spontaneous reports from the French, European, Japanese, WHO and Spanish Pharmacovigilance Database [8,9,[26][27][28].
These studies based on the pharmacovigilance databases all showed a significant association between DPP-4 inhibitors and BP.
FDA Adverse Event Reporting System (FAERS) was the pharmacovigilance database of the United States. We investigated the association between all types of diabetes medications (focused on DPP-4 inhibitors) and BP using the data from FAERS based on DMAs in this study. In addition, the pooled analysis based on DMAs between the DPP-4 inhibitors and BP was made by combining French, American, Japanese, WHO and Spanish Pharmacovigilance Database in the study. 8

Study Design
A retrospective analysis was conducted to comparatively assess BP reports with Diabetes Medications. Acetaminophen was considered as a negative control, whereas furosemide illustrated descriptive positive control [9,10].

Data source
Data in the present study were obtained from the public release of the OpenVigil FDA The data currently used in OpenVigil FDA was obtained from FAERS [29,30].
OpenVigil FDA is a pharmacovigilance tool to extract and analyze FAERS data using the OpenFDA API for accessing the FDA drug-event-database with the additional OpenFDA drug mapping and duplicate detection functionality, OpenFDA aims at providing clean and curated access to the underlying AERS and can count reports stratified to an extraction condition [29], and it overcame some disadvantages of FAERS.
In the study, DPP-4 inhibitors were limited to the approved drugs by the FDA  Table 1) to reduce the confounding bias.
Diabetes medications other than DPP-4 inhibitors analyzed in the study were listed in Supplementary Table 2. The study also analyzed the association between diabetes medications and BP after excluding the cases of combined use of DPP-4 inhibitors to reduce the DDP-4 inhibitors' interference.

Definition of adverse events
Adverse events in the OpenVigil FDA were coded according to the terminology preferred by the Medical Dictionary for Regulatory Activities (MedDRA) Preferred Terms (PTs). For the disproportionality analysis, pemphigoid (PT10034277) were selected for mining according to the MedDRA 22.0.

Data mining algorithms
Data mining algorithms (DMAs) can be classified in the frequentist and Bayesian approach. The frequentist methods are based on the same principles of calculation using the 2  2 table (Supplementary Table 3 For the study, when PRR > 2, χ 2 Yates > 4 (= p < 0.05), the lower limit of the 95% confidence interval of the ROR (ROR025) is greater than one and at least 3 cases as minimal criteria for a signal of disproportionality [31,32].

Case selection
During the study period (between 2004 and 2020), 12254196 adverse drug reaction reports were entered in the OpenVigil FDA. Among these, 89277 adverse drug reaction reports were related to DPP-4 inhibitors, and 3770 adverse drug reaction reports were related to BP. Among these DPP-4 inhibitors' reports, 383 reports were related to BP (alogliptin, n = 70; linagliptin, n = 51; sitagliptin, n = 250; saxagliptin, n = 17), 5 of them involved two or more DPP-4 inhibitors.

Characteristics of the DDP-4 inhibitors and control group
For the gender, the reaction tended to be more common in male ( to be more common in female (58.33%), but furosemide-related cases tended to be more common in male (52.58%). The entire control group tended to be elderly people-at least 11 75 years (45.00% and 55.32% of acetaminophen-and furosemide-related cases, respectively). The age distribution of these cases was similar to the general BP population, but the gender distribution was different from the general BP population [5]. The characteristics of DDP-4 inhibitors and the control group were summarized in Table 1.  For individual DPP-4 inhibitors, the disproportionality order was the same as in the 14 general disproportionality analysis. The results were summarized in Table 2.

BP and DDP-4 inhibitors in the Pooled databases
By combining the results of the study with those previous studies conducted over the FPVD (France), JADER (Japan), FEDRA (Spanish) and VigiBase (WHO) databases [8,9,26,27]. For the DMAs result between DDP-4 inhibitors and BP in the Pooled databases, it showed a signal, with the ROR025, PRR, the number of adverse events and χ 2 Yates of 60.276, 62.711, 1932, and 87122.550, respectively (Table 3).   Table 4.

Discussion
The DPP-4 inhibitors-related BP cases tended to be more common in males (presumably because DPP-4 inhibitors were used more often in males than in females [27]) and elderly people (at least 75 years). The effect of DPP-4 inhibitors on BP did 18 not have a statistical difference in gender in the FAERS. It was different from the result of a hospital-based Swiss-French study and a Finnish nationwide registry study, which found that the effect of DPP-4 inhibitors on BP had a statistical difference in gender [33,34].
These results showed disproportionality for BP and DPP-4 inhibitors in the entire pharmacological databases and the FAERS regardless of whether excluding cases where drugs other than DPP-4 inhibitors were suspected in the BP occurrence, which was consistent with those reported in previous studies conducted in other countries' pharmacovigilance databases [8,9,[26][27][28]. Analysis of each DPP-4 inhibitor separately also showed a significant association. Alogliptin showed higher ROR025 than other DPP-4 inhibitors, followed in decreasing order by linagliptin, sitagliptin and saxagliptin.
It was different from the previous studies [8,9,[26][27][28], presumably because the different regulatory Agencies approved the different DPP-4 inhibitors. For example, the FDA did not approve the vildagliptin, which appeared a higher risk than the others in other countries' pharmacovigilance databases' study [9,[26][27][28]. It was interesting to specify that sitagliptin was the most prescribed DPP-4 inhibitor in the USA [35].
However, disproportionality analyses confirmed a higher risk in alogliptin.  [27]. For the alpha-glucosidase inhibitors, meglitinides, nonsulfonylureas, and sulfonylureas, perhaps because the FAERS database had more reports than the JADER database or the association of diabetes mellitus with BP. In those early reports, the association of diabetes mellitus with BP had been analyzed, and possible underlying mechanisms that increased skin fragility due to elevated glucose levels and the induction of autoantibody production by glycosylation of dermal proteins were suggested [11]. The DMAs results between other individual diabetes medications and BP did not change after excluding case subjects who received DPP-4 inhibitors. It meant that maybe other types of diabetes medications did not interact with DPP-4 inhibitors on BP.
Our study has limitations. The FAERS database was a spontaneous reporting system rather than a mandatory reporting system, the reporters consisted of patients, caregivers, and manufacturers. FDA did not receive reports for every adverse event or medication 20 error that occurs with a product. This introduced an inevitable selection bias, and reporting biases may be differential across different drugs. There was no specific role to check the data in the report, the entry errors couldn't be controlled, such as typographical errors and spelling mistakes.
Moreover, concomitantly administered drugs, age groups and indications possibly introduced confounding bias. To exclude this possible effect, a sensitivity analysis that excluded the cases where drugs other than DPP-4 inhibitors were suspected in the BP occurrence had been made in the study, but BP events that may be caused by unknown drugs' interactions hadn't been excluded. And the patients' other concomitant diseases or drugs or indications were limits in the FAERS report.
Additionally, the FDA did not require that a causal relationship between a product and event be proven, and reports did not always contain enough detail to properly evaluate an event. Mapping names of pharmaceutical products to an active substance is still not sufficiently resolved the issue in pharmacovigilance and epidemiology [37]. So we can use this database to generate hypotheses rather than hypotheses testing, the database can't be used to calculate the incidence of an adverse event or medication error in the United States or establish any causal relationship.
In general, further study, particularly clinical trials, is required with better data sources and research design to ensure whether Diabetes Medications have any synergistic effect on BP.

21
In conclusion, this study suggests a strong signal between bullous pemphigoid and DDP-4 inhibitors in the FAERS and the combining data from French, Japanese, WHO, Spanish and American pharmacovigilance databases. The signal was significantly higher with alogliptin than with the other DPP-4 inhibitors in the FAERS. The effect of DPP-4 inhibitors on BP did not have a statistical difference between gender in the FAERS.
The study also suggests the association between alpha-glucosidase inhibitors, meglitinides, non-sulfonylureas, sulfonylureas and BP in the FAERS. And it doesn't suggest the association between the incretin mimetics, insulin, SGLT-2 inhibitors, thiazolidinediones and BP in the FAERS.