Gliptin Accountability in Mucous Membrane Pemphigoid Induction in 24 Out of 313 Patients

Mucous membrane pemphigoids (MMPs) and bullous pemphigoid (BP) are autoimmune bullous diseases that share physiopathological features: both can result from autoantibodies directed against BP180 or BP230 antigens. An association has been reported between BP and intake of gliptins, which are dipeptidyl peptidase-IV inhibitors used to treat type 2 diabetes mellitus. Clinical and immunological differences have been reported between gliptin-induced BPs and classical BPs: mucosal involvement, non-inflammatory lesions, and target BP180 epitopes other than the NC16A domain. Those findings accorded gliptins extrinsic accountability in triggering MMP onset. Therefore, we examined gliptin intrinsic accountability in a cohort of 313 MMP patients. To do so, we (1) identified MMP patients with gliptin-treated (challenge) diabetes; (2) selected those whose interval between starting gliptin and MMP onset was suggestive or compatible with gliptin-induced MMP; (3) compared the follow-ups of patients who did not stop (no dechallenge), stopped (dechallenge) or repeated gliptin intake (rechallenge); (4) compared the clinical and immunological characteristics of suggestive-or-compatible-challenge patients to 121 never-gliptin-treated MMP patients serving as controls; and (5) individually scored gliptin accountability as the trigger of each patient’s MMP using the World Health Organization-Uppsala Monitoring Center, Naranjo- and Begaud-scoring systems. 17 out of 24 gliptin-treated diabetic MMP patients had suggestive (≤12 weeks) or compatible challenges. Complete remission at 1 year of follow-up was more frequent in the 11 dechallenged patients. One rechallenged patient’s MMP relapsed. These 17 gliptin-treated diabetic MMP patients differed significantly from the MMP controls by more cutaneous, less buccal, and less severe involvements and no direct immunofluorescence IgA labeling of the basement membrane zone. Multiple autoantibody-target antigens/epitopes (BP180–NC16A, BP180 mid- and C-terminal parts, integrin α6β4) could be detected, but not laminin 332. Last, among the 24 gliptin-treated diabetic MMP patients, five had high (I4–I3), 12 had low (I2-I1) and 7 had I0 Begaud intrinsic accountability scores. These results strongly suggest that gliptins are probably responsible for some MMPs. Consequently, gliptins should immediately be discontinued for patients with a positive accountability score. Moreover, pharmacovigilance centers should be notified of these events.

Mucous membrane pemphigoids (MMPs) and bullous pemphigoid (BP) are autoimmune bullous diseases that share physiopathological features: both can result from autoantibodies directed against BP180 or BP230 antigens. An association has been reported between BP and intake of gliptins, which are dipeptidyl peptidase-IV inhibitors used to treat type 2 diabetes mellitus. Clinical and immunological differences have been reported between gliptin-induced BPs and classical BPs: mucosal involvement, noninflammatory lesions, and target BP180 epitopes other than the NC16A domain. Those findings accorded gliptins extrinsic accountability in triggering MMP onset. Therefore, we examined gliptin intrinsic accountability in a cohort of 313 MMP patients. To do so, we (1) identified MMP patients with gliptin-treated (challenge) diabetes; (2) selected those whose interval between starting gliptin and MMP onset was suggestive or compatible with gliptin-induced MMP; (3) compared the follow-ups of patients who did not stop (no dechallenge), stopped (dechallenge) or repeated gliptin intake (rechallenge); (4) compared the clinical and immunological characteristics of suggestive-or-compatible-challenge patients to 121 never-gliptin-treated MMP patients serving as controls; and (5) individually scored gliptin accountability as the trigger of each patient's MMP using the World Health Organization-Uppsala Monitoring Center, Naranjo-and Begaudscoring systems. 17 out of 24 gliptin-treated diabetic MMP patients had suggestive (≤12 weeks) or compatible challenges. Complete remission at 1 year of follow-up was more frequent in the 11 dechallenged patients. One rechallenged patient's MMP relapsed. These 17 gliptin-treated diabetic MMP patients differed significantly from the MMP controls by more cutaneous, less buccal, and less severe involvements and no direct immunofluorescence IgA labeling of the basement membrane zone. Multiple autoantibody-target antigens/epitopes (BP180-NC16A, BP180 mid-and C-terminal parts, integrin α6β4) could be detected, but not laminin 332. Last, among the 24 gliptin-treated diabetic MMP patients, five had high (I4-I3), 12 had low (I2-I1) and 7 had I0 Begaud inTrODUcTiOn Mucous membrane pemphigoids (MMPs) are rare diseases with very low annual incidences worldwide, ranging from 0.07 million inhabitants in Kuwait to 2 million inhabitants in Germany, and intermediate, with 1.25 million inhabitants, in France (1). These diseases are defined clinically. They cover a heterogeneous group of subepithelial autoimmune blistering diseases that predomi nantly affect the mucous membranes (2). They include the clas sical MMP, formerly called cicatricial pemphigoid, laminin 332 MMP, α6β4 integrin MMP, mucous membrane dominant epid ermolysis bullosa acquisita (MMEBA), and mucous membrane dominant linear IgA disease [MMlinear IgA bullous dermatosis (LABD)]. Abnormal scarring is the hallmark of MMPs: lesions heal via a fibrosing process leading to cicatricial lesions that can cause severe impairment of the eyes or can be lifethreatening in larynx or esophagus.
An association between BP and the intake of several drugs (spironolactone, amiodarone, sulfasalazine, allopurinol, furose mide, etc.) has been reported, since 1970 (6)(7)(8), and most recently with gliptins, which are dipeptidyl peptidaseIV (DPPIV) inhi bitors used to treat type 2 diabetes mellitus. Three gliptins are currently available in France: sitagliptin and vildagliptin, since 2007, and saxagliptin, since 2009. The first BP cases associated with gliptin intake were described in 2011. Since then, 42 cases of gliptinassociated BP have been published as case reports or in short series (9-23), 37 in two casecontrol studies (20,24), and 208 identified in pharmacovigilance databases (16,25). A study comparing 3,397 BP patients to 12,941 basocellular carcinoma controls from the Finnish nationwide registry and showing that vidagliptine increases the risk of BP has also been partially published very recently (26). Several authors have highlighted different clinical and immunological phenotypes of these gliptin associated BPs: mucosal involvement (15), noninflammatory lesions (18,23), and target BP180 epitopes outside the NC16A domain (18,23).
Because the role of gliptins in MMP had never been investi gated, we examined gliptin accountability in MMP induction in 24 gliptintreated diabetic MMP patients in our center cohort of 313 MMP patients. Our primary objective was to identify patients with a first gliptinintaketoMMPonset interval "suggestive" or "compatible" with MMP induction. Then we analyzed clinical and immunological findings and outcomes of these selected patients to evaluate other accountability criteria of gliptin MMP induction and, finally, indicate prognosis.

MaTerials anD MeThODs referral center Database
This singlemultisitecenter retrospective study (January 2007-June 2016), approved by our local Institutional Review Board (IRB 00003835 no. 2013/39NI), was conducted using the database of our Referral Center for autoimmune bullous diseases. The fol lowing information was systematically recorded in each patient's standardized medical chart. During their first consultation at our Center, all patients were asked about their medical history and treatments, evaluated by a multidisciplinary team that noted all cutaneous and mucous membrane lesions, clearly distinguish ing MMP reversible "active" mucous membrane lesions from irreversible "cicatricial" mucous membrane lesions (27)(28)(29). Skin and/or mucous membrane biopsy findings and immunosero logical results at diagnosis yielding a definite diagnosis were also recorded: direct immunofluorescence (DIF) immunedeposit pattern at the dermal-epidermal junction (linear) and Ig class(es) (IgA, IgG, IgM), ±C3 deposits; ultrastructural immunedeposit location by direct immunoelectron microscopy (IEM; when done, according to availability in each department); standard indirect immunofluorescence (IIF) on rat or monkey esophagus and 1 M NaCltreated human or commercially available (Euroimmun, Lübeck, Germany) monkey saltsplit skin (SSS), using polyva lent antiIgG, IgA, IgM as secondary antibodies; commercially available BP180-NC16A and BP230 enzymelinked immuno sorbent assays (ELISAs) using antiIgG secondary antibodies (MBL, Nagoya, Japan and/or Euroimmun, Lübeck, Germany); and immunoblot on amniotic membrane extracts (when done, according to availability in each department). In patients with a subepithelial autoimmune blistering disease [i.e., linear immunoglobulin (Ig) deposits along the basement membrane zone (BMZ) in DIF], diagnoses of MMP and BP were retained Keywords: autoimmune bullous diseases, mucous membrane pemphigoid, fibrosis, dipeptidyl peptidase intrinsic accountability scores. These results strongly suggest that gliptins are probably responsible for some MMPs. Consequently, gliptins should immediately be discontinued for patients with a positive accountability score. Moreover, pharmacovigilance centers should be notified of these events. on clinical criteria: the first when lesions predominantly affected the mucous membranes (2) and the second in patients who fulfilled criteria of Vaillant (3). Diagnoses of LABD and EBA were retained on immunological criteria: the first on class IgA of autoantibodies and the second when autoantibodies targeted type VII collagen (by ELISA and/or immunoblot) and/or are located in anchoringfibril zone (AFz) (by IEM). Diagnosis of MMLABD or MMEBA was retained in patients with predomi nant mucous membrane lesions and respectively a LABD or an EBA. Other MMP subgroups have been diagnosed according to the target antigen of autoantibodies (by IEM, IIF on SSS, ELISA, and/or immunoblot). In particular, the immunoblot on amniotic extract allowed the detection of antibodies to laminin 332 and α6/β4 integrin (30). Last, MMP was classified as severe or not according to Chan criteria (2,(27)(28)(29).

gliptin-Treated MMP Patients
First, we identified all consecutive patients with a definite MMP diagnosis, then selected those with diabetes mellitus, and, finally, included MMP patients prescribed gliptins to treat their diabetes.

chronology of gliptin intake and MMP Onset
Using Begaud's updated nomenclature (terms in bold type; 31) to impute a potential gliptin role in triggering MMP, dates of gliptin introduction (challenge), discontinuation (dechallenge), reintroduction (if any) (rechallenge), and the first MMP symp toms were extracted from the Center's database, collected from patient's chart and/or by contacting the patient's general practi tioner. Diabetics who started taking a gliptin before MMP onset and had a first gliptinintaketoMMPonset interval suggestive or compatible with MMP induction formed the suggestiveor-compatible challenge group (Figure 1). Those who discon tinued gliptin during the first year of MMP followup formed the dechallenge group and those who did not comprised the no-dechallenge group.

clinical and immunological Findings and MMP Follow-Up
Clinical and immunological characteristics at MMP diagnosis were extracted from the Center's database. Observation end points were collected retrospectively from patients' charts: time to first MMP complete remission (CR), CR rate at 1 year of MMP follow up, and relapse rates at 1 year and the end of MMP followup (1-3 years), according to Murrell et al. 's criteria (31).

reference series of MMP Patients Without gliptin intake
One hundred and twentyone consecutive MMP patients who had never taken a gliptin and had at least 1 year of followup served as controls. Patients with MMLABD were excluded, as were those with MMEBA. Clinical and immunological findings at diagnosis and CR and relapse rates at 1 year of followup were extracted from the Center's database and/or collected from each patient's chart.

statistical analyses
Quantitative variables, reported as mean ± SD or medians (range), were compared with Mann-Whitney Utests. Qualitative parameters, expressed as numbers (%) were compared with chi 2 or Fisher's exact tests, as appropriate. Twotailed pvalues less than 0.05 were considered statistically significant. Statistical analyses were computed with R version 3.4.3 (R Foundation for statistical Computing, Vienna, Austria).
Direct immunofluorescence microscopy of tissue samples from 14 (82%) of the 17 of suggestiveorcompatiblechallenge group patients had linear IgG deposits and 15 (88%) linear C3 deposits along the BMZ; none had IgA deposits ( Table 2). IEM (Figure 4) of nine patients' biopsies were positive for eight (89%): immune deposits were located on the lamina densa with/without the lamina lucida (LL) in seven (78%) of them and exclusively on the upper LL in the last one (11%); none had deposits under the lamina densa in the AFz. Autoantibodies directed against BMZ antigens were detected in 8 (47%) of the 17 patients whose sera were tested by IIF on rat or monkey esophagus. IIF on SSS was positive for 4 out of 15 patients and autoantibodies always labeled the cleavage roof; none labeled the cleavage floor. ELISA BP180 and BP230 were positive for 7 (47%) and 3 (20%) of the 15 patients tested, respectively. Immunoblots on amniotic mem brane extract ( Figure 5) of sera from seven of the nine ELISA BP180-NC16Anegative patients were negative for three patients and positive for four, detecting: a 200kDa band consistent with the β4 chain of α6β4 integrin (patient 8), whose immune deposits were located on hemidesmosomes by direct IEM; a 180kDa band (patient 3); a 120kDa band (patient 11); and 180 and 120kDa bands (patient 7).

Follow-Up of the suggestive-orcompatible-challenge group
Overall, median followup was 40 (range 0-164) weeks and median time to CR 8 (range 0-36) weeks. After the first year of followup, CR and relapse rates were 82 and 31%, respectively ( Table 3). No patient died.
Last, for the dechallenge group, MMP evolution was suggestive of gliptin imputability for six (1, 2, 5, 8, 9, and 11) of the seven patients who obtained CR, non-suggestive for the patient who did not, inconclusive for patient 4 in CR at 1 year of followup but who relapsed, and for patients 3, 7, and 10, with <1 year of followup, and patient 3 not seen at 1 year. For the no-dechallenge group, at 1 year of followup, MMP evolution was non-suggestive for patients 15 and 16 in CR, suggestive for patient 12 who had not achieved CR and inconclusive for remaining patients 13, 14, because of too short followup, and 17, who was not seen at 1 year.

control MMP Patients Without gliptin Use
Our controls were 121 patients with MMP (excluding MM EBA and MMLABD) seen consecutively in our Center, who had never taken a gliptin and were followed for at least 1 year ( Table 4). At MMP diagnosis (baseline), their ages ranged from 38 to 96 years, weights from 44 to 114 kg, and body mass indexes from 18 to 40. Half of them had only mucous membrane involvement and the other half had cutaneous and mucous membrane involve ments, and 78 had severe MMP. Controls had a median of 2 (1-5) involved sites: 50% skin, 89% mouth, 30% larynx, 31% genitals/ anus, 25% conjunctiva, and 31% esophagus. Immunologically, by DIF microscopy, 74% controls had linear IgG deposits along the BMZ, 71% had linear C3 deposits, and 26% had linear IgA deposits which were neither isolated nor predominate over IgG. Immune deposits were located on the lamina densa with/ without the LL in 60% of them and on the upper LL in 13%. IIF microscopy of their biopsies identified circulating autoan tibodies labeling the cleavage roof of SSS in 21%, the cleavage floor 2%, both sides in 3%, and no labeling of 74%. Control MMP patients whose sera labeled the cleavage floor in IIF on SSS had deposits on the lamina densa, but not the AFz by IEM, thereby excluding EBA and consistent with a laminin 332 MMP diagnosis. ELISA detected circulating antiBP180 autoantibodies in 43 (51%) and antiBP230 in 10 (13%) control sera. At 1 year of followup, 56% of the controls were in CR, 23% suffered relapses/ flares, and 2% had died.
Comparing the suggestive-or-compatible-challenge group's characteristics to those of the MMP controls, most were compa rable, especially the relapse and CR rates at 1 year of followup.  The suggestive-or-compatible-challenge patients were older than controls but not significantly so. However, suggestive-orcompatible-challenge group differed significantly from controls by weighing more, having higher body mass indexes, with more frequent cutaneous involvement, less frequent buccal, and severe involvements, and no IgA deposits.

accountability scoring
Using Begaud's scoring system, extrinsic accountability was rated B2 (sparse and/or unreliable publications) for all our gliptintreated MMP patients, in analogy with reported gliptin induced BP, a similar autoimmune bullous disease ( Table 5). For the suggestive-or-compatible-challenge group, the chronological accountability criterion was scored C3 (likely): for patients 1 and 2 because of their suggestive times to MMP onset and suggestive outcomes after gliptin dechallenge, and patient 8 because of the compatible time to MMP onset, positive rechallenge, and suggestive outcome; C2 (plausible): for patients 3 and 4 with suggestive times to MMP onset and inconclusive outcomes, and patients 5,9,11,12, and 17 with compatible times to MMP onset and suggestive outcomes; and C1 (doubtful) for the seven others because of compatible times to MMP onset but inconclusive or non-suggestive outcomes. The symptomatological accountability criterion was scored: S2 (evocative) for five patients whose clinical (cutaneous lesions, no buccal disease, no severe involvement) and immunological features (no IgA deposits), which differed significantly from controls; and S1 (not evocative) for the 12 other patients. No specific laboratory test can prove the link between gliptin intake and MMP.
Finally, the intrinsic accountability (combining C and S scores) was rated I4 for three patients, I3 for two patients, I2 for eight, I1 for four, and I0 for the seven patients with chronolo gically incompatible or undetermined challenge.
According to WHO-UMC accountability criteria, gliptin was probably responsible for triggering MMP for all patients of the suggestive-or-compatible-challenge group, a reason able time relationship between drug intake and first MMP manifestations; MMP regressed after gliptin withdrawal and relapsed after readministration; and because MMP could have been spontaneous.

DiscUssiOn
Our novel study on gliptin accountability in MMP induction was undertaken because of their extrinsic accountability, based on the following reports: MMP and BP have clinical and immu nological similarities (4), a demonstrated significant association between gliptin intake and BP onset in diabetic patients (16,24,25) and some gliptinassociated BPs have atypical clinical and immu nological phenotypes (15,18,23,25).
Mucous membrane pemphigoid and BP are subepithelial AIBDs, characterized by linear immune deposits along the BMZ, but have different clinical features. MMP is clinically defined by the predominance of mucous membrane lesions over skin lesions (2) and healing of its lesions leads to characteristic cicatricial scarring. BP, on the other hand, is typified by the absence of mucous membrane lesions, absence of predominant headand neck involvement, and absence of scars, and older age at onset (>70 years) (3).  Mucous membrane pemphigoid and BP share two autoanti bodytarget antigens, BP230 and BP180, but the dominant BP180 epitopes differ (4). The majority of MMP patients' sera react with the Cterminal domain of BP180, located in the lamina densa, combined or not with reactivity against the NC16A epitope, which is the membraneproximal noncollagenous region of the BP180 ectodomain in upper LL (5,38,39). Conversely, 80-90% of BP patients have IgG autoantibodies directed against the NC16A domain (40)(41)(42). Moreover, many authors reported that variable percentages (10-50%, depending on the study) of BP autoanti bodies targeted BP180 regions outside the NC16A domain (23,(43)(44)(45)(46)(47)(48)(49). Notably, that reactivity with extracellular epitopes of the BP180 Cterminal domain appeared suggestive of atypical BP, i.e., with skin and MM involvements (44,46) or lesions limited to the lower legs and scarring of the toenail beds (49).
Other target antigens associated with the clinical MMP phe notype have been characterized molecularly: laminin 332, both α6β4 integrin subunits, and type VII collagen (4), respectively defining laminin 332 MMP, α6β4 integrin MMP, and MMEBA. MMP also includes MMLABD, with predominant IgA immune deposits along the BMZ.
Dipeptidyl peptidaseIV is not specific to insulinotropic hor mones. It is abundantly distributed, notably in the skin, on the surface of keratinocytes, sebocytes, fibroblasts, and T cells. DPPIV is involved in the regulation of DNA synthesis and cytokine production by those cells, for example, CCL11/eotaxin (57) and transforming growth factorβ1 (TGFβ1) (58)(59)(60). DPPIV is also a cellsurface plasminogen receptor that activates plasminogen conversion leading to more plasmin (61), which is a major serine protease known to cleave the 120kDa ectodomain of BP180, the reby generating LABD97 antigen (62). Role of eotaxin and plas minogen-plasmin system is well known in BP pathogenesis (63,64) and that of TGFβ1 in MMP is suspected (65). How gliptins induce BP or MMP by acting as DPPIV inhibitors on eotaxin, TGFβ1, and/or plasminogen/plasmin system remains to be elucidated.
Between 2011 and 2017, 14 case reports or small series reported 42 patients who developed BP while taking gliptins for their dia betes (Table S1 in Supplementary Material). The authors of those original articles individually scored gliptin accountability for each patient with the WHO-UMC system for 17 of them (10,15,20) and Naranjo's score for 6 (19), KarchLasagna system for 1 (16), and accountability was assigned a posteriori for the remaining 18 (9, 11-14, 17, 18, 21, 22) (See Table S1 in Supplementary Material).
17 BPs were probable gliptininduced ADRs, 23 were possible ADRs, and 2 BPs were most likely not gliptininduced because of long interval (>48 months) between gliptin intake and BP onset.
The first two comparative case-noncase studies were pub lished in 2016, after analysis of pharmacovigilance databases (16,25). Comparing French patients with BP ADRs to those with nonBP ADRs, Béné et al. showed that the former were associated more significantly and frequently with gliptin expo sure (odds ratio 67.5; 95% CI 47.1-96.9) and vildagliptin carried a higher risk than other gliptins. The individual gliptin account ability for those 42 BPs was rated as probable for 10 andpossible for 31, and not reported for one (Table S2 in Supplementary Material). In the European pharmacovigilance database, Garcia et al. identified 166 BPs reportedly induced by gliptin exposure. Using proportional reporting ratios, they found that BP was relatively more frequently associated with gliptins than with other drugs, again with vildagliptin being most strongly associated. Unfortunately, detailed clinical features were not reported.
Recently, a third welldesigned case-noncase study was pub lished (24). Comparing 61 diabetic BP patients to 122 age and  , and 17 of the suggestive-or-compatible gliptin-induced mucous membrane pemphigoid group, was performed as previously described (30). Positive controls were α6 β4 integrin, laminin 332, and BP 180 antigen; C− was a negative control with normal human serum. Abbreviation: MW, molecular mass.
patients ≥80 years old. Once again, vildagliptin had a stronger association but the study was underpowered to detect differences among the other gliptins.
Very recently, a study comparing 3,397 BP patients with 12,941 basocellular carcinoma controls from the Finnish nationwide registry has shown that vidagliptine and BP are significantly asso ciated with an adjusted odds ratio of 10.4 (4.56; 23.80) (26). The GliptinonsettoBPdiagnosis interval was of 449 days. Clinical and immunological data were not available in this study.
Last, some reportedly gliptinassociated BPs had atypical clin ical and/or immunological phenotypes, raising doubts about the BP diagnosis. Izumi et al. (23) described seven gliptinassociated BPs and showed they differed significantly from conventional BP by the absences of inflammatory lesions and circulating autoan tibodies targeting the BP180-NC16A epitope and the presence of autoantibodies targeting the midportion of BP180 (120kDa ectodomain and LABD97). However, those patients had no autoantibodies targeting the BP180 Cterminal domain, which could have suggested an MMP diagnosis. Sakai et al. (18) also had a patient with similar gliptinassociated BP. Mendoça et al. (15) reported a patient with mucous membrane involvement at diag nosis, including arytenoid edema with several ulcerated lesions covered with fibrin, raising the question of gliptininduced MMP rather than BP.
Finally, seven cases reported in the French pharmacovigi lance database as BP ADRs to gliptin were excluded from Béné et al. 's study (25) because they did not meet Kershenovich BP, and Vaillant BP criteria, suggesting that they might really have been MMP. sexmatched diabetic controls, those authors demonstrated a significant association between gliptin use and BP onset in uni variate analysis and after adjustment: 28 (46%) of the 61 diabetic BP patients took gliptins vs. 18% of the diabetic controls (odds ratio 2.64, 95% CI 1. 19-5.85; p = 0.02 in multivariable analysis). Stratified analyses showed a stronger association for men and This retrospective, monocenter study on a historical cohort was limited by MMP rarity. However, because our Center rec ruits patients with autoimmune bullous diseases, our findings should provide a fairly accurate appreciation of this population. Its retrospective design often means that data collection was incomplete and, indeed, some patients immunological test results are missing. In addition, we did not compare diabetic MMP cases to diabetic controls and epitope mapping of autoantibody targeted antigens was not done.
The potential of gliptins to induce MMP was not investigated previously. We identified 24 gliptintreated diabetic MMP pati ents, representing 38% of all MMP diabetic patients, a rate similar to that of gliptintriggered BP in diabetics (24). We evaluated chronological gliptin accountability in MMP induction case by case: it was incompatible, excluding gliptin's role, for 5 patients but suggestive or compatible for 17.
Vildagliptin was the most frequently incriminated gliptin for our 17 MMP diabetic patients, as for BP in the literature (Table S2 in Supplementary Material), but the highest intrinsic accountability scores were equally distributed among gliptins. It is worth noting that sitagliptin (and not vildagliptin) is the most prescribed in France, and the rest of Europe (66), suggesting that vildagliptin has a greater capacity to induce autoimmune bullous diseases.
The female/male ratio of these 17 MMP diabetics was higher than that of the general diabetic population and BP diabetics (1.1 vs. 0.7 vs. 0.65-0.83, respectively) but similar to that of our MMP controls ( Table 4). MMP diabetics were younger than the general diabetic population (36) and BP diabetics tended to be older than our MMP controls. Median time to MMP onset was longer, with a wider range, than for gliptininduced BP, which can be explained by the insidious evolution of mucous membrane lesions in MMP. The arbitrarily chosen time to distinguish suggestive (4 patients) from compatible (13 patients) chronology was ≤12 weeks; the 13 patients with compatible chronologies had intervals exceeding 36 weeks.
At MMP diagnosis, most of our 17 MMP diabetics did not have severe involvement. They differed significantly from MMP controls by their higher weights and their body mass indexes. Indeed, overweight and obesity is known to be a risk factor for type 2 diabetes. This difference was, therefore, expected from our population of type 2 diabetic MMP patients. They also had more frequent cutaneous involvement, less buccal involve ment, and absence of DIFdetectable IgA deposits along the BMZ. The MMP outcomes of these 17 patients, during and at the end of the first year of followup, were the same as that for MMP controls. Indeed, gliptinassociated MMPs responded well to usual treatments after gliptin withdrawal. Intriguingly, the endocrinologists had discontinued gliptins for all the 11 patients because of insufficient diabetes control. Gliptin triggering of MMP had never been suspected by dermatologists treating MMP patients.
Our immunological study results suggested that in vivo-fixed and circulating autoantibodies targeted multiple BMZ antigens/ epitopes. IEM showed immune deposits in the lamina densa with/without the LL as in "classical MMP" in seven patients, four of them ELISA BP180-NC16Apositive. The three BP180-NC16Anegative ELISAs were immunoblotpositive Using three accountability methods, WHO-UMC's criteria, Naranjo's score (most used worldwide), and Begaud's method (most used in France and Europe), we assessed gliptin imput ability in MMP induction. With the WHO-UMC accountability criteria, gliptin triggering of MMP was probable for 17 patients and unlikely for 7. With Naranjo's system, ADRs were considered probable for only 1 patient, possible for 16, and doubtful for 7 patients. Last, according to Begaud's method, with scores ranging from I0 to I6, only 5 patients (I4 for three, I3 for two) were given high accountability scores, 12 had low accountability (I2 for eight, I1 for four), and 7 were scored I0, meaning chronologically incompatible or undetermined challenge.
The results of this study demonstrated that gliptins are prob ably responsible for some MMPs. Hence, all doctors prescrib ing gliptins must be made aware of this potential toxicity. The practical consequence of that finding is that, as soon as a positive accountability score is established, by precaution, gliptins, which can be easily switched in the case of inefficacy or ADR, should be replaced by another antidiabetic drug. Importantly, all such cases must be reported as possible ADRs to a pharmacovigilance center.
Large case-noncase comparative studies need to be per formed to confirm or refute MMP induction by gliptins and better understand their pathogenic mechanism. Targetepitope mapping might help to determine whether a particular immune response occurs in druginduced MMP.