**3. Results**

#### *3.1. Trial Population*

After screening 155 patients, 50 patients were included and randomized between May 2017 and August 2018 in three different centers (Lyon, Clermont-Ferrand, and Montpellier). The baseline characteristics are summarized in Table 1. All individuals except one met the inclusion criteria and were enrolled in the study. Due to one wrong allocation in the placebo group and to one wrong inclusion, the independent committee recommended that it was necessary to include two more patients (50 instead of the 48 patients initially scheduled) (Figure 1). One patient in the placebo group discontinued his follow-up due to a severe adverse event. All patients in the tacrolimus group filled in the epistaxis grids and completed six weeks of treatment. Twenty-five patients in the placebo group filled in the epistaxis grids and out of them 24 completed six weeks of placebo treatment.

**Figure 1.** Flow chart.



#### *3.2. Response to Treatment*

#### 3.2.1. Primary Outcome

All patients in the intention-to-treat population were analyzed (*n* = 50). The results are summarized in Table 2. According to our main outcome (30% reduction in the total duration of epistaxis over six weeks after treatment), no statistical difference was observed in the tacrolimus groups compared to the placebo group. Analysis of the per protocol population led to the same conclusions.

**Table 2.** Main outcome (*n* = 50) analysis: <sup>E</sup>fficacy of tacrolimus ointment on mean epistaxis duration six weeks before and after treatment.


Legends: \* risk difference (RD) and relative risks (RR) for binary outcomes; and Cohen's d for quantitatives.

## 3.2.2. Secondary Outcomes

Duration and number of nosebleeds before, during, and after treatment are presented in Figure 2. As for percentage of improvement, there was no difference between the tacrolimus group and the placebo group regarding evolution in the parameters during the six weeks following treatment. However, there was a difference in evolution when comparing epistaxis before and during treatment in terms of epistaxis duration (*p* = 0.04, Cohen's d: 0.53 (–0.04–1.11) and epistaxis number (*p* = 0.04, Cohen's d: 0.39 (−0.19–0.96)) (Figure 2).

**Figure 2.** Mean epistaxis duration and number on six weeks before, during and after treatment. Legends: Lines from bottom to top: minimum, 25th percentile, median, 75th percentile and maximum. Diamond: mean. Small squares: individual data.

*J. Clin. Med.* **2020**, *9*, 1262

The number of red blood cell transfusions did not di ffer before and during treatment (*p* = 0.57, Cohen's d: −0.21 (−0.77–0.36)), or before and after treatment (*p* = 0.69, Cohen's d: −0.23 (−0.8–0.34)). Three patients had blood transfusions before treatment and all of them were in the placebo group. Of them, two patients also had blood transfusions during treatment.

The SF-36 questionnaire revealed no di fferences in the dimensions of quality of life before, during, and after treatment (Table 3).

The di fference in the ESS after and before treatment in the tacrolimus group (mean = –0.43 (SD = 1.47)) and the placebo group (mean = −0.26 (SD = 0.99)) (*p* = 0.69, Cohen's d: 0.13 (–0.53–0.8)), and during and before treatment in the tacrolimus group (mean = -1.47 (SD = 1.55)) and the placebo group (−0.96 (SD = 1.26)) (*p* = 0.31, Cohen's d: 0.36 (−0.35–1.07)) were not significantly di fferent.

The biological criteria (hemoglobin and ferritin levels) did not significantly improve during or after treatment. The e ffect sizes for evolution in these parameters were 0.38 (−0.2–0.95) and 0.11 (−0.46–0.69) during treatment, and were even lower after treatment (0.19 (−0.38–0.76) and −0.32 (−0.9–0.26)), respectively. Mean levels on inclusion, six weeks after the beginning of the treatment, and six weeks after the end of the treatment in both the tacrolimus and placebo groups were 12.8, 13.3, 13.0 and 12.6, 12.6, and 12.6 for hemoglobin (g/dL), and 51.3, 75.6, 43.9 and 49.0, 59.9, and 69.3 for ferritin (μg/L), respectively. Using mixed models, we did not find a significant trend over time for hemoglobin levels (*p* = 0.61) or ferritin levels (*p* = 0.36), and no di fference in evolution between groups was observed (*p* = 0.59 for hemoglobin and 0.60 for ferritin).



*J. Clin. Med.* **2020**, *9*, 1262

#### *3.3. Safety Outcomes*

A total of 51 AE (25 in the tacrolimus group and 26 in the placebo group) and three severe adverse events (SAE) (one in the tacrolimus group and two in the placebo group) were recorded without differences between the groups. No SAE certainly or probably related to the treatment were recorded.

Of the 29 possibly or probably related AE, 13 patients had nose burning or a tingling sensation (10 in the tacrolimus group and three in the placebo group), two had infection, both in the tacrolimus group (genital HSV infection (*n* = 1), intercostal VZV infection (*n* = 1)), seven had a local sensation (burning eyes (*n* = 1), burning throat (*n* = 2), sneezing (*n* = 1), rhinitis (*n* = 1) and nose smell (*n* = 2) (four in the tacrolimus group and three in the placebo group) and five had other symptoms (thoracic pain (*n* = 1), back pain (*n* = 2), diarrhea (*n* = 1) and headache (*n* = 1), all in the placebo group.

Systemic absorption of tacrolimus: as expected, all FK506 dosages were < 5.0 ng/mL on day 8 (*n* = 48), day 22 (*n* = 46) and day 43 (*n* = 46) after the beginning of the treatment. Tacrolimus was detected in two patients (4.3%), both in the tacrolimus group on day 8 (1.2 and 1.02 ng/mL), four patients (8.3%) on day 22 (1.03, 1.02, 1.2 and 1.05 ng/mL), and not detected on day 43 (0%).

## **4. Discussion**

To our knowledge, this was the first phase II, double-blind, multicenter, randomized, placebo-controlled trial evaluating the efficacy of tacrolimus nasal ointment on epistaxis in HHT. In the present study, there was no significant benefit to using tacrolimus ointment administered twice a day on the nasal mucosa for six weeks after the end of the treatment, compared with placebo. We chose this time point for our main outcome (before treatment vs. after the end of the treatment) on the basis of physiology and on our previous studies using anti-angiogenic drugs which usually improved patients many weeks after the beginning of the treatment. Tacrolimus enhanced the ALK-1 signaling pathway in the endothelial cells of HHT patients, and inhibited increased VEGF signaling and hypervascularization in an HHT animal model [11], and we hypothesized that the effect which involved "blood vessel remodeling" would not be immediate. Furthermore, recent clinical data in one patient treated with low oral doses of tacrolimus showed that ESS improvement was observed three months after the beginning of the treatment [22]. However, we did not take into account that this treatment was local and with good local absorption, as confirmed by blood dosages which were very low but positive in four cases, and maybe had a quick effect during treatment which stopped after the end of the treatment. Moreover, the vast majority of patients in the tacrolimus group had a prior history of nasal surgery, compared to only half in the placebo group. It can be argued that this disequilibrium may have blunted the apparent effect of tacrolimus during the treatment period and/or contributed to the lack of an obvious effect after tacrolimus was halted. Similarly, fewer women were assigned to the tacrolimus than the placebo group, which may have influenced the results.

Importantly, we observed for our secondary outcome a significant improvement in epistaxis duration and number during the treatment phase in the tacrolimus group. These results sugges<sup>t</sup> that the effect of the drug occurred only during treatment and patients relapsed after they stopped the treatment. However, although this trial was randomized, we know that moistening nasal mucosa improves epistaxis in HHT [10], and we cannot exclude the idea that the efficacy observed in the treatment group was partly related to the effect of the ointment, not to the drug itself. Furthermore, this observation highlighted the fact that prolonged tacrolimus use would be necessary, and we do not have data on the long-term safety on mucosae. The ESS did not improve during treatment; however, even though the ESS has been shown to be a validated tool in the evaluation of epistaxis in HHT [23,24], we chose not to use it as our main outcome. This was firstly because our patients are used to completing epistaxis grids, and secondly because the ESS is self-administered by patients and is more subjective than an epistaxis duration measurement. It effectively includes subjective questions such as intensity, need for medical attention and anemia, not defined by hemoglobin level but by our patients' own evaluation.

Tolerance of the nasal ointment was good after a 6-week treatment. No severe adverse events were observed. The most frequent related adverse event was a sensation of burning in the nose in 34.5% of patients in the tacrolimus group, but it was transient in most cases. None of the patients stopped the treatment for this reason. This result was similar in non-HHT patients receiving tacrolimus ointment on mucosae (oral lichen planus) in randomized studies. Corrocher et al. [16] and Vohra et al. [17] observed a burning sensation in 56% and 35% of patients, respectively, treated for oral lichen planus, and, similarly, this event resolved rapidly within four to five days. In the present study, we observed two infections in the tacrolimus group but not locally, one genital and one skin infection, and the pharmacokinetics monitoring performed revealed moderate systemic absorption, almost undetectable in most cases. It is thus unlikely that these complications were related to an immunosuppressive e ffect of tacrolimus. Other adverse events were observed with the same frequency in both groups. Nasal cartilaginous septum perforation was followed closely and not observed in either group after treatment.

This trial had several significant limitations. First, patients completed epistaxis grids and noted epistaxis duration, which are not directly observed outcomes, partly subjective and imprecise, and are subject to error. Second, we included all HHT patients with nosebleeds and did not take into account a history of nasal surgery or nasal crusts or septal perforation, which may change mucosal drug absorption. Almost all patients had undergone di fferent types of surgery and most of the mucosa of the nasal cavity cannot be touched by a cotton swab, thus possibly resulting in under treatment. Third, the blinding process was successful in most cases; however, burning sensations were more frequent in the tacrolimus group. This could lead to underestimation of the placebo e ffect and overestimation of the relative treatment e ffect because patients can deduce that they are in the placebo arm and may be less likely to report improvement [25]. Finally, we observed a decrease in epistaxis duration in the placebo group, during and after treatment, which could be partly due to a Hawthorne e ffect. While nasal moisturizing is a recommended treatment for preventing HHT bleeding, participating in a study would probably improve the way it was done.
