Characterizing Self‐Reported Daily Diary Epistaxis Measures in Hereditary Hemorrhagic Telangiectasia

ABSTRACT

Objectives

To characterize a proposed clinical trial outcome measure of weekly cumulative bleeding duration from daily nosebleed diaries, termed PRO‐CB (patient‐reported outcome of cumulative bleeding). We also assessed cumulative weekly gushing bleeding duration (PRO‐CGB) and evaluated the accuracy of patient recall on Epistaxis Severity Score (ESS) items 1–3 compared with daily diary data.

Methods

In this prospective observational study at the Toronto HHT Centre (St. Michael's Hospital, Unity Health Toronto), 23 patients from two clinical trials were observed during three‐month run‐in periods. Participants recorded each nosebleed episode daily, noting duration and intensity (gushing vs. non‐gushing). Data were analyzed using linear mixed‐effects models, correlation tests, and partial autocorrelation analyses.

Results

Diary compliance was 98.8%. Mean weekly PRO‐CB was 71.2 min (SD 81.1); mean episode duration was 5.9 min (SD 12.6). Daily and monthly means were 10.1 min (SD 20.1) and 307.6 min (SD 311.6), respectively. Gushing episodes lasted significantly longer than non‐gushing (mean difference: 5.81 min; 95% CI: 4.53–7.09). Bleeding patterns varied widely across individuals. PRO‐CGB also showed inter‐patient variability. PRO‐CB and ESS total scores were not significantly correlated (p = 0.18, r = 0.39). ESS items 1 and 2 correlated strongly with diary data (p < 0.001); item 3 did not. Agreement was 83% for item 1% and 70% for items 2 and 3.

Conclusion

Daily nosebleed diaries are a feasible and accurate tool for tracking epistaxis severity in HHT. Despite inter‐patient variability, PRO‐CB offers a meaningful outcome measure for epistaxis clinical trials.

Level of Evidence

2

This study investigates the use of daily self‐reported nosebleed diaries to measure epistaxis severity in patients with Hereditary Hemorrhagic Telangiectasia (HHT), minimizing recall bias. Two novel metrics—PRO‐CB (cumulative bleeding duration) and PRO‐CGB (gushing bleeding duration)—are introduced and compared with the traditional Epistaxis Severity Score (ESS). Findings demonstrate that diary‐based measures are feasible, reliable, and offer a practical alternative for clinical trials assessing epistaxis in HHT.

1. Background

Hereditary hemorrhagic telangiectasia (HHT) causes chronic recurrent nosebleeds (epistaxis) in more than 90% of adults and impacts quality of life. Long‐term outcomes with surgical and ablative therapies have been disappointing, with concerns regarding the safety of multiple recurrent ablative therapies. There is an urgent need for effective new therapies for epistaxis in HHT and for clinical trials testing newly identified potential therapies. As such, well‐characterized epistaxis measures are needed for efficient design of clinical trials.

Epistaxis severity can vary greatly between individuals and developing a meaningful epistaxis outcomes measure for clinic trials in HHT has been a challenge. The epistaxis severity score (ESS) was first developed in 2010 [1] from patient‐reported survey data, ultimately including six items (questions about typical symptoms over the past 3 months) that correlated with patient‐reported severity (mild, moderate, and severe) of their epistaxis. The score was also validated by comparing values to historical invasive treatment requirements and was later associated with physical and mental components of a health‐related quality of life survey [2]. While the ESS has been used to measure epistaxis in HHT clinical trials, its utility in this context may be limited by recall bias, as items ask about the last 3 months of epistaxis symptoms and ask the respondent to report the “typical” symptoms during this period. Further, the 3‐month time horizon of the ESS makes it impractical for assessment of symptom response to clinical trial interventions which may be over much shorter durations. In the first HHT trial for epistaxis, even the placebo group saw a reduction in ESS score, highlighting the possibility of patient recall bias contributing to the placebo effect [3].

The Nasal Outcome Score for Epistaxis (NOSE) in HHT is another validated measure which is calculated by assessing the impact of epistaxis on individuals with HHT across domains of physical problems, functional limitations, and emotional consequences [4]. Similar to the ESS, it relies on patient recall of severity and includes additional variables not directly related to bleeding characteristics.

We have proposed a new clinical trial outcome measure, PRO‐CB (patient reported outcome, cumulative weekly bleeding duration), calculated from patient‐reported daily nosebleed diary data regarding nosebleed duration, frequency, and intensity. Here, we report daily epistaxis diary data to characterize PRO‐CB, from the “untreated” portion of two clinical trials in HHT patients.

2. Methods

This study was conducted at the Toronto HHT Centre, Division of Respirology, St. Michael's Hospital, Unity Heath Toronto. Data were collected from the participants of two clinical trials which were approved by the local Research Ethics Board (19–243 and 17–294), NCT03397004 and NCT04646356. All participants provided informed consent for the anonymous use of their data. We collected natural history data from the initial 3‐month run‐in period before any intervention. Inclusion criteria for these two trials required that individuals were at least 18 years of age and had at least 15 min per week of cumulative epistaxis in the prior month. Data were collected using paper‐based daily nosebleed diaries, completed independently by participants, with initial guidance and then check‐ins by the research coordinator in weeks 1, 2, 4, 6, 8, 10, and 12 to ensure adherence. Participants recorded each individual bleeding episode, its duration, and whether it was gushing or not. Gushing bleeds were described to participants as a steady stream of blood instead of just dripping. We reported a high diary completion rate (98.8%) of daily nosebleed diaries in HHT, using the same strategy, in the full report of the Doxycycline for Epistaxis in HHT trial (NCT03397004) which is one of the two clinical trials analyzed in the current study [5].

2.1. Primary Outcome Measure

The primary outcome measure for this study is the weekly cumulative bleeding duration (in minutes), referred to as PRO‐CB. This measure is calculated by summing all the bleeding minutes that occurred within a single week.

2.2. Secondary Outcome Measures

1. Weekly cumulative gushing bleeding duration (PRO‐CGB): This measure is calculated by summing all the gushing bleeding minutes that occurred within a single week, similar to PRO‐CB but without any non‐gushing bleeds included.

2. Episode, daily, and monthly cumulative durations: The episode duration is the individual bleed duration in minutes. The daily cumulative bleeding duration is calculated by summing all the bleeding minutes within 24 h. The monthly cumulative bleeding duration is calculated by summing all the bleeding minutes within 30 days.

3. Percentage of gushing of overall duration: Calculated as the total bleeding minutes classified as gushing, divided by the cumulative total bleeding minutes.

4. ESS: The epistaxis severity score (ESS) assesses the severity of nosebleeds reported by participants recalling a 3‐month period, as previously defined [1, 5].

2.3. Analyses

We use descriptive statistics to characterize participants' nosebleed data and calculated the coefficient of variation expressed as a percentage (standard deviation divided by the mean × 100) for episodic, daily, weekly, and monthly bleeding minutes. Nocturnal bleeds were pre‐defined as 5 min in duration when duration was not recorded by the participant. A linear mixed‐effects model with a random intercept for each patient was used to compare the episode bleeding durations between gushing and non‐gushing bleeds. We used graphical displays (line plots, kernel density plots) to display individual and aggregate distributions of bleeding and partial autocorrelation function plots to assess the correlation between subsequent measurements for both PRO‐CB and PRO‐CGB. Partial autocorrelation analyses were conducted to determine if PRO‐CB data points are systematically related to any data points at previous time points. Jonckheere's trend test was selected to evaluate trends across ordered ESS categories without assuming normality of the underlying distributions. Partial autocorrelation was used to assess temporal dependence in repeated weekly diary measures while accounting for lag effects inherent to within‐subject longitudinal data. A Pearson correlation test was used to assess the correlation between PRO‐CB means and ESS itself, while a Jonckheere's trend test was used to compare item 1 and 2 from ESS to the corresponding diary data. For item 3, a two‐sample t‐test was used to compare percent gushing bleeds for everyone, based on their gushing or not selection in ESS. For item 1, the categories were divided into numerical ranges to allow for agreement analyses. For item 2, mean episode duration was used for agreement analyses. For item 3, a gushing % of total episodes of over 50% was considered equivalent to “typical.” Future study sample size estimates were calculated based on summary statistics of the PRO‐CB. The estimation was conducted to determine the sample size needed to achieve 80% or 90% power with a 5% Type I error rate. All analyses were performed using R version 4.3.0.

3. Results

3.1. Summary Statistics

Data from 23 individuals were included (9 male, 14 female, mean age = 61 years). Diary compliance was 98.8% over 3 months. Mean baseline ESS was 5.3 (SD = 0.99). Weekly duration (PRO‐CB) mean was 71.2 min (SD = 81.1, median = 40, IQR = 20–95, range = 0–555). Episode duration mean over a 3‐month baseline period was 5.9 min (SD = 12.6, median = 3, IQR = 1–5, range = 1–300). Daily duration mean was 10.1 min (SD = 20.1, median = 4, IQR = 0–11, range = 0–310). Monthly duration mean was 307.6 min (SD = 311.6, median = 190, IQR = 91–434, range = 16–1667). Coefficient of variation for episodic bleeds was 214%, for daily was 199%, for weekly 114%, and for monthly was 101%. Gushing bleeds comprised 555/2963 (18.7%) of total episodes over 3 months, ranging from 0% to 57% of episodes across patients. Mean gushing episode duration significantly exceeded non‐gushing (difference: 5.81 min, 95% CI: 4.53–7.09 min, p < 0.001).

3.2. PRO‐CB Analyses

The PRO‐CB demonstrated considerable variability within and between patients (Figure 1). Partial autocorrelation plots demonstrated no significant correlation between subsequent PRO‐CB measures. While cohort‐level summaries suggest modest central tendency, individual patients demonstrated markedly different bleeding distributions, highlighting substantial inter‐individual heterogeneity that is not fully captured by summary statistics alone. A kernel density plot showed a skewed distribution of PRO‐CB scores (Figure 2). Individual distributions are also reported to highlight the differences in peak shape and individual bleed spreads (Figure 3).

FIGURE 1.

PRO‐CB line graph over 3‐month period (N = 23).

FIGURE 2.

PRO‐CB kernel density plot (N = 23).

FIGURE 3.

PRO‐CB kernel density plot, individual (N = 23).

The PRO‐CGB also demonstrated considerable variability within and between patients (Figure 4). A kernel density plot showed a skewed distribution of PRO‐CB scores (Figure 5). Individual distributions highlight the differences in peak shape and individual bleed spreads (Figure 6).

FIGURE 4.

PRO‐CGB line graph over 3‐month period (N = 23).

FIGURE 5.

PRO‐CGB kernel density plot (N = 23).

FIGURE 6.

PRO‐CGB kernel density plot, individual (N = 23).

3.3. Patient Recall Assessment

Collected ESS data was compared with diary data to assess patient recall. Table 1 summarizes the data used for comparison. The numerical categories used to determine agreement for ESS item 1 are shown in Table 2. Comparison of PRO‐CB mean and ESS score showed no significant correlation (p = 0.18, corr = 0.39). For Item #1 (typical bleed frequency), a statistically significant positive correlation was found between ordered categories from ESS and total number of episodes over entire 3‐month period from diary (p < 0.001). In addition, Item #1 agreement between ESS categories and diary data frequency was 83%. For Item #2 (typical bleed duration), a similar positive correlation was found between ordered categories from ESS and mean episode duration from diary (p < 0.001). However, Item #2 agreement between ESS categories and mean episode duration was only 70%. A two‐sample t‐test was used to compare the mean percentage of gushing episodes (from diary data) between participants who selected “typically gushing” and those who selected “not typically gushing” on ESS Item #3. The difference between the two groups was not statistically significant (p = 0.19), indicating that the self‐reported ESS classification of gushing was not associated with a higher actual percentage of gushing bleeds. Agreement for Item #3, between binary category and gushing percentage of total episodes was only 70%.

TABLE 1.

ESS comparison data with respective diary data.

Participant	ESS	PRO‐CB	Item 1	# of bleeds in 3 months	Item 2	Average bleed duration	Item 3	Gushing % of total bleeds
1	4.35	277	Several per day	152	6–15	23.7	No	7
2	3.84	40	Once per day	82	6–15	6.3	No	10
3	4.24	59	Several per week	78	6–15	9.8	Yes	27
4	4.53	30	Several per day	122	1–5	3.2	No	0
5	4.53	26	Several per day	153	1–5	2.2	No	13
6	5.25	30	Several per day	204	1–5	1.9	No	46
7	7.28	146	Several per day	147	6–15	12.9	No	20
8	4.96	99	Several per week	64	16–30	20.0	No	41
9	2.62	17	Several per week	40	1–5	5.4	No	0
10	3.44	36	Several per day	96	1–5	4.9	No	2
11	3.44	21	Several per day	155	1–5	1.8	No	1
12	4.53	79	Several per day	250	1–5	4.1	No	14
13	4.35	79	Several per day	250	6–15	4.1	No	14
14	5.25	20	Several per day	67	1–5	4.0	No	49
15	4.75	138	Once per day	179	6–15	10.0	Yes	35
16	6.67	9	Once per week	11	6–15	11.1	Yes	27
17	6.09	28	Several per week	31	6–15	10.7	Yes	45
18	5.29	88	Several per day	215	1–5	5.3	No	16
19	5.43	185	Several per day	261	6–15	9.2	No	1
20	4.96	12	Several per week	38	6–15	4.1	Yes	5
21	6.2	120	Several per day	100	6–15	15.7	No	53
22	3.44	28	Several per day	172	1–5	2.1	No	14
23	4.69	92	Several per day	210	6–15	5.7	No	0

Note: Item 1 ESS question was “How often do you typically have nosebleeds?” Item 2 was “How long is your typical nosebleed (minutes)?” Item 3 was “Is your typical bleed gushing or not?”.

TABLE 2.

Categories used to determine agreement for Item 1 of ESS score.

Category	Range (bleeds in 3 months)
Several per day	> 90
Once per day	60–89
Several per week	12–59
Once per week	4–11
Once per month	1–3
Less than once a month	0

3.4. Examples of Sample Size Estimate for Future Clinical Trials

Sample size estimates were conducted using the summary statistics of PRO‐CB over 3 months (mean = 71.1, SD = 82.1), as well as PRO‐CGB (mean = 38.1, SD = 42.2), monthly total cumulative duration (mean = 307.6, SD = 311.6), and monthly gushing cumulative duration (mean = 125.9, SD = 145.9). For example, achieving a 50% reduction in PRO‐CB with 90% power and 5% alpha would require approximately 56 participants using a paired t‐test. As expected, larger effect sizes (i.e., greater reductions in bleeding duration) required fewer participants, while smaller effect sizes required substantially larger sample sizes to maintain adequate power.

4. Discussion

This study characterized the PRO‐CB measure using daily nosebleed diaries from HHT patients. The demonstrated variability of bleeding duration and intensity across patients highlights the challenge of measurement of this outcome for clinical trials. However, we also demonstrated the feasibility of collecting prospective daily diary data, rather than relying solely on recall measures. As such, we propose PRO‐CB as an additional clinical trials outcome, providing prospectively collected shorter term detailed data. We report the basic characteristics of PRO‐CB measurement for incorporation into clinical trials design and discuss some areas that require further development.

Analyzing the nosebleed diary data aggregated over various time frames has provided valuable insights into its characteristics. The high variability observed across all time frames, as indicated by the coefficients of variation, is not unexpected. Temporal fluctuations in bleeding patterns are common and typical in HHT. Despite this variability, these data remain accurate representations of individual bleeding patterns and are our best resource for developing effective epistaxis therapies. The distribution of PRO‐CB varies by individual and highlights the importance of assessing baseline epistaxis severity on a case‐by‐case basis. Even with inclusion criteria in place, individual distributions vary widely, especially in individuals with longer and more frequent bleeding in general. It is important to note that the “outliers” in the data do represent real occurrences and are comprised of the longest bleeds (and mostly gushing). Removal of these outliers for analyses would omit some of the most significant bleeds that may have the largest impact on patients' blood loss and likely quality of life; therefore, there must be careful consideration before eliminating these as statistical outliers.

PRO‐CB has several other interesting characteristics. The lack of partial autocorrelation suggests that weekly bleed durations are independent of each other within an individual, making statistical analysis more straightforward. Further, the observed agreement and correlations between diary data and specific items within the ESS score, but not with the overall ESS score itself, further reinforce the use of diary data measures, rather than aggregated scores of the ESS and NOSE. High daily diary compliance in the relevant studies also indicates that patients can adhere to the symptom tracking over extended periods of time.

Despite the benefits of incorporating a diary‐based method to collect epistaxis data for HHT trials, recall based measures like the ESS or NOSE scores are still valuable tools for HHT clinicians and researchers. These measures are less labor intensive and can provide an estimate of epistaxis severity over a long period of time. As such, recall based measures can be valuable for epistaxis severity stratification for clinical trials design, even when using daily diaries for trial outcomes measures. Recruitment of a study population who are dedicated and diligent with diary tracking may not be feasible for all study designs. In addition, recall based measures remain practical in clinical practice for clinicians to assess the severity of their patients' symptoms quickly.

While PRO‐CB provides a complete measure of weekly epistaxis duration, we have also defined PRO‐CGB as a relevant outcomes measure for HHT trials. Despite its higher variability, the fact that gushing bleeds represent a higher volume of blood by definition and that they are significantly longer than non‐gushing episodes means that these episodes should be carefully considered and will have meaningful impacts on patients' blood loss. Furthermore, the binary “gushing or not” distinction within the diary provides for a simple and quick way to include intensity data without excessive participant burden. However, other studies have weighted measures or scales to capture the intensity of bleeding episodes which, while more detailed, may increase the tracking burden on participants [6]. This suggests that a scaled approach is also a feasible method for assessing intensity, with potential trade‐offs in participant effort. Notably, despite the availability of diary‐based measures, contemporary randomized clinical trials for epistaxis in HHT have continued to use the ESS as a primary endpoint [7].

Lastly, we have presented researchers with sample size estimates based on epistaxis characteristics from our study participants. These estimates are intended to inform future clinical trials, aiding in the determination of appropriate recruitment targets needed to detect clinically meaningful changes in PRO‐CB.

Beyond informing sample size calculations, PRO‐CB may also support the future development of thresholds for clinically meaningful change using anchor‐based or distribution‐based approaches, although such thresholds were beyond the scope of the present study. To facilitate broader implementation in larger or less closely monitored trial populations, strategies such as electronic diaries, simplified reporting frameworks, or hybrid diary–recall designs may help mitigate diary fatigue while preserving data quality.

Limitations of this study include the small sample size, missing duration data for some nocturnal bleeds, and the use of a single specialized HHT center, where many participants had prior trial experience. This highly engaged cohort likely contributed to the unusually high diary adherence observed and may not reflect the broader HHT population. As a result, the generalizability of these findings to less engaged populations or non‐specialized clinical settings may be limited. The paper‐based diary format also introduces potential bias, including delayed entries, back‐filling, and recall error, and compliance may be lower outside a closely monitored trial environment. The definition of “gushing” bleeds relies on patient perception and may vary between individuals, contributing to the modest agreement with ESS item 3. Finally, PRO‐CB and PRO‐CGB were not compared with quality‐of‐life or other clinical outcomes, and future studies using electronic time‐stamped diaries and more diverse cohorts are needed to further validate these measures.

5. Conclusion

Here we characterized the clinical trial measure PRO‐CB as a feasible method to monitor epistaxis severity over time in a controlled clinical trial setting environment. We demonstrate the benefits and also challenges of patient‐reported daily diary measures. The high variability of PRO‐CB between and within individuals stems from the nature of the disease and further reinforces the use of daily diaries to capture and understand variability and outliers. In future clinical trials, we hope this work can be used in the experimental design of HHT‐related clinical trials to ensure comprehensive evaluation of drug efficacy.

Funding

This study was supported by U.S. Department of Defense grant (W81XWH‐17‐1‐0429). Dr M.E.F. was also supported by the Li Ka Shing Knowledge Institute. Dr. A.M.B. was supported by the Baxter and Alma Ricard Chair in Inner City Health at Unity Health Toronto and the University of Toronto.

Ethics Statement

This study was conducted in accordance with the ethical standards outlined by the Unity Health Toronto Research Ethics Board and received approval prior to study commencement. All participants provided full consent for the anonymous use of their relevant data.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.