FDA draft guidance blood pressure monitoring in clinical trials – questions answered
Jeff Heilbraun, MS – VP Medical Affairs, Cardiac Safety
Within the clinical trials arena, there are many references of blood pressure monitoring in protocols across all study phases and therapeutic indications. Blood pressure is often associated within a study’s schedule of events as a standard vital sign or safety assessment.
Specific to the development of hypertension therapeutic treatments, blood pressure assessment becomes a primary endpoint. Methods of capturing blood pressure data have evolved from initially being solely captured in the clinic setting, using a Mercury manometer and stethoscope to capture the BP values to more recent focus on implementation of standardized automated oscillometric BP devices. In addition, the implementation of Ambulatory Blood Pressure Monitoring (ABPM), which allowed for 24-hour, out-of-clinic (ambulatory) data collection, has become a recognized standard for defining a hypertension drugs BP effect over a 24-hour period. The continuum of modalities then expanded to include remote or telemonitored home BP for the acquisition of BP data over an extended period of time (days/weeks/months), providing additional insight into BP trends. These technologies become complimentary methods of evaluating a study participant’s BP and response to developing medications.
Stepping forward into 2018, the FDA released a draft Pressor Effect guidance, which brought a unique and new focus on the “off-target” effect (increase) in blood pressure of all therapeutic agents in development. The draft Pressor Effect guidance went through public review and discussion leading to an updated release of the draft guidance in February 2022.
On April 27, 2022, the cardiac safety team at Clario provided an overview of the updated draft guidance, which can be found through the following link.
This blog post provides the Q&A responses from questions that were generated through the webinar. An additional purpose of the blog is to continue the discussion related to the draft Pressor Effect guidance as well as general considerations of blood pressure monitoring within the clinical trials arena, including developing technologies and additional hemodynamic endpoints outside of traditional brachial BP values.
On behalf of the team at Clario, we look forward to addressing additional questions and an engaging and insightful dialogue with members of the clinical research, medical and scientific research, medical device and regulatory community.
1 a) Is there any guidance consideration on preferred ABPM durations, e.g., 24 or 48 hours for pediatric or adult clinical trials; b) Are there any specific recommendations by the FDA on cardio-oncology trials or stress upon significance of BP assessments in such trials?
In answer to (a), the FDA hasn’t opined on the duration of ABPM sessions, but it has always accepted 24-hour assessments. To be honest, wearing an ABPM device for 48 hours is challenging. It can interfere with sleep, and even during the day, is somewhat obtrusive. Asking a patient to wear this for a safety assessment, when they are not necessarily going to personally benefit from the information, is difficult enough. Asking for 48 hours leads to compliance issues.
In relation to the second part of the question (b), the FDA hasn’t made any specific comments about oncology trials. As you know, many TKIs and VEGF inhibitors do have BP effects, and the FDA considers it important to understand the magnitude of the BP increases. This isn’t about approval, but simply about informing prescribing oncologists about the safety concern and what type of monitoring and mitigation strategies are important.
2 One of the concerns with an ABPM sub-study is the size of the trial. Knowing the variability within and between subjects and being able to assess small changes in BP may be challenging. Use of home BP monitoring in the overall Phase III clinical study will allow for 24 monitoring but with the advantage of a larger sample size. Home BP monitoring will also be more acceptable to a greater number of patients. Why did the guidance not include home BP monitoring as an alternative to the ABPM sub-study?
As noted, the sample size needed to exclude a small treatment effect of only 3 mmHg is large, due to the inherent variability in this data type. This presents challenges to budget, and feasibility, for sponsors. To clarify, a sub-study with ABPM would be performed typically on an outpatient basis (subjects wear the 24 hour monitor at home and come to the clinic only for it to be initialized properly, and subsequently to return the device). “Home BP” as we referenced it in the webinar refers to a provisioned, standardized BP cuff programmed for triplicate inflations that is used at home by the patient independently, at set time points. This augments the data collected at scheduled clinic visits. Home BP isn’t used overnight; it is just for collection of individual triplicate readings which are initiated by the subject when awake. The role of Home BP is not specifically addressed in the guidance currently. I suspect this is because the agency is focused on assessing full 24 hour data (including overnight BP response), with the concern that daytime-only data could be falsely negative. From a patient safety perspective though, Home BP is quite valuable, providing insight to changes in blood pressure at time points that would be otherwise missed, unless the clinic visit schedule was made much more dense. Timely detection of a hypertension safety signal at home could allow for earlier intervention, if necessary. And, in phase II of development, home BP data in patients could allow a sponsor to detect a potential hypertension signal more readily by supplementing the limited dataset typically available from clinic vital signs only.
3 I did not notice anything in the guidance document regarding standardization of BP assessments at baseline to correct for variabilities within/between subjects (e.g., differences in absolute BP, use of antihypertensives and other meds that have known pressor effects, differences in baseline BP — normotensives vs. hypertensives, co-morbidities, CV risk status). Higher variability at baseline makes interpretation of BP findings challenging.
Some of the variables noted here could be managed through eligibility criteria for the study, but care should be taken not to refine the study population to the point at which it no longer accurately reflects the intended treatment population. The principle is clear, that FDA requires this ABPM data to be generated in the intended treatment population. If a new indication or new population is to be explored, additional ABPM data could be requested for this submission rather than relying on negative data from a different population. The rationale for this incorporates the same issues raised in this question. Comorbidities, concomitant medications, baseline CV-risk, could all influence how a patient responds to a new chemical entity. Certain populations, such as chronic renal disease, have higher BP variability than others, and this may also need to be taken into account when considering an appropriate sample size for a dedicated blood pressure study.
4 An earlier speaker mentioned designing studies “to exclude a 3 mmHg increase in BP” when talking about statistical design. Could you talk more about that (or what is meant by “exclude” in that context) for those who may not be as strong in stat design?
The updated draft guidance now states the following: “the study should be powered to exclude a 3-mmHg increase in 24-hour average systolic blood pressure using an upper bound of the two-sided 95% confidence interval, assuming the true effect is 0 mmHg.”
Importantly, this indicates that the objective of the analysis is not just to demonstrate that the mean change from baseline is less than 3 mmHg. Rather, the objective is to calculate the 95% confidence interval around that mean change from baseline and demonstrate that the upper bound of this confidence interval is less than 3 mmHg. This approach is taken to increase likelihood that a negative analysis result is a true negative, and that a BP effect was not missed due to variability in the data (i.e., the risk of a false negative result is reduced).
5 In the sample size estimates for ABPM once at baseline and once on-treatment. At what point can a study take advantage of the repeated assessments at home and do you think it will reduce the number of required samples?
The selection of time points for ABPM is highly program-specific, as there are several factors that need to be considered. As an example, drugs with a direct, exposure-dependent effect on vascular tone will be handled differently than drugs that, indirectly, have an effect on sodium retention and fluid status over time. The guidance emphasizes the importance of selecting a time point after the drug has reached a “steady-state effect on blood pressure,” which is different to a simple PK-based steady state exposure definition. The mechanism by which the drug influences blood pressure must be understood to determine steady state of the pharmacodynamic effect. It is plausible that an initial hypertensive effect could wane with time, and, therefore, an ABPM session at day 7 versus week 8 could have different results for the same drug. For the intentions of this guidance, interest is on the longer term/steady-state effects, and the change from baseline comparison at this time point.
The guidance does not suggest having multiple on-treatment ABPM time points, but rather describes selecting a single time point most appropriately. If the analysis included multiple time points and the burden was to demonstrate an upper bound less than 3 mmHg at each of those time points, the sample size could actually be increased overall (as we’d assume that standard deviation at each time point is the same).
6 Regarding data cleaning and setting implausible ranges, if a site mistakenly enters, for example, 21 instead of 121 for systolic, is it preferred that you exclude this data point instead of having them update?
Manual transcription data entry errors like this can certainly happen. This is a key advantage to using a blood pressure collection modality with automated electronic capture of results into the clinical trial database. A process for correction of data entry errors by site staff would fall within the scope of the sponsor’s SOPs and need to be handled as determined by those standards. As discussed on the webinar though, we do typically recommend having a process for excluding biologically implausible data from the analysis. A range of plausible values can be included in the statistical analysis plan (pre-specified), and if a value of 21 mmHg for SBP was present, for example, it would typically be excluded from the analysis.
7 How is the steady state defined? FDA guidance does not explicitly call out small drug vs. the scientific background was calling small drugs explicitly. Does the guidance apply to any medical product in development?
First- the definition of ‘steady state’ for BP effects is not fixed. When we think about QT prolongation, things are simple — drug-induced QTc prolongation due to direct hERG channel block is exposure dependent and tracks exposure very nicely. Steady state for QTc effects means PK steady state (for parent and metabolites). For drugs that alter hERG channel trafficking, things are not as clear, but still follow PK, if with a long delay of hours to days.
In contrast, drugs may increase BP by multiple different mechanisms. Some drugs are direct arterial vasoconstrictors, and produce BP effects immediately, tracking plasma concentration very closely. Other drugs affect the renin-angiotensin axis or fluid status, and the PD effects don’t track the PK. A number of drugs have effects that wax or wane over time. In the guidance, the FDA recommends doing the on-treatment ABMP session at 4 weeks, but, in practice, I’ve seen them recommend waiting as long as 8-12 weeks. It really depends on the PK of the drug as well as the mechanism of the BP increase. However, in general, 4 weeks is about the minimum unless you have a known vasoconstrictor.
In response to the second question, the guidance does not discuss molecule size as a consideration. We know that some biologic macromolecules can affect BP directly or indirectly. The FDA pays more attention to duration of use — for instance, a vaccine or monoclonal antibody that is administered once or twice would not be required to collect ABPM data. However, a biologic that is administered over a period of months or years might be subject to this guidance. Part of the equation would be figuring out the “BP load” — in other words, the product of the magnitude of the BP increase and the duration that it will be present. If a biologic is given once a month and increases systolic BP by 4 mm Hg for only 3 days a month, the concern would be much lower than for a drug that produces a 3 mm Hg increase for 20 days a month.
8 When having sites take triplicate, is it advised to throw out the first measurement and use the last two for analysis, since some guidelines suggest that the first can be slightly elevated?
The present draft guidance references capture the triplicate BP assessment and generate an average of the three readings (similar to the SPRINT Trial).
Depending on the specific scientific working group, there have been different inflation schedules promoted for use with AOBP.
The AHA Scientific Position Paper (Munter et al. Hypertension. 2019;73:e35–e66) states that the sequence could include 3 BP readings up to 5 BP readings, taking the average of the last 2 readings.
The ESH 2021 guidelines also reference capturing triplicate and averaging readings 2 and 3.
Stergiou et al. 2021European Society of Hypertension practice guidelines for office and out-of-office blood pressure measurement J Hypertens 39:1293–1302 C
Regarding the rejection of the initial reading of the triplicate assessment, many studies do discard the initial reading based on a potential “White Coat” response (even with a 5-minute rest period) and average readings 2 and 3. This approach has been included in previous protocols.
This also gets into the discussion of “attended vs. unattended” BP assessment, about which there have been a number of publications.
We have seen a variety of inflation sequences, including:
- Take (4) readings, discard the 1st and average the last three (active study)
- Take (5) readings, discard readings 1, 2, 3 and average 4 and 5 (active study)
- Take (3) readings and generate the average of all three (active study)
- Take (3) readings, discard the 1st and generate the average based on readings 2 and 3 (active study)
Some of the considerations are linked to specific Automated Office BP devices and functionality to configure the inflation sequence and average calculation.
9 What impact does not having a placebo cohort have on sample size? Should a placebo cohort be included?
If the intent is to placebo adjust, then variability in both arms must be considered, and the impact on sample size is quite substantial, 119 vs. 234 subjects per arm, using a previous example. The long interval between baseline and on-treatment ABPM sessions may lead to a period effect, which would be lessened by having a placebo comparator. In addition, if the active treatment has a beneficial effect on BP, having a placebo group will also make this easier to demonstrate.
10 What protocol considerations should we be looking at?
One important point is reducing risk for missing data — selecting ABPM time points that allow for a session to be repeated if it fails. For example, schedule it at 8-10 weeks, if treatment duration is 12 weeks.
11 The FDA suggests that more than 50% of expected BP measurements constitutes a valid session; this seems somewhat low — any thoughts?
Discuss random distribution of missing points vs. missing multiple consecutive hours. The FDA has done simulation studies of 70% vs. 50% and has shown little impact on accuracy.
12 What about cuffless BP modalities? Can these be used to collect BP data?
They are, potentially, useful in earlier phase studies in an exploratory way; they are informative but may not meet regulatory expectations for submission-worthy data.
13 How do you suggest handling triplicate BP readings from home or clinic visits. Do we just take the average? Or do we average just the 2nd and 3rd?
- For Home BP, historically we have implemented a triplicate BP assessment at pre-defined AM and PM windows with an average of the AM and PM set of readings
- There have been protocols that just reference taking two readings at AM and PM windows
- Readings have not been excluded in the Home BP averaging as the assumption is that there is limited risk to “white coat hypertension” similar to ABPM and out-of-office assessment
For a clinic visit, this is an open discussion, and it depends on which scientific group and the interpretation of the position paper.
For example, the Mineralys team would not consider an alternate approach based on an AHA position paper interpretation, which ended up being configured from (5) measurements and an average reading of 4 and 5 (see attached article).
- Historically, what worked well in a large Idorsia study, was to capture (4) readings and discard the first reading when generating the average.
- As referenced in the guidance, they noted a triplicate and an average (this was used in the SPRINT Trial, which is often referenced).
- As you know, the team at KBP Biosciences decided on taking (3) measurements and discarding the 1st reading and averaging readings 2 and 3. I am not sure if there is any benefit to this approach.
14 When looking at the draft guidance, is there any consideration around BP assessment by age group, as there is a difference when measuring pediatric versus adults?
For the clinic BP assessment, the SPRINT Trial introduced the discussion of “attended vs. unattended” BP measurement; it does not look like this is addressed in the guidance. Is this something we need to address in our trials?
15 Why include warnings about a small BP increase when the drug itself may be very beneficial for the patient?
The adverse effects of a small BP increase may be overshadowed by the positive effects of the drug. However, a prescribing physician needs to have this information to adequately inform the patient about the risks and benefits of the medication and needs to know whether additional BP monitoring and treatment may be appropriate.
16 Why did testosterone get a black box warning for such a small BP increase?
This was probably related to the target population already having a high cardiovascular risk. As explained in the updated guidance, the absolute risk increase from a specific mean BP increase will depend heavily on the underlying risk of the patient population. In contrast, for patient populations with lower CV risk, the same magnitude of BP increase might just have the standard warning language.