Understanding temporal relationships of patient experiences and their importance to patient safety

[#1]

In the development of new medical products, it is important to understand the temporal relationship of outcomes that arise for individual patients. Over the years, JMP Clinical has offered ways to understand the timing of medications, adverse events (AEs), and laboratory abnormalities through such reports as Patient Profiles and Adverse Events Narrative. These features give users the opportunity to understand a wealth of information from clinical trials, including:

• Which medications may have caused notable safety issues.
• Which treatments may have been used in response to the appearance of AEs.
• Which events or medications may have preceded a patient’s death.

These reports provide greater insight into the cause and effect of patient experience.

However, it can be challenging to subset to a particular set of clinically relevant outcomes and explore a user-defined time window around those outcomes across all patients. Further, summarizing the frequency of outcomes that took place within those windows is a manual process.

No longer!

The forthcoming Nearby Occurrences report in JMP Clinical 19 enables users to define reference occurrences of interest and find nearby occurrences that appear within time windows around start or stop dates, as well as those nearby occurrences that may have any overlap in time between reference and nearby occurrences. What is an occurrence? Per the CDISC documentation on Occurrence Data:

“Occurrence analysis is the counting of subjects with a given record or term and often includes a structured hierarchy of dictionary coding categories. Examples of data that fit into this structure include those used for typical analysis of adverse events, concomitant medications, and medical history.”

Nearby Occurrences currently considers occurrences such as adverse events (AE), concomitant medications (CM), and deaths.

How can this be used in practice? For example, the Nearby Occurrences report empowers users to:

1. Identify medications that were started on or within a few days subsequent to the start date of a specific AE.
2. Identify AEs or medications that started within a week prior to a patient’s death.
3. Identify any AEs that start within a few days of stopping particular medications.
4. Identify any AEs that start or are ongoing while particular medication classes are used.

The dialog for Nearby Occurrences is shown in Figure 1. Users can refer to AEs or medications according to their individual coded names as specified by MedDRA or WHODrug Global, or they can refer to terms according to their reported text or higher level classification (e.g., High Level Term or Original Category for Medication). These designations become the “Terms of Interest” for Nearby Occurrences. Users can opt to explore all AEs or medications or subset to terms that occur during specific periods of the study (i.e., Pre-Treatment, On Treatment, Off Treatment Follow-up, Treatment Emergent).

Figure 1. Nearby Occurrences dialog

Running the report for the Nicardipine example provides the following output in Figure 2.

Figure 2. Nicardipine example

At this point, the output assumes all occurrences are of interest as reference occurrences (which isn’t particularly useful). However, suppose we are interested in Serious Adverse Events (SAEs) of the Cardiac Disorders System Organ Class (SOC). Selecting these terms in the data filter provides an occurrence level and patient level summary of the terms that have been selected (Figure 3). In this example, all of our reference occurrences are AEs. From the left count plot at the occurrence level in Figure 3, we can ascertain the following:

1. There are 31 reference occurrences that are SAEs from the Cardiac Disorders SOC. Of these 31 occurrences, 55% and 45% of these events take place in patients in the NIC .15 and Placebo arms, respectively.
2. The most common occurrence is cardiac arrest with 12 occurrences, with 58% and 42% in NIC .15 and Placebo, respectively.
3. The least common Cardiac Disorders SAEs are atrioventricular block first degree, ventricular extrasystoles, and ventricular tachycardia with one occurrence each, which were all experienced in patients in the Placebo arm.

Figure 3. Reference occurrences: Cardiac Disorders SAEs

From the right count plot at the patient level in Figure 3, we can ascertain the following:

1. There were 24 patients that experienced Cardiac Disorders SAEs. Of these 24 patients, 58% and 42% are in the NIC .15 and Placebo arms, respectively.
2. The frequencies of patients for each occurrence type (n = 12, 3, 3, 3, 3, 2, 2, 1, 1, 1) do not change from the occurrence level count plot, indicating that each occurrence type occurs in a different patient. Had the frequencies been higher for an occurrence type in the occurrence level plot, that would indicate that patients would have experienced more than one instance of an occurrence.
3. The sum of the frequencies of the individual occurrences (n = 31) is greater than the number of patients (n = 24), indicating that some patients experienced multiple Cardiac Disorders SAEs.

The tabular summaries provided show the frequency and column percentage of each reference occurrence. For example, from the tables we can conclude that:

1. 41.2% (7/17) and 35.7% (5/14) of the Cardiac Disorders SAEs experienced in the NIC .15 and Placebo arms, respectively, were cardiac arrest.
2. 50.0% of the patients who experienced Cardiac Disorders SAEs in the NIC .15 (7/14) and Placebo (5/10) arms experienced a cardiac arrest.

The 31 reference occurrences are available for review by opening a Data outline box.

With the selection of these reference occurrences, the Query outline box of the Data Options panel enables the user to run a query to identify events (AEs or deaths) or medications that meet the criteria of the query (Figure 4).

Figure 4. Query dialog

Two types of queries are possible:

1. Specific Query: This query searches for nearby occurrences by comparing the start or end date of the reference occurrences to the start or end date of the nearby occurrences. If the nearby occurrences occur within the selected number of days After, Before, or Before or After the selected dates, they will be summarized in the Nearby Occurrences area of the output. Note that a window of zero days will identify occurrences that occur on the same date. For example, if the start date is used for both reference and nearby occurrences, Nearby Occurrences will identify terms that start on the same day as reference occurrences.
2. Any Overlap: This query searches for nearby occurrences by assessing whether there is any overlap in time based on the start and end dates of the reference and nearby occurrences.

In practice, missing dates for deaths and AEs are unlikely to occur. However, medications for chronic disease may have unknown start dates (indicating the medication was started at some point in the past) or unknown end dates (implying the drug is still being used to manage disease). While JMP Clinical makes every effort to impute dates according to predefined rules, in some instances missing dates may still occur. Nearby Occurrences makes the following assumptions for missing dates:

1. For specific queries, records can only be determined to be "near each other" when the dates used in the query are both non-missing. In other words, missing start dates are assumed to have occurred so far in the past that they could not be interpreted to be "nearby." Similarly, missing end dates are assumed to have not occurred, thus implying that the medication or event is still ongoing.
2. For queries of any overlap, a missing start date is assumed to be at the beginning of time, while a missing end date is assumed to be at the end of time. These assumptions make it possible to identify nearby occurrences with some overlap in time with reference occurrences when one or both dates are missing.

For instances where the stop dates of medications or AEs were not collected as part of the eCRF design, the stop dates are assumed to be equal to the start dates. In other words, the occurrences start and end on the same day. This assumption is always made for deaths.

Let’s run the following query to identify nearby occurrences that start within three days before the start of the reference occurrences (Figure 5).

Figure 5. Specific query

Clicking Run Query produces figures and tables in the Nearby Occurrences area of the output window. You can close the Reference Occurrences outline box to easily view this output. Recall that the Nicardipine trial has daily visits for approximately two weeks for a very severe disease. The initial output is extremely crowded, though when we subset to CM using the Data Filter, we get the following results in Figure 6.

Based on the results, the following conclusions are true:

1. Occurrence level
   1. There are 192 CM occurrences that started within three days of the start date of at least one Cardiac Disorders SAE. Of these 192 CM occurrences, 54% and 46% of these CM occurrences take place in patients in the NIC .15 and Placebo arms, respectively.
   2. The most common nearby occurrence is dexamethasone with 22 occurrences, 50% in both NIC .15 and Placebo.
   3. Potassium supplements has 21 occurrences.
   4. Twenty-one out of 192 (10.9)% of all nearby CM occurrences were potassium supplements.
2. Patient level
   1. Twenty-one of 24 patients have nearby CM occurrences for least one Cardiac Disorders SAE.
   2. Twelve patients took phenobarbital. Given there are 18 instances of phenobarbital at the occurrence level, some patients took it more than once in those three days). Eleven patients took potassium supplements. Given there are 21 instances at the occurrence level, some patients took it more than once in those three days).
   3. Twelve out of 21 (57.1%) patients had phenobarbital, two-thirds of whom were on NIC .15.

Figure 6. Nearby Occurrences

Here, the Data window may be of more interest to the user to see the specific details for nearby occurrences and the particular reference event(s) to which they belong. Note that the figures and tables in Figure 4 display unique nearby occurrences. However, the Data window shows all of the nearby occurrences that apply to each reference occurrence, even if the nearby occurrences are presented multiple times.

The report includes reference occurrences in the output even when no nearby occurrences are present; they are listed as having “No Nearby Occurrences.”

Nearby Occurrences is also available for crossover trials or trials where treatments vary according to study periods. Patient level summaries are based on the planned or actual arm. Occurrence level summaries are based on occurrence level treatment values.

This upcoming feature makes it much easier to identify patients of interest according to the events they experience.