I’ll start with a little bit of background, I think it will really help set us up to explain the results of the study and what we did. Today when oncologists want to make decisions regarding cancer treatments they often use the concept of performance status, which basically tells us how functional or active the patient is, to try and predict how the patient will do with treatment – are they going to have complications or are they going to end up in the emergency room? Are they going to have a lot of side effects? Is the treatment going to be helpful or not? Typically, patients who are very frail and, sick and weak tend not do very well and have a lot of complications. The concept of performance status is very effective, it is predictive pretty much of all the results we look into in cancer treatment, whether it’s good outcomes or bad outcomes. It’s cheap and easy to use but there are more and more data that it is biased, it’s really affected by the relationship between the patient and the provider. Also, providers tend to have an assessment of performance status that is not necessarily up to date. They might think of their patient and how they were a few months ago and not realise that today their function is really different. 

Because of this issue a lot of different approaches have been brought up in recent years to try and look at different ways to not only assess performance status but also to predict complications and address events in cancer patients getting treatment. So you of course know a lot of work has been done in patient-reported outcomes where in those questionnaires, cancer patients would say how they feel every week or at any interval and that would be used to predict complications and other outcomes in cancer care. That has been really successful but it takes a lot of toll from the patient and from the provider, looking at all that data. Other groups have started looking at smart watches – can we track patients’ movements and activity, to get a sense of their performance status and maybe use that digital information to predict complications in cancer patients? 

Our group, in this study, we have partnered with the Division of Research at Kaiser Permanente, together with a local tech company called Medable. Medable is a company that develops digital platforms for clinical trials in oncology. Together we wanted to explore whether or not reports by caregivers about how the patient is doing can also be used as predictors of clinical outcomes. So in our study we recruited 54 dyads of patients and their caregivers. We recruited them over a few months at the end of 2020 and we gave each patient and each caregiver an Apple watch for the duration of the study, which was about a month, and also a smartphone app. We collected the physical activity using the smartwatch as well as we collected regular surveys from the patients about how they were doing, collecting patient reported outcomes, self-rated ECOGs and other questionnaires, and from the caregivers. Using those observer-reported outcomes or caregiver-reported outcomes, we tried to see if what the caregiver thinks about the patient, how they are doing, is actually predictive of the patient’s outcomes.

As a background to presenting the findings I do also want to  mention that last year we presented at ASCO Quality Symposium a survey that we ran through with all of our oncologists at Kaiser Northern California and we had 38 respondents. The survey had to do with how often and in what ways do oncologists use information from caregivers. We found out that all of our oncologists often rely on caregiver reports on how the patient is doing to evaluate the patient’s performance status and symptoms and also to make decisions. What was really interesting in our findings is that, as a lot of people know, often the patient would say one thing and the caregiver would say another thing. The patient would say, ‘I’m doing great,’ and the caregiver would be like, ‘Eh, you know, you remember last week you couldn’t get out of bed?’ or something to that effect. We asked oncologists, ‘What do you do in that scenario?’ and most oncologists either rely on what the caregiver is saying more than what the patient is saying or they look for some kind of objective data, maybe like a smartwatch or other objective collected data to make a determination.

So, coming to our study now, we presented the first findings of our clinical study with the patients and their caregivers. We collected from the caregivers two main types of reports. The first one was what we coined caregiver-reported outcomes, basically looking at asking the caregiver about the patient’s symptoms. Once a week the caregiver would go on the app and rate the patient’s symptoms using a scale called the PRO-CTCAE. The caregiver would say, for example, ‘Over the last week the patient had severe, very severe, nausea or moderate nausea or mild nausea or no nausea. Then going over other symptoms as well such as fatigue, depression and others. So that was one set of answers that we collected and the other one was asking the caregiver how is the patient doing physically, how active are they? For that we used a scale called NIH-PROMIS which is a collection of questionnaires and specifically a subset of those questions had to do with physical function. Then we correlated the answers of the caregivers about the patient’s function, the patient’s symptoms, with very important outcomes that we look for in oncology. We associated them with the outcomes of mortality, hospice referrals, emergency room visits, hospitalisations, treatment breaks and regimen changes, things that are very important for clinicians. 

We had a pretty small study, this was a pilot study in preparation for hopefully a larger study in the future, so we didn’t have quite a lot of events in terms of those outcomes that I mentioned – ED visits, hospitalisations, mortality etc.- so that was the limitation of our study. But we did find out that the reports from the caregiver on the patient’s symptoms were predictive of a lot of the outcomes that we were looking at. So, for example, the count of how many severe or very severe symptoms the patient had was a statistically significant predictor of hospitalisations and ED visits. There was also a trend towards statistical significance to predict high grade adverse events of chemotherapy and dose reductions as well as a trend towards predicting death. When we looked at specifically the PROMIS questionnaires that correlated the patient’s function and their outcomes we also found that this would predict hospice referrals.

Now, these are very interesting findings. This is one of the first publications to show that if you ask the caregiver how the patient is doing that’s very relevant information because it can help you predict how the patient will do in the future. The next part of our study, the next part of our publication, would focus on trying to put everything together, putting all the different predictors we collected – the patient-reported outcomes, the caregiver-reported outcomes, the mobility data from the smartwatch and what our nurses collected as well, which is the nurse-rated ECOGs and information on symptoms – trying to put all that together into one predictive model to develop an algorithm that can be very accurate in predicting complications in real time, and hopefully developing a clinical trial where we use this prediction model to identify those cases that are going to have a complication and try to intervene and prevent that complication. That would be the holy grail of what we’re trying to achieve eventually. 

How will the trial results impact patients?

Personally for me, and I think for the rest of our group as well, we feel more confidently now to make sure we also collect information from the caregiver when we see patients in the clinic, knowing that adding that question to the caregiver can improve how we predict those clinical outcomes in our cancer patients. In the future I hope that once we develop this predictive model we’ll be able to have routine collection of different aspects or different data from the patients and their caregivers such as their mobility and the caregiver-reported outcomes and the patient-reported outcomes and have a workflow where if there are certain situations that predict an impending complication that we would have some kind of intervention, depending on the situation. Either if it’s something that is symptom-related might be a nurse or a pharmacist that would intervene; if it’s something other it might be the oncologist that would intervene.