Follow-Up of Cancer Patients Receiving Anti-PD-(L)1 Therapy Using an Electronic Patient-Reported Outcomes Tool (KISS): Prospective Feasibility Cohort Study

Introduction

Cancer patients experience a variety of symptoms derived from the malignancy itself as well as side effects of the given treatment. Many symptoms are left unnoticed due to factors such as limited symptom follow-up between prescheduled health care visits, nonsystematic evaluation of symptoms, and inadequate communication [1-7]. In general, worsening of symptoms indicates cancer progression or severe side effects of the treatment and is linked to poorer cancer survival [8].

Patient-reported outcomes (PROs) consist of health-related questionnaires completed by the patients themselves, which can capture symptoms and signs and their severity. Web-based reporting of PROs has many advantages compared to paper questionnaires such as reducing time to complete and overcoming geographic location limitations. Scheduled electronic patient-reported outcomes (ePROs) enable timely and continuous collection of symptoms in a cost-effective manner [9-14]. Furthermore, use of ePROs in cancer patient monitoring has shown impressive improvements in overall survival compared to standard follow-up [15,16]. In addition, ePROs can be coupled to an urgency algorithm, which sends an alert to the care unit upon report of severe or altering symptoms by a patient. This enables rapid reaction to and treatment of important medical events.

In the past 5 years, there has been significant advancement in the development of cancer immunotherapies with the introduction of immune checkpoint inhibitor (ICI) therapies such as anti-PD-(L)1 and anti-T-lymphocyte-associated protein 4 (CTLA-4) antibodies [17]. ICI therapies have become the most important medical therapies in many malignancies such as melanoma, non-small cell lung cancer, and urogenital cancers [18-27]. ICIs differ from traditional cancer therapies due to potentially severe side effects in all organs of the body and late timing of side effect occurrence [27-29]. Therefore, there is a need for comprehensive and ongoing assessment of symptoms.

Approximately 15% of patients receiving ICI monotherapies reportedly have severe grade 3-4 side effects, and about 30% have lower grade adverse events (AEs). Even life-threatening side effects can occur, but they can, in most cases, be managed with early detection, by delaying or stopping the ICI therapy, and with the initiation of immunosuppressive medication [30-32].

To our knowledge, this is the first prospective trial investigating ePROs in the follow-up of cancer patients receiving ICIs. The study aim was to investigate the feasibility of ePRO symptom follow-up and to analyze the spectrum of patient-reported symptoms, number and aetiology of urgency algorithm alerts, correlations between different symptoms and treatment benefit, and patient compliance.

Methods

Study Design and Participants

KISS was an investigator-initiated, multicenter, prospective, one-arm study, which was undertaken in 3 multidisease cancer centers in Finland. Patients were recruited during routine doctors’ appointments at study centers by study doctors. The inclusion criteria included advanced cancer to be treated with anti-PD-(L)1 in outpatient settings, initiation of anti-PD-(L)1 therapy had occurred ≤2 weeks prior to study recruitment, age ≥18 years, Eastern Cooperative Oncology Group ≤3, and availability of internet access and email. Baseline information such as basic laboratory values, age, and gender were collected from electronic health care records. After providing written informed consent, study patients received a short (5-15 minutes) instruction on how to use the Kaiku software by a study physician. At the initiation of the treatment phase (within 0-2 weeks from the first anti-PD-(L)1 infusion) and weekly thereafter until treatment discontinuation or 6 months of follow-up, patients received an email notification to complete the baseline electronic symptom questionnaire of 17 questions. If a weekly symptom questionnaire was not completed on the day of email receipt, daily email reminders were sent for 6 days. In addition, patients were asked to fill in a monthly electronic patient experience survey until treatment discontinuation or 6 months of follow-up. The use of the ePRO tool was free of charge for the patients and study centers. Online technical support by Kaiku Health for the users was available from 8 am to 4 pm Monday to Friday. The investigators evaluated the treatment response according to Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria at 8-10 weeks after treatment initiation. Clinical benefit rate was selected as a benefit measure instead of objective response rate since (1) we had a small number of study subjects, (2) responses were analyzed only up to 12 weeks from inclusion, and (3) the correlation between clinical benefit and objective response rate is not as clearly defined with immunotherapies as with traditional cancer medications.

According to the protocol, study results were analyzed when the last included patient had 12 weeks of follow-up. The major endpoints of the study included (1) patient-reported symptoms and their severity; (2) number of triggered alerts by the ePRO tool and their correlation to treatment side effects, cancer progression, other medical events, or survival; (3) correlations between different symptoms and treatment side effects, cancer progression, other medical events, or survival; and (4) patient compliance using the patient experience survey and response rates to symptom questionnaires. Sample size was based on the estimation that 15% of patients receiving ICI monotherapies will experience severe (grade 3-4) side effects and about 30% will experience lower grade AE. In a 40-patient cohort, 3-6 patients will experience a severe immune-related AE. It was estimated that the expected study population is sufficient to evaluate the feasibility of the symptom questionnaire in detecting severe AEs. Questionnaires from several timepoints were estimated to be collected from 90% of the study population (~35 patients), which would enable a more comprehensive assessment of feasibility, patient experience, and correlation of ePRO changes to treatment response and survival.

All data collection was carried out according to national legislation and under permit from the medical director of each research center. The study was approved by the Pohjois-Pohjanmaan sairaanhoitopiiri (PPSHP) ethics committee (number 9/2017), Valvira (number 361), and Oulu University Hospital Ethics Committee (9/2017). The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines.

ePRO Follow-Up

The Kaiku Health ePRO tool is a web-based solution scaled to be used easily on smartphones and home computers. The Kaiku Health immune-oncology module designed for the study consists of 17 questions. The symptoms selected for the Kaiku Health symptom tracking tool for cancer immunotherapy are based on the most common AEs that have occurred during clinical trials of anti-PD-1, anti-PD-L1, and anti-CTLA-4 monotherapies. The symptoms tracked by the instrument are potential signs and symptoms of immune-related AEs. The symptom selection was based on publications from the following clinical trials: CheckMate 017 (NCT01642004), CheckMate 026 (NCT02041533), CheckMate 057 (NCT01673867), CheckMate 066 (NCT01721772), CheckMate 067 (NCT01844505), KEYNOTE-010 (NCT01905657), and OAK (NCT02008227). Food and Drug Administration labels for nivolumab, pembrolizumab, and atezolizumab were also used in the symptom selection for the instrument. The questions for each symptom in the instrument were developed based on the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) register by converting the description of a grading into patient-friendly language. Any criterion that would be impossible for patients to report has been excluded from the available questions. Developing the symptom questionnaire in this manner enabled self-reporting by patients and development of an algorithm that provides an assessment and approximation of the severity of each symptom according to NCI-CTCAE criteria. NCI-CTCAE grades the symptoms from 0 to 4: no (0), mild (1), moderate (2), severe (3), and life-threatening (4).

Questions assess the presence of blood in stool, blood in urine, blurred vision, chest pain, cough, loss of appetite, diarrhea, dizziness, fatigue, fever, headache, itching, nausea, pain in joints, rash, shortness of breath, stomach pain, and vomiting. Besides recording the presence of a symptom, a severity algorithm that grades the symptom according to NCI-CTCAE was applied. The severity algorithm triggered an email alert to the study physician of the care unit based on preset limits (presence of a grade 3 or higher symptom or increase in symptom severity from grade 0 to 2). The patients were informed that the care unit would react to the alerts promptly within 3 days; thus, the ePRO follow-up was intended only for nonurgent communication, and in urgent matters, patients were advised to contact emergency care.

Patient Experience Survey

Study participants were requested to reply to a monthly patient experience survey. The patient experience survey consisted of 6 yes/no or multiple-choice questions. The survey was developed by the investigators for the study and has not been previously validated.

Statistical Analysis

The analysis was carried out when the last patient included had 12 weeks of follow-up data available. Correlations of different patient-reported symptoms were analyzed using heat maps with Pearson product-moment correlation. In the heat map analysis, the intensity of the color signifies the level of correlation: red, negative correlation; blue, positive correlation. In other words, a large effect correlation was defined as 0.5; medium as 0.3, and small as 0.1 (absolute values).

Results

Study Accrual and Patient Characteristics

Patient recruitment took place between June 2017 and March 2019, and the last study patient visit was in June 2019. Anticipated recruitment for the study was 40 patients in 12 months, but due to a slow recruiting pace, the period was extended. Informed consent was provided by 43 patients, and analysis was limited to a total of 37 patients who had anti-PD(L)-1 therapy initiated and answered at least 2 symptom questionnaires (baseline and one following; Figure 1). No technical issues nor security breaches related to the web-based tool occurred during the study period.

‎View this figure

The median age of the study participants was 62 years (range 32-80 years). The majority of patients were male (27/37, 73%), and 5 patients had a history of an autoimmune disease, with hypothyreosis (4/5, 80%) being the most common. Tumor types included lung cancer (15/37, 41%), melanoma (9/37, 24%), genitourinary (GU) cancer (9/37, 24%), and head and neck cancer (4/37, 11%), and 28 (28/37, 76%) patients had stage IV disease (Table 1).

Patient-Reported Symptoms and Alerts

During the study, 889 completed symptom questionnaires were registered. The range of answered questionnaires was 0.583-1.27 per patient per week, with high response rates throughout the complete follow-up period up to 24 weeks (Table 2).

During the first 12 weeks of ePRO follow-up, the most common grade 1-2 symptoms were fatigue (346/889, 39%), cough (187/889, 21%), pain in joints (160/889, 18%), itching (151/889, 17%), loss of appetite (151/889, 17%), nausea (151/889, 17%), and shortness of breath (133/889, 15%). The most common grade 3-4 symptoms were cough (53/889, 6%), loss of appetite (36/889, 4%), and nausea (36/889, 4%). None of the patients (0/37) reported blood in stool or hematuria (Table 3).

Of the 391 answered symptom questionnaires during the first 12 weeks, the ePRO tool triggered 67 (67/391, 17.1%) alerts. The most common reasons for alerts were loss of appetite, shortness of breath, pain in joints, blurred vision, and cough. The treating physicians were asked to evaluate the etiology of alerts by grading them to cancer, treatment, or unclear categories. Unclear reasons were the most common cause of alerts (38/67, 57%), followed by treatment (21/67, 31%) and cancer (8/67, 11%; Table 4).

Patient Compliance

Patient compliance was assessed every 4 weeks based on the electronic patient experience survey provided through Kaiku software. During the first 12 weeks, 31 patients replied to the survey, and analysis was limited to these. All the patients replied that using the Kaiku software was easy or very easy, and only 1 of 6 patients reported that they needed assistance using the software. Over 90% of the patients (29/31, 94%) reported that the questions were understandable. In addition, 90% of the patients (28/31, 90%) felt that the Kaiku ePRO follow-up improved their cancer care, and 95% (29/31) said they would recommend using it in the follow-up of cancer patients (Table 5).

Correlations Between Patient-Reported Symptoms and Treatment Benefit

Correlations between ePRO-collected symptoms were analyzed using heat maps. According to the results, the symptom correlations during the first 12 weeks and beyond were very similar (Figure 2). During the first 12 weeks, large positive correlations were seen between nausea, diarrhea, loss of appetite, and vomiting; stomach pain and decreased appetite; and rash and itching. Only small negative correlations were detected between cough and vomiting, itching and chest pain, and itching and fever (Figure 2A).

‎View this figure

Of the 37 patients, 34 were evaluated for objective treatment response (RECIST 1.1) by the investigators and included in the treatment benefit analysis. Of the 34 patients, 22 (65%) patients had complete response (CR), partial response (PR), or stable disease (SD) as the best response, while 12 (12/34, 35%) patients had progressive disease (PD). The heat map analysis suggested a small positive correlation between clinical benefit (CR/PR/SD) and itching (0.23 for the first 12 weeks, Figure 2A; 0.25 for all data, Figure 2B) and medium correlations between PD and chest pain (–0.41 for the first 12 weeks, Figure 2A; –0.47 for all data, Figure 2B). We further analyzed symptom progression and severity for itching and chest pain. During the first 12 weeks, 15-23% of the patients with clinical benefit reported itching, while the rate was much lower for patients with PD (0-14%; Figure 3). Furthermore, the average grade was much higher for patients with clinical benefit (weeks 1-12, 0.26-0.37; all 12 weeks, 1.18) compared to patients with PD (weeks 1-12, 0-0.17; all 12 weeks, 0.75; Table 6). For the complete follow-up period, most of the patients with clinical benefit had itching (14/22, 64%), while this was much lower for patients with PD (4/12, 33%; Figure 3). The severity of itching for the patients with clinical benefit was mainly low grade (grade 1: 6/22, 27%; grade 2: 4/22, 18%; Figure 3). During the complete follow-up period, chest pain was much more common in patients with PD (7/12, 58%) than in the patients with clinical benefit (4/22, 18%; Figure 4). In the first 12 weeks, patients with PD had a tendency for gradually increasing average grades for chest pain; conversely, a continuing decrease in the average grade was seen for patients who responded to the therapy (Table 6).

‎View this figure

‎View this figure

Discussion

According to previous studies, ePRO follow-up has improved survival and quality of life compared to routine surveillance when used with cancer patients receiving chemotherapy and lung cancer patients treated with curative intention [15,16]. However, the ePRO approach remains virtually unstudied in the context of cancer immunotherapies [33]. For optimal follow-up of patients receiving ICIs, there is a need for comprehensive assessment, grading, and long-term surveillance of symptoms. ePROs could provide a cost-effective follow-up tool to meet these 3 requirements. We previously reported a retrospective pilot study of ePRO follow-up of cancer patients treated with ICIs [34]. To our knowledge, this study is the first prospective clinical trial investigating ePRO follow-up of cancer patients treated with anti-PD-(L)1 therapies.

In this study, we used an ePRO module with 17 questions and an algorithm grading the PROs according to NCI-CTCAE. The questionnaire was designed specifically for patients receiving ICIs based on the published side effect profile of these agents. The symptom variety based on patient reporting and the grading algorithm performed well, and the symptom data followed closely what has been reported in clinical trials investigating ICIs. A recent meta-analysis with more than 20,000 patients suggested that fatigue (18%), itching (11%), and diarrhea (9%) are the most common AEs reported in patients treated with anti-PD-(L)1 agents [35]. The incidence of AEs in clinical trials are generally lower than in our study, which might be related to better capture of patient-reported low-grade symptoms, which are often overlooked in physician-based AE reporting in clinical trials [36-42].

In the present study, the symptom questionnaire was also coupled to an urgency algorithm, which generated alerts in 17% of the answered questionnaires during the first 12 weeks. Loss of appetite and fatigue were among the most common symptoms generating alerts. These symptoms very rarely alter the cancer treatment, and symptomatic treatments are scarce. Furthermore, physicians determined that most of the alerts were caused by unclear reasons, which is probably related to the high frequency of symptoms with unclear etiology. Fine-tuning of the alerts to focus not only on the symptom grade but also the nature of the symptom could lower the number of alerts and staff workload without sacrificing the performance of ePROs.

Patient adherence to and experience with ePRO follow-up was found to be very good in this study. The patients were requested by email to complete symptom questionnaires weekly, and the number of completed questionnaires was very close to one per patient per week for the first 12 weeks. Based on the patient experience surveys, the system was easy to use, and patients felt that ePRO follow-up improved their cancer care, which is in line with previous studies [40,41].

Our previous retrospective study with patients treated with ICIs suggested that some ePRO-reported dermatological, gastrointestinal (GI), and pulmonary symptoms co-occur [34]. Similarly, we saw large positive correlations between treatment response and GI symptoms as well as between treatment response and dermatological symptoms. Furthermore, the data showed small negative correlations between pulmonary symptoms and some GI symptoms and between itching, pulmonary symptoms, and fever. In our previous retrospective study, which did not include data on the treatment responses, we generated a hypothesis that GI and skin symptoms might be related to immune activation and treatment benefit, while pulmonary symptoms could signal tumor progression. Since this study also included data on treatment benefit, it enabled us to investigate our hypothesis. The results showed that there was a small positive correlation between treatment benefit (CR/PR/SD) and itching (ePRO) and between PD and chest pain (ePRO). Previous studies have linked autoimmune skin toxicity (rash) to PD-1 agent benefit [43-45]. Our results are hypothesis-generating while suggesting that ePRO-collected symptom data can mimic physician-assessed symptoms and correlate with treatment benefit. Furthermore, compared to physician-based AE reporting, it is possible that ePROs enable enhanced capturing of low-grade AEs without visible presentation such as itching and therefore facilitate predicting clinical treatment benefit.

ePROs enable cost-effective capture of symptoms and their change over long periods [46]. Changes over time might better predict treatment side effects and benefit than just a single presentation of a symptom. Furthermore, data from this study showed that early (in the first 12 weeks) changes in symptoms correlate with treatment benefit as well as symptoms from the whole follow-up period. This further highlights the possibility that early changes in symptoms predict outcomes. Large-scale symptom data coupled with treatment benefit and side effects could be used to build prediction models using artificial intelligence methods. These models could predict an individual’s risk for symptom development, treatment-related side effects, and treatment benefit.

Our study has some limitations. The sample size is small (n=37); however, the size is typical for feasibility studies. The small sample size prevents us from making strong generalizations based on the data. The one-arm design of the study precludes comparison of the effectiveness of the intervention. However, we feel that our study is important since it lays the groundwork for future studies on the topic.

In conclusion, this study is the first reported prospective clinical trial investigating the use of ePROs in the follow-up of cancer patients treated with ICIs. The results of this study suggest that follow-up of cancer patients using ePROs is feasible, enabling comprehensive capturing of symptoms over long periods with good patient adherence and satisfaction. Moreover, some early patient-reported symptoms were found to correlate with treatment benefit suggesting that individual prediction models for treatment benefit could be generated.