Patient-reported outcomes from the TBCRC 022 study of neratinib and ado-trastuzumab emtansine for HER2-positive breast cancer brain metastases

Abstract

Purpose

In TBCRC 022 (NCT01494662), neratinib and ado-trastuzumab emtansine (T-DM1) demonstrated intracranial activity among patients with HER2-positive breast cancer brain metastases. However, gastrointestinal (GI) toxicities—particularly diarrhea—were common, potentially impacting quality of life. Clinician-reported adverse event (CTCAE) grading may underestimate the patient experience. We report GI toxicities using patient-reported outcomes (PROs) in TBCRC's neratinib-T-DM1 cohort.

Methods

Patients received neratinib (160 mg daily) and T-DM1 (3.6 mg/kg IV every 21 days). A pre-planned analysis assessed patient-reported GI toxicities during Cycles 1–4 using the Patient-reported Outcomes Measurement Information System (PROMIS) GI Diarrhea scale, Systemic Therapy-Induced Diarrhea Assessment Tool (STIDAT), and PRO-CTCAE. Descriptive statistics and linear mixed effects models evaluated symptom trajectories over time. We evaluated agreement for PRO and clinician-reported data.

Results

Forty-four patients enrolled; all completed ≥1 PRO. GI symptom burden increased over Cycles 1-3. PROMIS scores worsened significantly by Cycle 2, with a peak mean increase of 6.62 (95% confidence interval (CI): 2.67–10.59; p = 0.002) at Cycle 3. STIDAT scores also worsened by Cycle 3 (mean change 0.50; 95%CI: 0.11–0.90; p = 0.015). Based on maximum PRO-CTCAE scores, moderate-to-severe symptoms were reported by 20.5% (diarrhea), 24.4% (appetite loss), and 41.0% (constipation). Agreement between PRO-CTCAE and clinician-reported CTCAE was low (kappa <0.2) with clinicians reporting less toxicity. PROMIS and STIDAT scores were significantly correlated.

Conclusion

This GI-focused PRO analysis highlights the value of PROs in capturing patient-experienced toxicities that impact quality of life yet are underestimated by clinicians. Incorporating PROs into clinical trials can inform supportive care and prophylactic strategies.

Highlights

• •First GI-focused PRO analysis of neratinib plus T-DM1 in HER2+ brain metastases.
• •Patient-reported diarrhea, appetite loss, and constipation peaked by cycle 3.
• •PROMIS and STIDAT tools were highly correlated for symptom reporting.
• •Clinician-reported CTCAE underestimated patient-reported GI toxicities.
• •Discordance in symptom reporting, not severity, drives low CTCAE–PRO agreement.

1. Introduction

Approximately 15–20% of people with metastatic breast cancer (MBC) have HER2-positive disease [1]. Over time, patient survival has significantly improved with the advent of HER2-targeted therapies, including monoclonal antibodies—often conjugated with chemotherapy—and tyrosine kinase inhibitors [2]. However, due to the distinct biology of HER2-positive tumors and the extended course of metastatic disease, brain metastases develop in up to 50% of those with MBC over time [3]. Central nervous system (CNS) involvement contributes substantially to morbidity and mortality, representing a critical clinical need [4] [5].

The Translational Breast Cancer Research Consortium (TBCRC) 022 trial (NCT01494662) was a multicenter, multi-cohort, phase 2 study investigating neratinib-based treatments for both pretreated and untreated HER2-positive breast cancer brain metastases (BCBM) [[6], [7], [8]]. Cohort 4 assessed the intracranial efficacy of neratinib (160 mg daily) in combination with ado-trastuzumab emtansine (T-DM1; 3.6 mg/kg intravenously every 21 days) across three cohorts: 4A – patients with previously untreated brain metastases; 4B – patients with progression after local CNS-directed therapy and no prior T-DM1 exposure; and 4C – patients with progression after local CNS therapy and prior T-DM1 treatment.

The primary endpoint for the Cohort 4 analysis was CNS objective response according to Response Assessment in Neuro-Oncology–Brain Metastases (RANO-BM) criteria, evaluated for each cohort. As previously reported, the CNS objective response rate (ORR) was 33.3% [95% confidence interval (CI), 4.3%–77.7%] in cohort 4A, 35.3% (95% CI, 14.2%–61.7%) in cohort 4B, and 28.6% (95% CI, 11.3%–52.2%) in cohort 4C. As a result of these data, neratinib-based regimens have been added to the National Comprehensive Cancer Network Guidelines for treatment of CNS disease [9]. In Cohort 4 of TBCRC 022, using traditional Common Terminology Criteria for Adverse Events (CTCAE) V4.0, the most frequently reported adverse event (AE) was diarrhea (despite diarrhea prophylaxis with loperamide and colestipol during cycle 1), occurring as a maximum of grade 2 in 31.8% and grade 3 in 22.7% of patients; no grade 4 diarrhea was observed. Fatigue, primarily grade 2, was reported by 27.3% of patients [8,10]. With neratinib administration, diarrhea has been typically reported as an early symptom and often becomes less severe and more manageable after the initial treatment cycles. Prophylactic loperamide and dose-escalation strategies, as evaluated in the CONTROL trial [11], are routinely used in practice and were incorporated into the Cohort 4 study design.

Patient-reported outcomes (PROs) are increasingly used in oncology to capture symptoms and treatment-related toxicities from the patient perspective, complementing clinician-reported AEs [12,13]. Several comparative studies have consistently shown that clinician-reported CTCAE grades underestimate patient symptom burden. Concordance with PROs is typically only moderate, with underreporting most evident for subjective toxicities such as fatigue, pain, gastrointestinal (GI) symptoms, and appetite loss [14]. Patients frequently describe higher severity and interference than clinicians record, particularly for grade ≥2 toxicities [15], highlighting that clinician assessments alone may not capture the full extent of treatment-related adverse events.

Given the wide recognition that clinician-reported symptoms do not fully capture the patient experience [12,13], in a pre-planned analysis for patients on Cohort 4 of TBCRC 022, we prospectively captured additional AE details with a focus on GI toxicity, using patient reported symptoms. Herein, we report the results of the PRO analysis for patients enrolled to Cohort 4 who were treated with neratinib in combination with T-DM1. To fully assess GI PROs beyond grade, we incorporated three validated instruments with the goal to describe the comprehensive experience with regard to frequency, severity and impact of diarrhea on one's quality of life.

2. Materials and methods

TBCRC 022 trial (NCT01494662) was a multicenter, phase II, multi-cohort, nonrandomized study enrolling patients with HER2-positive breast cancer brain metastases (BCBM) across three sub-cohorts. All participants provided informed consent, and the trial adhered to the Declaration of Helsinki [8].

Eligible patients had measurable CNS disease (≥10 mm parenchymal brain metastasis), Eastern Cooperative Oncology Group (ECOG) performance status 0–2, adequate end organ function, and no recent seizure activity or steroid escalation. Cohort 4A included patients with no prior CNS-directed therapy, while cohorts 4B and 4C enrolled patients with CNS progression after such therapy—without prior T-DM1 exposure (4B) or with prior T-DM1 exposure (4C). Exclusion criteria included unmeasurable CNS disease (e.g., leptomeningeal disease only), significant chronic diarrhea, or active hepatitis [8].

2.1. Treatment plan

All patients across Cohort 4 received neratinib (160 mg orally, daily) and T-DM1 (3.6 mg/kg IV, every 21 days), per the National Surgical Adjuvant Breast and Bowel Project (NSABP) FB-10 regimen [16]. Mandated diarrhea prophylaxis was provided to participants and included loperamide 4 mg every 8 h for days 1-14 followed by 4 mg every 12 h during days 15-21, and colestipol 2 g twice daily, but clinicians could stop treatment earlier per their discretion (e.g., if patient experienced constipation) or continue past cycle 1 if desired. Treatment diaries were provided but not required for prophylactic medications. Treatment with neratinib and T-DM1 continued until progression, unacceptable toxicity, or patient/clinician decision. Clinical evaluations, including CTCAE evaluation, occurred every 21 days, with brain and body imaging at 6 weeks once, then every 9 weeks.

2.2. PROs and collection

The Patient-Reported Outcomes Measurement Information System (PROMIS) GI Symptom Scale (Supplementary Data 1) consists of 5 items assessing frequency, severity, and impact of diarrhea on daily activities, each rated on a 5-point Likert scale (from “never” to “always” or “not at all” to “very much”) [17,18]. The Systemic Therapy-Induced Diarrhea Assessment Tool (STIDAT; Supplementary Data 2) is a 13-item instrument covering dimensions such as the patient's perception of diarrhea, stool frequency, fecal incontinence, and associated abdominal symptoms, with items rated on 5-point Likert scales or numeric frequency counts [19]. The PRO-CTCAE module used in this study comprised 20 GI symptom questions, each scored with 5-point Likert-like response options for frequency, severity, and/or interference (Supplementary Data 3). Patients were asked about decreased appetite, nausea, vomiting, heartburn, flatulence, abdominal bloating, constipation, loose/watery stools/diarrhea (frequency), abdominal pain, loss of control of bowel movements, hiccups [20]. Because diarrhea is an early AE observed with neratinib-based treatments, surveys were administered at study visits for Day 1 of Cycles 1–4 (every 3 weeks), if protocol treatment was still ongoing, including the PRO-CTCAE, PROMIS, and STIDAT questionnaires. Survey timepoints occurred on the same day as clinician-reported CTCAE assessments.

Higher PROMIS and STIDAT scores indicate greater GI symptom burden. PROMIS scores use T-scores, where the mean is 50 and the standard deviation (SD) is 10. Based on HealthMeasures guidance, a change of approximately 5–7 T-score points is generally considered clinically meaningful, corresponding approximately to 0.5–0.7 SD. The STIDAT scoring system uses a weighted algorithm based on several items, with a cutoff score of 1.35 used to predict the presence of clinically significant diarrhea. [17,19]. PRO-CTCAE items were scored according to National Cancer Institute (NCI) guidelines, using 5-point ordinal response scales; each PRO-CTCAE symptom composite score has range 0-3 [21]. For scoring, the highest reported value for each item dimension was retained [20].

2.3. Missingness of PROs at each time point

See Supplementary Data 4 and 5 for missingness summaries for the PRO-CTCAE score, STIDAT score, and PROMIS GI Diarrhea score at each time point and how these participants’ data were analyzed. If a subject completed at least one of these scores at any time point within cycles 1-4, we were able to calculate the maximum score across cycles 1-4 for this subject.

2.4. Clinical CTCAE

Clinical CTCAE were assessed by clinical and research staff on Day 1 of every cycle using the NCI CTCAE, CTEP Version 4.0. See Supplementary Data 4 and 5 for details on how clinical CTCAE was reported and the missingness summaries for clinical CTCAE toxicities during cycles 1-4.

2.5. Objectives

The primary objective of this analysis was to evaluate patient-reported GI toxicity (PROMIS GI Diarrhea and STIDAT scores; PRO-CTCAE) in Cohort 4 patients receiving T-DM1 and neratinib during cycles 1 to 4. We also aimed to compare patient-reported vs. clinician-reported GI adverse events during the initial four cycles of therapy.

2.6. Statistical analysis

We reported PROs and GI CTCAE outcomes for each cycle separately and as the maximum reported value across all time points for each subject. For these analyses, data from cohorts 4A, 4B, and 4C were combined. These cycle-specific and maximum PRO and clinical CTCAE outcomes were then summarized using descriptive statistics, including median and range for continuous scores and frequencies for categorical symptoms. Kaplan-Meier methods were used to evaluate the time from the date of first T-DM1-neratinib treatment to the first grade 3 PRO-CTCAE event, censoring patients at their day 1 of cycle 4 if no grade 3 was reported.

The changes in the PROMIS and STIDAT scores over time were evaluated using linear mixed effects models, where time (cycle number) was treated as a categorical variable and a random intercept for each subject was included to account for repeated measures. This model reported the average change in the PROMIS and STIDAT scores between cycles 2-4 and cycle 1 while accounting for repeated measures.

The relationship between the PROMIS and STIDAT scores using all available time point measures was evaluated using a linear mixed effects model, where a random intercept for each subject was included to account for repeated measures. The relationship between the max PROMIS and max STIDAT scores was evaluated using a linear model.

In a final analysis, the agreement between the maximum PRO-CTCAE and clinical CTCAE symptoms—each categorized as 0-1, 2, or 3 (higher number = more severe)—were evaluated using kappa statistics. To further evaluate agreement between maximum PRO-CTCAE and clinician-reported CTCAE, we performed a complementary analysis. The goals of this complementary analysis were to evaluate the agreement between reporting the symptom vs. not reporting the symptom while ignoring the level of severity of the reported symptom. We collapsed the maximum PRO-CTCAE and CTCAE scores into the categories of 0-1 (“No: Symptom not reported”) vs. 2-3 (“Yes: Symptom reported”) and evaluated the agreement between these categories for each symptom using kappa statistics.

3. Results

3.1. Patient characteristics and treatment

As previously reported, between November 7, 2018, and November 1, 2021, TBCRC 022 enrolled in total 44 patients to cohort 4 (6, 17, and 21 patients in cohorts 4A, 4B, and 4C, respectively) [8]. Participant characteristics are summarized in Table 1. Additional details on patient characteristics by cohort have been previously published [8]. The median number of cycles of neratinib and T-DM1 received on protocol were 6 (range 1-67) for each treatment. Across Cohorts 4A, 4B, and 4C, 5 of 44 patients (11.4%) discontinued treatment due to unacceptable toxicity. The most frequently reported grade ≥3 toxicities in these 5 patients were diarrhea (3 patients, 60%), increased aspartate aminotransferase (2 patients, 40%), and increased alanine aminotransferase (1 patient, 20%).

Table 1.

Patient characteristics for the 44 patients with patient-reported outcome (PRO) data.

Characteristic	Overall (N = 44)
Age, years
Median	48.5
Range	34.0 - 67.0
Sex (Female)	43 (97.7%)
Race
White	38 (86.4%)
African American or Black	2 (4.5%)
Asian	1 (2.3%)
Other	3 (6.8%)
Ethnicity
Hispanic or Latino	1 (2.3%)
Non-Hispanic	39 (88.6%)
Unknown	4 (9.1%)
CNS Parenchymal (Yes)	44 (100.0%)
CNS Leptomeningeal (Yes)	3 (6.8%)
Lung/Pleural (Yes)	10 (22.7%)
Breast or Chest Wall (Yes)	8 (18.2%)
Lymph Node (Yes)	4 (9.1%)
Liver (Yes)	13 (29.5%)
Bone (Yes)	27 (61.4%)
Soft Tissue (Yes)	3 (6.8%)
Number Non-CNS Metastatic Sites
Median	1.5
Range	0.0 - 6.0
ER Status
Negative	19 (43.2%)
Positive	20 (45.5%)
(Missing)	5 (11.4%)
Number Prior Lines of Chemotherapy
Median	2.0
Range	0.0 - 10.0
Prior surgery for CNS tumors (Yes)	14 (31.8%)
Prior SRS (stereotactic radiosurgery) (Yes)	23 (52.3%)
Prior WBRT (whole brain radiotherapy) (Yes)	23 (52.3%)
Number Neratinib Cycles
Median	6.0
Range	1.0 - 67.0
Number T-DM1 Cycles
Median	6.0
Range	1.0 - 67.0

Abbreviations: CNS, central nervous system; ER, estrogen receptor; T-DM1, ado-trastuzumab emtansine.

3.2. Patient-reported GI toxicity

Overall, all patients in Cohort 4 completed at least one PROMIS Diarrhea, STIDAT, or PRO-CTCAE assessment in one of the cycles 1-4. Completion rates for each instrument declined over time. For PROMIS scores, 39/44 (88.6%) patients completed surveys at Cycle 1, 31/44 (70.5%) Cycle 2, 20/44 (45.5%) Cycle 3, and 22/44 (50.0%) Cycle 4. For the STIDAT, completion was 41/44 (93.2%) at Cycle 1, 27/44 (61.4%) at Cycle 2, 20/44 (45.5%) at Cycle 3, and 20/44 (45.5%) at Cycle 4. For PRO-CTCAE, 30/44 (68.2%) provided a full symptom set at Cycle 1, 19/44 (43.2%) at Cycle 2, 12/44 (27.3%) at Cycle 3, and 10/44 (22.7%) at Cycle 4.

Patients frequently reported GI symptoms related to treatment with neratinib and T-DM1. The evolution of STIDAT scores and PROMIS Diarrhea scores is shown in Fig. 1. Over time (and compared to baseline scores prior to the first cycle), there was a statistically significant worsening in both measures. For theSTIDAT score, a statistically significant worsening compared with Cycle 1 was observed starting at Cycle 3, with an average difference of 0.50 ([95% CI: 0.11–0.90]; p = 0.015), Fig. 1A. A statistically significant increase compared to Cycle 1 in the PROMIS Diarrhea score was observed as early as the start of Cycle 2, with a maximum average difference at the start of Cycle 3 of 6.62 ([95% confidence interval (CI): 2.67–10.59]; p = 0.002) (Fig. 1B).

Fig. 1.

Box plots of STIDAT and PROMIS scores across cycles

Box plots of changes in patient-reported outcomes using STIDAT (A) and the PROMIS GI Diarrhea (B), among 44 patients treated with neratinib plus T-DM1 per treatment cycle (Cycles 1–4).

Abbreviations: STIDAT, Systemic Therapy-Induced Diarrhea Assessment Tool; PROMIS, Patient-reported Outcomes Measurement Information System; GI, gastrointestinal; T-DM1, ado-trastuzumab emtansine.

PRO-CTCAE assessments over Cycles 1–4 are displayed in Fig. 2. Of those completing diarrhea assessments at each time point, 40.0–66.7% reported grade 1 (mild) symptoms and 5.0–16.7% reported grade 2 (moderate) symptoms. Diarrhea at cycle 3 was among the most frequently reported symptoms. Specifically, 5 patients (41.7%) reported grade 1, 2 patients (16.7%) reported grade 2, and 1 patient (8.3%) reported grade 3 diarrhea. However, based on the maximum reported diarrhea PRO-CTCAE score across cycles 1-4, 20.5% of patients reported moderate to severe diarrhea, including 6 patients (15.4%) with grade 2 and 2 patients (5.1%) with grade 3 diarrhea; diarrhea PRO-CTCAE data across all cycles 1-4 were missing for 5 patients.

Fig. 2.

Bar plots of PRO-CTCAE scores across cycles

Patient-reported gastrointestinal symptoms over treatment cycles using PRO-CTCAE scores among 44 patients receiving neratinib plus T-DM1. Each panel shows the proportion of patients reporting symptoms at each severity level per treatment cycle (Cycles 1–4) for different GI symptoms: abdominal pain, heartburn, bloating, appetite loss, nausea, constipation, diarrhea, and hiccups. Symptom scores are defined as follows: 0 = none, 1 = mild, 2 = moderate, 3 = severe. Bar heights represent the percentage of patients (out of 44) reporting each level of symptom severity during the respective cycle.

Abbreviations: PRO-CTCAE, Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events; T-DM1, ado-trastuzumab emtansine; gastrointestinal, GI.

With regard to other relevant symptoms, appetite loss was also commonly reported, with 9 of 12 patients (75.0%) experiencing mild to severe symptoms at Cycle 3. In addition, constipation was frequently reported, with 8 of 12 (66.7%) of patients experiencing mild to severe symptoms at Cycle 3 (and diarrhea prophylaxis was not required after Cycle 1). Specifically, 2 patients (16.7%) reported grade 3 (severe) constipation at cycle 3. Based on the maximum reported PRO-CTCAE assessments across cycles 1-4, 41.0% of patients experienced moderate to severe constipation, including 5 patients (12.8%) with grade 2 and 11 patients (28.2%) with grade 3 symptoms; constipation PRO-CTCAE data across all cycles 1-4 were missing for 5 patients. Bar plots of PRO-CTCAE scores by cycle and bar plots of the maximum reported PRO-CTCAE scores across cycles 1-4 are presented in Fig. 2andFig. 3, respectively. Of note, although diarrhea and constipation were frequently experienced by patients, reporting both grade ≥2 diarrhea and constipation occurred <5 patients overall during cycles 1-4.

Fig. 3.

Bar plot of maximum PRO-CTCAE scores

Barplot of the maximum patient-reported PRO-CTCAE symptom —defined as the worst score reported across all time points—for the 10 gastrointestinal-related symptoms among 44 patients treated with neratinib plus T-DM1 in Cohort 4 of the TBCRC 022 trial. Symptom severity is represented by the highest score reported across treatment cycles (1–4) for each patient. Scores are defined as: 0 = none, 1 = mild, 2 = moderate, 3 = severe. Bar heights represent the percentage of patients (out of 44) reporting each level of symptom severity during the respective cycle.

Abbreviations: PRO-CTCAE, Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events; T-DM1, ado-trastuzumab emtansine; TBCRC, Translational Breast Cancer Research Consortium.

The median time from cycle 1 day 1 to the onset of any grade 3 PRO-CTCAE symptom was 6 weeks (95% CI 3 weeks–not reached), corresponding to the start of Cycle 3 (3 weeks per cycle). The 9-week survival probability (corresponding to the start of Cycle 4) without grade 3 PRO-CTCAE symptoms was 39% (95% CI 22%–68%) (Fig. 4). Among the five patients who discontinued treatment due to unacceptable toxicity (Supplementary File 6), most (60%) stopped after just one cycle. The median PROMIS GI Diarrhea T-score among these patients was 54.3 (range: 39.9–70.8), and the median STIDAT score was 1.9 (range: 0.6–3.6). Severe PRO-CTCAE symptoms (grade 3) were reported in one patient for nausea, loss of appetite and constipation.

Fig. 4.

PRO-CTCAE Time-to-Event analyses- Time from first treatment to first PRO-CTCAE score 3 (any symptom)

Kaplan-Meier curve depicting the time from initiation of treatment to the first patient-reported grade 3 PRO-CTCAE score for any gastrointestinal symptom among 41 patients treated with neratinib plus T-DM1 in Cohort 4 of the TBCRC 022 trial.

Abbreviations: PRO-CTCAE, Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events; T-DM1, ado-trastuzumab emtansine; TBCRC, Translational Breast Cancer Research Consortium.

3.3. Association between PRO-CTCAE and CTCAE scores

Agreement between PRO-CTCAE and clinician-reported CTCAE scores was weak, with kappa statistics consistently <0.2 (slight agreement at best) across symptoms (Supplementary 7). The highest concordance estimate was for abdominal pain (κ = 0.19) and the lowest was for vomiting (κ = −0.19) (Table 2). Using two-level categories (Supplementary Data File 8) (No: symptom not reported = 0-1, Yes: symptom reported = 2-3), agreement between PRO-CTCAE and clinical CTCAE scores remained weak across most symptoms, consistent with the 3-level analyses. The only exception was constipation, which shifted from slight to fair agreement (κ = 0.33).

Table 2.

Kappa statistics evaluating agreement between PRO-CTCAE and clinical CTCAE symptoms.

Symptom	Kappa Estimate	95% CI
Diarrhea	0.09	(-0.11, 0.29)
Nausea	0.08	(-0.18, 0.33)
Vomiting	−0.19	(-0.29, −0.09)
Abdominal Pain	0.19	(-0.04, 0.42)
Constipation	0.18	(-0.01, 0.36)
Bloating	0.15	(-0.03, 0.32)

Abbreviations: PRO-CTCAE, Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events.

3.4. Correlation between PROMIS diarrhea and STIDAT scores

A significant correlation was observed between PROMIS Diarrhea and STIDAT scores. In a linear mixed-effects model using all available time points and a linear model comparing maximum scores, each one-unit increase in the STIDAT score was associated with a 7.09-unit increase (95% CI 5.59–8.61; p < 0.001) and a 6.75-unit increase (95% CI 4.47–9.02; p < 0.001) in the PROMIS Diarrhea score, respectively.

4. Discussion

In this study, we report the results of the GI PRO analysis using three different scoring tools during the first 4 cycles of treatment with neratinib and T-DM1 in TBCRC 022. With a high PRO completion rate by study participants, we observed a significant increase compared to baseline in GI symptoms across all three scoring systems (PROMIS, STIDAT, and PRO-CTCAE), primarily observed in the first two treatment cycles, despite use of prophylactic anti-diarrhea medication. The most common GI symptoms were diarrhea, appetite loss, and bloating. We observed only limited concordance between PRO-CTCAE and clinician-reported CTCAE grades; however, there was a significant correlation between PROMIS Diarrhea and STIDAT scores. It is of note that these scales do differ somewhat in content outside of what was analyzed, such as the detailed questions included on the STIDAT regarding the impact of bowel symptoms on mood, family life and social life. Similar to other studies that include PROs, our findings reinforce the importance of symptom reporting as described by patients themselves.

To date, this study presents the first analysis to focus specifically on patient reporting of GI symptoms for a regimen with a high rate of GI toxicity, and we used various scales to obtain a deeper understanding of the impact of GI symptoms on quality of life and function. Prior phase I and II studies of the neratinib plus T-DM1 combination have reported exclusively clinician-assessed GI toxicity (CTCAE), with rates of ≥ grade 2 diarrhea of 27%–36%, typically emerging around Cycle 2 [8,22]. In the ExteNET trial, which evaluated neratinib in the extended adjuvant setting following adjuvant trastuzumab and did not mandate antidiarrheal prophylaxis, grade 3 diarrhea was reported in up to 40% of patients [23]. Our findings are consistent with these reports, showing a significant worsening in GI symptoms when measured by PRO tools. Specifically, the PROMIS Diarrhea score increased by 6.62 points [95% CI: 2.67–10.59] after two cycles—a change that is clinically meaningful, exceeding the 0.5 SD threshold. In contrast, the STIDAT score also worsened by 0.50 [95% CI: 0.11–0.90], though this change did not reach clinical significance. PRO-CTCAE analysis further confirmed the early onset of diarrhea as perceived by patients and also revealed appetite loss and constipation, with 9.8 and 28.2% of patients reporting grade 3 severity, respectively. These symptoms were notably underreported in clinician-assessed CTCAE, where only 2.3% of patients were recorded as having grade 2 appetite loss or constipation [8]. Understanding the timing and severity of symptoms from the patient perspective enables more targeted and proactive supportive care interventions. It may also support the development of dose-escalation strategies, starting with lower initial doses to improve tolerability during the early treatment cycles, a strategy that is now part of the US Prescribing Information and National Comprehensive Cancer Network^Ⓡ (NCCN) guidelines based on the CONTROL trial [11,24,25].

As mentioned, we observed limited agreement between traditional CTCAE assessments and patient-reported data. Furthermore, our sensitivity analysis confirmed that this limited agreement persisted even when considering only the presence or absence of symptoms, suggesting that differences in agreement are not driven by differences in severity but rather by discordance in symptom reporting. A pooled analysis of individual participant data from the CLEOPATRA, EMILIA, and MARIANNE trials identified significant associations between PROs, assessed using the Functional Assessment of Cancer Therapy—Breast (FACT-B) questionnaire, and the occurrence of grade ≥3 adverse events [26]. While these specific tools were not evaluated in our study, the findings from that analysis also showed that PROs can serve as independent prognostic factors for both overall survival (OS) and progression-free survival (PFS).

Importantly, among the small subset of patients who discontinued treatment due to toxicity, patient-reported symptom severity was heterogeneous, suggesting that treatment discontinuation may reflect overall treatment burden and patient experience rather than the intensity of a single symptom. These findings further support the complementary value of PROs beyond clinician-reported toxicity grading alone.

Based on this protocol and PRO findings, several considerations may guide clinical evaluation and management of GI toxicities with neratinib plus T-DM1. First, close monitoring during the initial two to three cycles is essential, as symptoms typically emerge early and are most severe in this period. Second, prophylactic antidiarrheal therapy (e.g., loperamide, colestipol) should be initiated at treatment start, with careful adjustment to avoid overtreatment and constipation. Third, patient education and proactive symptom reporting should be emphasized, as clinician assessments alone underestimate symptom burden. Finally, emerging evidence supports neratinib dose-escalation strategies, starting with lower initial doses to improve tolerability during early cycles; this approach is now reflected in the United States Prescribing Information and NCCN guidelines. Together, these strategies may optimize tolerability and maintain treatment adherence.

Missing PRO data remains a major challenge in oncology clinical trials and may affect the validity and interpretation of patient-centered endpoints. In our study, PRO completion rates declined over time, which is not unexpected in a heavily pretreated metastatic population with progressive disease and treatment-related toxicities. Recent recommendations to improve PRO completion emphasize the importance of integrating PRO assessments into routine clinical workflows, using electronic PRO platforms with automated reminders, implementing real-time monitoring of missing questionnaires, and providing dedicated site-level support and patient education regarding the importance of PRO collection [27]. However, these strategies must also balance patient burden, particularly in advanced cancer populations where frequent reminders or lengthy questionnaires may themselves contribute to distress or non-completion.

A notable limitation is that data were collected for only the first four cycles, a choice made prospectively to focus on the expected peak window of toxicity while minimizing attrition due to disease progression. Although PROs were only collected through Cycle 4, we observed that diarrhea peaked at Cycle 3 and slightly improved thereafter; however, grade 1 symptoms remained common. The clinical relevance of persistent low-grade symptoms should not be overlooked, as chronic grade 1 diarrhea may still negatively affect quality of life and treatment experience over time. Additionally, the study was small and excluded patients with pre-existing diarrhea, which may limit generalizability. Another limitation is that treatment interventions, such as dose reductions in response to toxicity, were not recorded alongside PRO data, making it difficult to contextualize changes in symptom severity over time. Further, the adherence to anti-diarrheal medications and prophylaxis in cycle 1 varied, and this wasn't consistently documented or available for analysis. The balance between antidiarrheal prophylaxis and diarrhea control is patient-specific and may contribute to constipation symptoms in some cases. Beyond maximum-grade reporting, longitudinal assessment of symptom trajectories and time spent at each grade may provide a more accurate picture of the sustained burden of treatment — particularly when low-grade symptoms persist over extended periods.

In summary, this study is the first to report patient-reported gastrointestinal toxicities with neratinib and T-DM1 in HER2-positive breast cancer brain metastases. Using PROMIS, STIDAT, and PRO-CTCAE, we found that diarrhea, appetite loss, bloating, and constipation were common, often underrecognized by clinicians, and most severe in the first treatment cycles. These results highlight the value of PROs in capturing symptom timing and severity as well as impact on quality of life, supporting their routine use to guide supportive care and improve treatment tolerability.

Data availability statement

Due to the nature of the research, data from this trial are not available for sharing or use. The participants of this study did not give written consent for their data to be shared publicly or with other researchers.

Funding

This work was supported by funding provided by Puma Biotechnology, Inc. We appreciate their support. We are also grateful for funding from the Breast Cancer Research Foundation to Nancy U. Lin; and the Translational Breast Cancer Research Consortium (TBCRC) through its three foundation partners: the AVON Foundation, the Breast Cancer Research Foundation, and Susan G. Komen for the Cure. None of the above funders played any role in the analysis or interpretation of the study findings, and no grant numbers apply.

CRediT authorship contribution statement

Thomas Grinda: Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Hillary M. Heiling: Conceptualization, Data curation, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing. Nabihah Tayob: Conceptualization, Data curation, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing. Karen L. Smith: Conceptualization, Methodology, Writing – review & editing. Raechel Davis: Data curation, Formal analysis, Methodology, Project administration, Writing – review & editing. Christine Cotter: Data curation, Formal analysis, Methodology, Project administration, Writing – review & editing. Michelle K. DeMeo: Data curation, Formal analysis, Methodology, Project administration, Writing – review & editing. Cesar A. Santa-Maria: Conceptualization, Methodology, Writing – review & editing. Catherine Van Poznak: Conceptualization, Methodology, Writing – review & editing. Beverly Moy: Conceptualization, Methodology, Writing – review & editing. Adam M. Brufsky: Conceptualization, Methodology, Writing – review & editing. Michelle E. Melisko: Conceptualization, Methodology, Writing – review & editing. Ciara C. O'Sullivan: Conceptualization, Methodology, Writing – review & editing. Claudine Isaacs: Conceptualization, Methodology, Writing – review & editing. Yasmeen Rauf: Conceptualization, Methodology, Writing – review & editing. Julie R. Nangia: Conceptualization, Methodology, Writing – review & editing. Robyn T. Burns: Project administration, Resources, Writing – review & editing. Jennifer Savoie: Project administration, Resources, Writing – review & editing. Antonio C. Wolff: Conceptualization, Methodology, Writing – review & editing. Eric P. Winer: Conceptualization, Methodology, Writing – review & editing. Erica L. Mayer: Conceptualization, Methodology, Writing – review & editing. Sara M. Tolaney: Conceptualization, Methodology, Writing – review & editing. Mothaffar F. Rimawi: Conceptualization, Methodology, Writing – review & editing. Kathryn J. Ruddy: Conceptualization, Methodology, Writing – review & editing. Gabrielle Rocque: Conceptualization, Methodology, Writing – review & editing. Ian E. Krop: Conceptualization, Methodology, Writing – review & editing. Nancy U. Lin: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing. Rachel A. Freedman: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: KS reports employment at and stock holdings in Merck; and spousal ownership of stock in Abbott Labs and Abbvie. CI reports consulting work for Arvinas, AstraZeneca, Genentech, Novartis, Pfizer, Gilead, Merck, and Seattle Genetics; royalties from Wolters Kluwer and McGraw Hill; and research support (to institution) from Tesaro/GSK, Seattle Genetics, Pfizer, AstraZeneca, Bristol Myers Squibb, Genentech, Novartis, and Regeneron. ELM reports consulting for Novartis, Genentech, Lilly, AstraZeneca, and Aktis. CVP reports royalties from UpToDate for writing on topics not related to the manuscript at hand. EPW reports consulting for 4D Path. JRN reports advisory board service for Gilead. CO’S reports research funding to her institution from Genentech, Pfizer, and Jazz Pharmaceuticals; service on the steering committees of AstraZeneca and Pfizer; general consulting for Daiichi Sankyo; and employment with Atlas Oncology Partners. GR reports research funding from Pfizer, Gilead, and Daiichi Sankyo; consulting for Gilead and Pfizer; and employment with Atlas Oncology Partners. AMB reports consulting for AstraZeneca, Pfizer, Novartis, Lilly, Genentech/Roche, Daiichi Sankyo, Merck, Agendia, Sanofi, Puma, Myriad, Gilead, Bria-Cell, and Celcuity; and research support from Agendia and AstraZeneca. TG reports travel fees from AstraZeneca, Gilead, Lilly, Novartis and Pfizer; consulting/advisor roles for Pfizer, AstraZeneca, Roche, Gilead and Cancerologie-pratique; and a personal grant from Philippe Foundation. SMT reports consulting or advisory roles for Novartis, Pfizer/Seagen, Merck, Eli Lilly, AstraZeneca, Genentech/Roche, Eisai, Bristol Myers Squibb/Systimmune, Daiichi Sankyo, Gilead, Blueprint Medicines, Reveal Genomics, Artios Pharma, Menarini/Stemline, Bayer, Jazz Pharmaceuticals, Cullinan Oncology, Circle Pharma, Arvinas, BioNTech, Launch Therapeutics, Zuellig Pharma, Johnson&Johnson/Ambrx, Bicycle Therapeutics, BeiGene Therapeutics, Mersana, Summit Therapeutics, Avenzo Therapeutics, Aktis Oncology, Celcuity, Boehringer Ingelheim, Samsung Bioepis, Olema Pharmaceuticals, Tempus, Boundless Bio, Denali Therapeutics, Relay Therapeutics, Corcept, Ottima Pharma, and Ellipses Pharma; science advisory board service for Valanx BioTech; research funding from Genentech/Roche, Merck, Exelixis, Pfizer, Lilly, Novartis, Bristol Myers Squibb, AstraZeneca, NanoString Technologies, Gilead, Seagen, OncoPep, Daiichi Sankyo, Menarini/Stemline, Jazz Pharmaceuticals, and Olema Pharmaceuticals; and travel support from Lilly, Gilead, Jazz Pharmaceuticals, Pfizer, Roche, and AstraZeneca. NUL reports institutional research support from Genentech, Pfizer, Merck, Seattle Genetics, Zion Pharmaceuticals (as part of Roche/GNE), Olema Pharmaceuticals, AstraZeneca, Iksuda, and Stemline/Menarini; consulting honoraria from Pfizer/Seagen, Daiichi Sankyo, AstraZeneca, Olema Pharmaceuticals, Stemline/Menarini, Artera Inc., Eisai, Shorla Oncology, Denali Therapeutics, and Genentech; royalties from UpToDate (book); and travel support from Olema, AstraZeneca, and Daiichi Sankyo. RAF reports institutional funding from Puma. The remaining authors do not report any conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: