Introduction

Treatment-induced nausea (TIN) is a prevalent and highly distressing symptom experienced by the 2.3 million women globally [1], undergoing treatment for breast cancer each year, spanning interventions such as surgery, radiation, chemotherapy, and immunotherapy [2]. Despite its widespread occurrence and its profound impact on patients’ quality of life, the mechanisms underlying TIN remain mostly unknown [3, 4]. This gap in knowledge poses a significant challenge for both clinicians and researchers, limiting the development of effective prevention and treatment strategies.

Although nausea and vomiting are commonly grouped together, such as postoperative nausea and vomiting (PONV) or chemotherapy-induced nausea and vomiting (CINV), recent advances in antiemetics have been clinically effective for managing treatment-induced vomiting (TIV) [5, 6]. However, nausea has remained a significant clinical problem for up to 68% of patients treated for breast cancer even when prescribed guideline-consistent antiemetics [7, 8].

Same-day surgery is now the standard for most breast cancer surgeries [9]; thus, patients report very little postoperative nausea (PON) due to multimodal guidelines to prevent PON and early discharge to home following surgery [6]. However, women face post-discharge nausea (PDN) at home from day one, and PDN persists past the usually reported 72-h post-surgery period. Our research and that of others show that at least 30% of women continue to experience PDN for 48 h to up to a week [6], following breast cancer surgery despite adherence to PONV antiemetic guidelines [6, 10]. Additionally, the relationship between PONV and pregnancy-induced nausea and vomiting (PNV) has been reported [11, 12].

Following surgery, at least 20% of women with breast cancer begin chemotherapy. CIN can lead to treatment plan interruption, dehydration, malnutrition, inability to return to work, a decrease in quality of life, and additional health care utilization [13]. Evidence-based treatment guidelines [14] for CIN are, in most cases followed closely; yet, in women with breast cancer, 40–50% of those prescribed chemotherapy experience CIN even though clinical guidelines are followed [15]. The risk for radiation-induced nausea is considered minimal [16]; however, there are several studies that have focused on patient-reported radiation-induced nausea in women treated for breast cancer [17,18,19].

Growing evidence suggests that the inability to treat the 30% of women who do not respond to antiemetic medications may be explained by genomic risk factors [20]. Recent attention has been given to a significant association of pregnancy-induced nausea and Growth/differentiation factor 15 (GDF15) [21, 22]. Originally identified as a stress-responsive cytokine with diverse physiological functions, GDF15 is now recognized as a growth-differentiation factor with cytokine-mediated metabolic signatures for maintaining homeostasis. Recent studies suggest that GDF15 has a significant role in metabolism, appetite regulation, and stress response [23, 24]. GFRAL-RET (GDNF family receptor alpha-like and rearranged during transfection) are the high-affinity receptors for GDF15 [25]. GFRAL-NET is exclusively expressed in the area postrema (AP) and the nucleus of the solitary tract (NTS). The AP/NTS are two critical regions in the hindbrain involved in the emetic pathway [26].

While the association between the GDF15 gene and pregnancy-induced nausea and vomiting (PNV) is beginning to be understood [23], recent research measuring serum level for GDF15 protein suggests that elevated levels may also contribute to postoperative nausea and vomiting (PONV) in women undergoing breast cancer surgery [27]. In addition, high levels of circulating GDF15 have been associated with receipt of chemotherapy. Thus, we hypothesize that GDF15 may mediate nausea and vomiting [24] related to patient cancer treatment. Therefore, the purpose of this study was to determine if variability in GDF15 and a positive history of pregnancy-induced nausea and vomiting are predictors of treatment-induced nausea severity across the trajectory of the first 6 months following surgery in women treated for early-stage breast cancer.

Methods

Research design, setting, and sample

This prospective, 6-month, longitudinal cohort study recruited patients scheduled for breast cancer surgery (without reconstruction) at a large, NCI-designated cancer center hospital. Patients were eligible for study participation if they met the following criteria: (1) biological female aged 18–90 years; (2) diagnosed with early-stage breast cancer (stage I, II, IIIA); (3) had a negative history for previous neurological conditions that could also be a cause of nausea and vomiting; (4) were assigned an American Society of Anesthesiologists (ASA) physical status of I, II, or III; (5) had access to a computer or a telephone to report symptoms experienced; and (6) were able to read and write in English. The study was approved by the University of Pittsburgh Institutional Review Board. The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments All study participants engaged in active informed written consent prior to the initiation of data collection.

Procedures and phenotype data collection

Socio-demographic data associated with increased risk for nausea were collected at the time of recruitment and included age, race/ethnicity, history of pregnancy-induced nausea and vomiting, and smoking status. Female gender (which included entire sample) is also a major risk factor for treatment-induced nausea and vomiting across the trajectory of treatments [28, 29]. Nausea severity was measured with a 0–10 numeric nausea scale [30] following surgery, 48 h after surgery, and then weekly during adjuvant therapy. On the nausea numeric scale, 0 indicated no nausea and 10 was the most severe nausea ever experienced. All study participants completed the data collection at every time point.

Genetic data collection

Saliva samples were obtained at the time of surgery using the Oragene® Saliva Collection Kit following the standardized protocol provided by DNAGenotek (Ontario, Canada). DNA data were extracted from the saliva collection system. Genotyping was conducted using the Illumina Infinium Global Diversity Array (San Diego, California). Based on recent research and those available from the Global Diversity Array, the genetic factors selected for this analysis were single nucleotide polymorphisms (SNPs) of the GDF15 gene. Selected SNPs were rs810804, rs1227731, rs8101249, and rs1059369. The selected SNPs were selected based on previous studies [31], with sufficient variability in the minor allele frequency for analysis. The genotype distribution for the four GDF15 SNPs included in the analysis is shown in Table 1.

Table 1 Distribution of genotypes of the GDF15 polymorphisms included in the analysis
Full size table

Statistical analysis

Data were analyzed using SAS version 9.4 (SAS Institute, Inc., Cary, NC) with the level of statistical significance established at 0.05 (two-tailed) for non-directional hypothesis testing and confidence intervals set at 95%. Data were first screened for any anomalies (i.e., outliers, large amounts and nonrandom patterns of missing data, and non-normality of residuals) that may invalidate study results when applying the desired statistical strategies, and as needed remedial strategies were applied (e.g., data transformations, meaningful collapsing of categories). Participant characteristics were summarized for the full sample using appropriate descriptive statistics given the variable’s measurement level and empirical data distribution. A spaghetti plot was first generated for numeric nausea severity scores to visually explore the variability of scores within and between participants from 48-h post-discharge through the first 6 months of follow-up. Given the high interpersonal variability in responses expected over time, group-based trajectory modeling (GBTM) using PROC TRAJ in SAS [32] was used to identify distinct groups or classes of participants whose numeric nausea severity scores over the first 6 months after breast cancer surgery followed a similar temporal pattern or trajectory. Using a recommended approach [33] and assuming a censored normal distribution, the complexity of the model for the numeric nausea severity score was determined in terms of the number of groups and for each group the shape of the trajectory (flat/stable, linear, quadratic, cubic) using the approximated Bayes Factor to compare the Bayesian Information Criteria (BIC) between competing models using the BIC based on the total number of repeated assessments (BIC1) and that based on the number of participants (BIC2). The adequacy of the unconditional group-based trajectory model was assessed using recommended group-specific diagnostics including the average of posterior probabilities based on group assignment after applying the assignment rule, the odds of correct classification, the differences between the observed proportions based on the assignment rule and the estimated group membership probabilities, and the confidence intervals of the estimated group membership probabilities. Results reported for the final unconditional model included the estimated group-specific trajectory coefficients and their corresponding p-values from t-tests, the BIC1 and BIC2 values for the final model, and the estimated group membership probabilities and observed proportions for each group. Plots of the observed and predicted group-specific trajectories based on the final unconditional model were also generated.

The best-fitting group-based trajectory model for the numeric nausea severity scale was then expanded to include each targeted predictor variable (i.e., four GDF-15 SNPs and a history of nausea and vomiting during pregnancy) first singly to yield crude/unadjusted estimated model coefficients (i.e., log-odds) and then jointly considering each SNP with the history of nausea and vomiting during pregnancy to yield adjusted estimated model coefficients. As most of the SNPs for GDF-15 had limited variability for the homozygous variant category, the heterozygous and the homozygous variant categories were also collapsed to yield binary versions of the SNPs (i.e., having at least one copy of the variant allele versus no copies of the variant allele). The full multivariate model with all predictors (i.e., binary predictors for SNPs and history of nausea and vomiting during pregnancy) simultaneously was also estimated for the numeric nausea severity score. Lastly, from the full multivariate model, a parsimonious model derived using a manual backward elimination approach where the p-value to retain a predictor was set at 0.05. Post hoc model assessment was conducted for all models.

Results

Study participants (n = 290) were mostly White (87%) with an average age of 60 years, an average BMI of 30, and very few (7%) were smokers (Table 2). More than one-third (n = 108; 37%) reported a positive history of pregnancy-induced nausea. TIN for adjuvant radiation therapy was reported in 42% of study participants (n = 222) while receiving radiation therapy and 64% of study participants (n = 81) who received chemotherapy. Though the number of women receiving targeted therapies was low (n = 37) more than half reported nausea or gastrointestinal disturbance months following surgery.

Table 2 Demographic and treatment characteristics for study participants (n = 290)
Full size table

The spaghetti plot displayed in Fig. 1 shows considerable heterogeneity in the individual response trajectories between and within participants for the monthly numeric nausea severity scores from the 48-h post-discharge period following breast cancer surgery and monthly for the first 6 months of breast cancer treatment. Based on GBTM applied to numeric nausea severity scores assuming a censored normal distribution the best-fitting unconditional model based on BIC1 and BIC2 values and demonstrating model adequacy using the recommended diagnostics and cutoffs (including somewhat narrow confidence intervals the estimated membership probabilities) for revealed three distinct, stable (i.e., intercept only) trajectories: extremely low (n1 = 119), low (n2 = 135), and moderate (n3 = 35) (see Fig. 2 and Table 3). Model adequacy was supported based on recommended model diagnostics.

Fig. 1
figure 1

Spaghetti plots of the numeric nausea severity scores over the first 6 months following breast cancer surgery (N = 290)

Full size image
Fig. 2
figure 2

Observed and predicted group trajectories from unconditional group-based trajectory modeling of the numeric nausea severity score (solid line, observed mean; dashed line, predicted mean)

Full size image
Table 3 Results from the unconditional group-based trajectory modeling of the numeric nausea severity score (N = 290)
Full size table

As reported in Table 4, the addition of the targeted phenotypic and genotypic predictors one by one to the trajectory model showed significant positive effects for the history of nausea and vomiting during pregnancy for membership in the stable low severity and the stable low/moderate severity groups compared to the extremely low severity group and for at least one copy of the variant allele for rs8101249 for membership in the low severity group compared to the extremely low severity group. In contrast, a negative effect was observed for rs1227731 for either having at least one copy of the variant allele for membership in the stable low severity group compared to the extremely low severity group. When all predictors were considered simultaneously in the full multivariate model, only the positive effect for the history of nausea and vomiting during pregnancy was maintained. However, when the included predictor variables were limited to only those that were significant using a backward elimination approach (i.e., the parsimonious model), both the positive effect for the history of nausea and vomiting during pregnancy and the negative effect of having at least one copy of the variant allele for rs1227731 were retained, both for membership in the low severity group compared to the extremely low severity group.

Table 4 Predictors of trajectory group membership using conditional group-based trajectory analysis for nausea severity (N = 290)
Full size table

Discussion

Consistent with current research, this study confirms the persistence of treatment-induced nausea (TIN) among women undergoing breast cancer therapy across the trajectory of treatments [4, 34]. As expected, very few women in this study experienced vomiting; new medication protocols have significantly decreased the occurrence of vomiting. Despite adherence to evidence-based antiemetic protocols, a significant proportion of patients reported persistent nausea across different treatment phases, including post-discharge nausea, radiation-induced nausea, and chemotherapy-induced nausea. Importantly, pregnancy-induced nausea and vomiting was found to be a significant predictor of membership in higher nausea severity groups, and genetic variations in GDF15, particularly rs8101249 and rs1227731, showed promising associations with treatment-induced nausea severity. The SNP rs1227731 is an intronic GDF15 SNP that may modulate GDF15 expression through the GDF15–GFRAL-RET signaling pathway. This single intron of GDF15 harbors a microRNA (miR-3189), though the SNP’s impact on this miRNA or splicing is not yet understood [31, 35]. The other SNP rs8101249 is located in the 3′untranslated region (3′UTR) of GDF15. Though the 3′UTR is part of the final exon, it does not code for protein, but a variant in this location may affect mRNA stability or regulation perhaps by disrupting microRNA binding sites [31]. The identification of genetic factors is consistent with prior work linking GDF15 to pregnancy-induced nausea and highlights its potential association to TIN [36].

These findings underscore the need to consider inter-individual variability in the experience of TIN, moving beyond a one-size-fits-all approach to symptom management. The significant association between pregnancy-induced nausea and TIN aligns with the growing evidence linking GDF15, a responsive cytokine implicated in metabolism and appetite regulation [24], and its receptors located in the AP/NTS with nausea. The persistence of this association in the multivariate model highlights the potential for leveraging clinical history to identify patients at higher risk of TIN.

The findings have important clinical implications. First, incorporating clinical history, such as pregnancy-induced nausea, into risk assessment tools may improve identification of patients at heightened risk for all TIN. Second, the emerging role of GDF15 as a biomarker for nausea suggests that targeting its signaling pathways could yield new therapeutic approaches. Functional studies of GDF15 variants may clarify their role in nausea pathophysiology and guide the development of precision medicine strategies, which hopefully in the future will include insurance coverage for genetic testing for symptom management.

A limitation of this study is the reliance on self-reported pregnancy-induced nausea and vomiting. Due to the time elapsed since participants’ pregnancies, we could not assess symptom severity for direct comparison with treatment-induced nausea. However, because pregnancy-related nausea and vomiting is typically memorable, we were confident using this as a binary variable. This supports our hypothesis that GDF15, along with a positive history of pregnancy-induced nausea and vomiting, may influence nausea and vomiting associated with cancer treatment in women with breast cancer.

Conclusion

This study highlights the persistence and variability of TIN among women across the trajectory of treatment modalities for breast cancer, emphasizing the need for tailored symptom management approaches. The associations between pregnancy-induced nausea, GDF15 genetic variants, and TIN severity provide a foundation for future research aimed at improving symptom prediction and intervention strategies. By bridging clinical and genetic insights, this work contributes to the broader effort to improve quality of life for women undergoing breast cancer treatment.