Introduction

Despite dramatic improvements in antiemetic control conferred by serotonin (5-HT3) and neurokinin (NK-1) receptor antagonists and prescriptive antiemetic guidelines, approximately 50 % of cancer patients still experience nausea and vomiting (N&V) during treatment, with nausea proving to be more pervasive than vomiting (incidence: 37–70 vs. 13–34 %) [1–3]. Several patient, clinical and quality of life (QoL) factors associated with the development of N&V during treatment have been identified [3–6], the most important being the emetogenic potential of chemotherapy received [7]. Other factors include female gender [3–5], younger age [4–6], a history of N&V [5] or low alcohol consumption [4], preexisting anxiety [5, 6], a high expectation of developing N&V after chemotherapy [4, 5], emetogenic potential of radiotherapy [7] and undergoing surgery [4–6] or combined therapy [7].

Uncontrolled treatment-induced nausea and/or vomiting (TIN+/−V) due to chemotherapy, radiotherapy and/or surgery results in a range of physical and psychosocial symptoms, which impact considerably on patients’ QoL and healthcare costs [2, 3, 6, 8–11]. Nevertheless, few prospective studies have quantitatively demonstrated that aspects other than physical functioning are adversely affected by uncontrolled TIN+/−V [8, 12]. Moreover, studies examining TIN+/−V have commonly assessed the impact of chemotherapy-induced nausea and vomiting (CINV) per se on QoL, and have done so using the Functional Living Index–Emesis (FLIE) questionnaire [8–10, 13]. Unfortunately, the FLIE has some limitations, most notably, the use of aggregate scores that may lack sensitivity to detect differences within and across individual QoL domains, a relative lack of published psychometric evaluation and the omission of concomitant symptoms that patients may experience (e.g. fatigue) [14, 15]. Patients most at risk of TIN+/−V include those who have received previous chemotherapy, are planned for concurrent chemoradiation or have poor performance status. Most TIN+/−V studies have been RCTs which have often excluded these groups; thus, TIN+/−V outcomes from these trials cannot be generalised [2–4, 8, 10].

Current evidence-based antiemetic guidelines (e.g. Multinational Association of Supportive Care in Cancer/European Society for Medical Oncology (MASCC/ESMO) [7], American Society of Clinical Oncology [16], National Comprehensive Cancer Network) [17] maintain that the goal of antiemetic therapy is complete N&V prevention. The inability, thus far, to achieve a desirable level of control for all patients is multifactorial. Reasons include the predominating failure to control nausea, continued use of emetogenic chemotherapy regimens, incomplete understanding of the mechanisms underlying TIN+/−V, underestimation of TIN+/−V incidence, lack of compliance with antiemetic guidelines, infrequent/inadequate assessment of TIN+/−V and its risk factors in routine clinical practice, antiemetic development targeting vomiting at the expense of nausea and previous studies examining TIN+/−V within the constraints of clinical trials [18, 19].

Another reason, albeit little-considered, is that N&V may be part of a more extensive symptom cluster, a concept which has only recently become prominent in cancer QoL research. Symptom clusters are defined as follows:

“A symptom cluster comprises 2 or more symptoms that are related to each other and that occur together. Symptom clusters are composed of stable groups of symptoms, are relatively independent of other clusters, and may reveal specific underlying dimensions of symptoms. Relationships among symptoms within a cluster should be stronger than relationships among symptoms across different clusters. Symptoms in a cluster may or may not share the same etiology.” [20] (p. 278)

Symptom clustering may be therapeutically important for TIN+/−V because treatment of nausea or vomiting may be affected or influenced by other symptoms occurring in a cluster with them (e.g. bloating, difficulty swallowing) [21, 22].

The goals of the present study were to assess the “real-world” incidence of TIN+/−V (i.e. acute N&V developing within 24 h of treatment + delayed N&V developing more than 24 h, up to several days, after treatment) [7] in cancer patients receiving combined modality treatment in a routine clinical setting; identify the presence and composition of any symptom cluster(s) involving N&V over time at the onset, during and at the end of treatment; and, most importantly, examine the impact of nausea, vomiting and any symptom clusters containing them, respectively, on the QoL and psychological adjustment of patients across combined treatment.

Materials and methods

A secondary analysis of data collected from a prospective, longitudinal, observational design involving a heterogeneous group of 200 cancer patients was performed. The ethics committees of the participating institutions approved the study. All patients provided written and informed consent.

Patients

Participants were recruited for a larger study evaluating QoL and psychosocial distress from a consecutive series of 287 eligible medical oncology outpatients, who largely received combined treatment (including chemotherapy) at Royal Perth Hospital, Western Australia, between 1997 and 2003. In the week before treatment began, patients were approached and given a verbal explanation of the study plus an information sheet.

Eligibility criteria

Eligibility criteria included histological confirmation of cancer, age of 18 years and above, absence of acute psychiatric symptoms or conditions that could cause emesis (i.e. central nervous system/head and neck metastasis, (ongoing) gastrointestinal obstruction, a history of motion sickness or alcohol/drug abuse, pregnancy, metabolic/electrolyte imbalances, pain conditions treated with non-stable doses of opioids and any other uncontrolled medical condition that may cause nausea and vomiting, such as increased intracranial pressure or hypercalcaemia; see Table 1 footnote), no prior cancer treatment for the current diagnosis (excluding surgery) and adequate English literacy and cognitive ability to complete the study questionnaires.

Table 1 Patient and clinical characteristics of the sample (N = 200)
Full size table

Data collection

Data concerning demographics, clinical characteristics and potential risk factors for TIN+/−V were collected from patients, oncologists and medical records. Questionnaires were completed by patients at pretreatment (within 7 days of the start of chemotherapy ± radiotherapy), on-treatment (8 weeks ± 1 week), post-treatment (within 7 days of the last treatment received) and follow-up (6 months). As this study was specifically aimed at assessing TIN+/−V (acute + delayed N&V) and associated symptom clusters in patients during treatment, follow-up data were excluded from analysis. Data from patients not completing questionnaires within the specified intervals were also excluded.

The on- and post-treatment assessment, in each instance, occurred at 7 days following the end of a treatment cycle to capture both acute and delayed N&V (i.e. TIN+/−V incidence), as well as other treatment effects. Also, the on-treatment assessment was performed at 8 weeks (± 1 week), chiefly to minimise patient burden associated with more frequent or ongoing assessment (e.g. patient diaries) that would have added to the more extensive battery of questionnaires (100 items) employed in the larger QoL study. Moreover, it was considered an optimal time point to measure the incidence of acute and delayed N&V (and other side effects) during treatment, as it coincided with the administration of chemotherapy (undertaken by all patients, unlike radiotherapy and surgery) that was generally given weekly or every 3–4 weeks for a period of 3–12 months.

Questionnaires

Patients were administered standardised questionnaires assessing QoL and psychological distress, which have been described in detail previously [23]. They included the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30 (EORTC QLQ-C30) V2.0 [24, 25], Selby Quality of Life (QoL) Uniscale (overall QoL) [26], Physical Health (PH) Uniscale (overall physical health) [24, 26], Beck Depression Inventory Short Form (BDI-SF; clinical depression) [27], Impact of Event Scale-Intrusion Subscale (IES-IS; cognitive–emotional distress related to cancer) [28] and Eysenck Personality Questionnaire-Revised (EPQ-R)—Neuroticism Short Form (SF) (premorbid neuroticism at pretreatment only) [29]. The EORTC QLQ-C30 assessed the study’s primary endpoint for N&V, defined in the present study as the absence of total control (i.e. no nausea, vomiting or retching and no rescue medication).

Statistical analysis

Exploratory factor analysis (EFA) in the form of principal components analysis, using both oblique and orthogonal rotations, was performed on QoL (EORTC QLQ-C30, Selby QoL Uniscale, PH Uniscale) at each treatment time point to identify any symptom cluster (factor) containing nausea and/or vomiting. Criteria to determine the best factor model included: (a) exclusion of items exhibiting Kaiser’s measurement of sampling adequacy, Kaiser–Meyer–Olkin or communality values < 0.5, (b) Bartlett’s test of sphericity being significant (p < 0.0001), (c) retention of factors with eigenvalues > 1 that were supported by scree plot analysis and clinical/theoretical plausibility of factors, (d) factor loadings ≥ 0.4 for interpretation of factors (explaining ≥16 % of total variance), (e) Cronbach’s alpha ≥ 0.7 for each factor and (f) stability of factor solutions across analyses employing oblique and orthogonal rotations [30, 31]. Stability or persistence of the symptom cluster (factor) containing nausea and/or vomiting across treatment time points was also established quantitatively by the criterion that at least 75 % of symptoms in the initial cluster (pretreatment) must be present in subsequent clusters (on- and post-treatment) [31].

The impact on QoL domain scores and psychological functioning of any TIN+/−V symptom cluster identified and its individual symptoms was examined across treatment (pre-, on- and post-treatment) via univariate analyses using split-plot analyses of covariance. Additionally, the impact on overall QoL of any TIN+/−V symptom cluster identified was assessed at: (a) each treatment time point via multivariate analyses using hierarchical multiple regression (p ≤ 0.1 for retention of predictor variables in each regression model) and (b) across treatment using a multiple regression model with repeated measures (linear mixed model analysis, LMM), with the added ability to retain patients in the analysis who have missing data across one or more treatment time points [30, 32]. The level of significance for all analyses was set at p ≤ 0.05.

Results

Patient characteristics

Two hundred (70 %) of the eligible patients consented and completed baseline questionnaires. Of these, 178 (89 %) patients advanced to on-treatment (8 weeks ± 1 week) and 153 (76 %) progressed to post-treatment. During treatment, 47 patients were withdrawn due to death (15), changes in medical care (15), patient withdrawal (15) and loss to follow-up (two).

Patient and clinical characteristics have been described in detail previously [23] and are summarised in Table 1; they are typical of Western Australian cancer patients at the time of the study [33]. Briefly, all patients received chemotherapy; 161 (80 %) were chemotherapy-naïve. Overall, 239 chemotherapy regimens were administered (range of cycles: 2–12); fluorouracil + leucovorin (FU/LV, 19 %) and cyclophosphamide + methotrexate + FU (CMF, 11 %) were most prevalent. According to the MASCC/ESMO antiemetic guidelines [7], 102 (51 %) patients received moderately/highly emetogenic chemotherapy (M/HEC; ≥ 30 % emetic risk without prophylaxis), and the remaining 98 (49 %) received minimally/low emetogenic chemotherapy (M/LEC; <30 % emetic risk without prophylaxis).

Antiemetic therapy consistent with guidelines in force at the time and place of the study was used for most patients (this predated the NK-1 antagonists). Overall, 130 (65 %) patients received 5-HT3 receptor antagonists (e.g. ondansetron) + corticosteroids (e.g. dexamethasone), 55 (28 %) were prescribed dopamine receptor antagonists (e.g. metoclopramide) and 43 (22 %) patients received no antiemetics.

Prevalence/incidence, persistence and severity of TIN+/−V

Overall, TIN+/−V was experienced by 123 (62 %) patients, TIN only by 70 (35 %) patients, TIV only by four (2 %) patients and concurrent TIN+/−V by 50 (25 %) patients. Thus, TIN occurred in 120 (60 %) patients and TIV in 54 (27 %) patients overall. At on-treatment (8 weeks), TINV was reported by 87 (49 %) patients, TIN only by 55 (31 %) patients, TIV only by four (2 %) patients and concurrent TINV by 28 (16 %) patients. Thus, TIN occurred in 83 (47 %) patients and TIV in 32 (18 %) patients at on-treatment (8 weeks). Finally, TINV was reported by 86 (56 %) patients, TIN only by 51 (33 %) patients, TIV only by three (2 %) patients and concurrent TINV by 32 (21 %) patients at post-treatment. Thus, TIN occurred in 83 (54 %) patients and TIV in 35 (23 %) patients at post-treatment.

Impact of chemotherapy emetogenicity on patients’ quality of life and psychological adjustment

No significant differences were found on any QoL or psychological dimension for patients treated with M/HEC (n = 96) when compared to those who received M/LEC (n = 57; 0.131 ≤ p ≤ 0.97) after controlling for covariates (previous chemotherapy experience, surgery prior to adjuvant chemotherapy ± radiotherapy, variable treatment duration, antiemetic use, baseline QoL/psychological functioning score at pretreatment).

Identification of a TIN+/−V symptom cluster and its pattern across time during treatment

EFA was conducted on QoL scores at each treatment time point. Near-identical factor solutions were produced by orthogonal (varimax) and oblique (oblimin) rotations; thus, results of the latter are reported only (Table 2). At pretreatment, a six-factor solution explaining 68.2 % of the total variance in QoL was extracted. Of this variance, 7.2 % (approximately one tenth) was accounted for by a gastrointestinal (GI) symptom cluster consisting exclusively of nausea, vomiting and appetite loss. Similarly, a seven-factor solution emerged at both on-treatment and post-treatment, explaining 67.6 and 72.6 % of the total variance in QoL, respectively. Of the variance, 6.9 and 8.3 % (approximately one tenth) was accounted for by the same GI symptom cluster at on- and post-treatment, respectively. Cronbach’s alpha for the GI cluster across treatment was 0.74, 0.73 and 0.73, respectively, indicating acceptable internal consistency. Consequently, all symptoms identified exclusively in the GI cluster (nausea, vomiting, appetite loss) at pretreatment were replicated at on- and post-treatment, thus establishing cluster stability across time during treatment.

Table 2 Gastrointestinal symptom cluster structure and symptom prevalence across treatment undertaken by a heterogeneous sample of cancer patients (N = 200)
Full size table

Validation of the identified GI symptom cluster (nausea, vomiting and/or appetite loss)

It was hypothesised that previous chemotherapy experience coupled with higher scores (unweighted means) across symptoms would demonstrate clinical relevance of the GI symptom cluster. Indeed, patients with previous chemotherapy experience reported significantly greater GI symptom distress across time during treatment than those without (31.6 vs. 14.1, p = 0.029) after controlling for covariates (surgery prior to adjuvant chemotherapy ± radiotherapy, variable treatment duration, baseline GI symptom distress scores at pretreatment).

Impact on patients’ quality of life and psychological adjustment of nausea, vomiting and appetite loss alone and as part of a GI symptom cluster

Patients with nausea experienced significantly greater QoL impairment/psychological distress overall across treatment (pre-, on-, post-treatment) than those unaffected in the areas of physical, role and social functioning, fatigue (39.6 vs. 31.0, p = 0.003), appetite loss, overall physical health and overall QoL (0.003 ≤ p ≤ 0.048) after controlling for covariates. Results also approached significance (i.e. p ≈ 0.05–0.06) for sleep disturbance (p = 0.056) and premorbid (pretreatment) neuroticism (EPQ-R Neuroticism SF; p = 0.052). Comparably, patients with vomiting experienced significantly greater QoL impairment/psychological distress overall across treatment in physical, role and social functioning, fatigue, appetite loss, sleep disturbance, overall physical health and cancer distress (IES-IS; 0.001 ≤ p ≤ 0.034) than those unaffected. Results approached significance for cognitive functioning (p = 0.062) and overall QoL (p = 0.052) also. Patients with appetite loss, however, experienced significantly greater QoL impairment overall across treatment in physical, role and social functioning, fatigue, nausea/vomiting and overall physical health (0.001 ≤ p ≤ 0.049) than those unaffected, while results approached significance for cognitive functioning (p = 0.054) and sleep disturbance (p = 0.056; Tables S1–S3 of the “Electronic supplementary material”).

Turning to the impact of multiple symptoms, patients with nausea and/or vomiting experienced significantly greater QoL impairment/psychological distress overall across treatment in physical, role and social functioning, fatigue, premorbid neuroticism, overall physical health and overall QoL (0.003 ≤ p ≤ 0.038) than those unaffected after controlling for covariates (Table S4 of the “Electronic supplementary material”). More prominently, however, patients with GI cluster symptoms (nausea, vomiting and/or appetite loss) experienced significantly greater QoL impairment overall across treatment than those unaffected in physical, role and social functioning, fatigue, nausea/vomiting (p < 0.001), appetite loss (p < 0.001), overall physical health and overall QoL (0.003 ≤ p ≤ 0.05 otherwise) after controlling for covariates (Table 3). Results approached significance for cancer distress (p = 0.058) also.

Table 3 Cancer patients’ quality of life and psychological functioning across treatment in relation to nausea, vomiting and/or appetite loss as part of a gastrointestinal symptom cluster, after controlling for demographics, clinical variables and pretreatment (baseline) scores
Full size table

Impact of nausea, vomiting and appetite loss as part of a GI symptom cluster on patients’ overall quality of life

Two-stage hierarchical multiple regression analysis examining the influence of the GI symptom cluster on patients’ overall QoL at each treatment time point was performed (Table 4). Forced entry controlling for confounding variables or covariates (demographics; clinical characteristics, such as disease stage; pre-/on-treatment QoL) identified in previous studies and those inherent in the present study design (e.g. variable treatment duration) [7, 34, 35] plus stepwise elimination of exploratory predictors contained in the GI cluster (i.e. absence vs. presence of nausea, vomiting and appetite loss, respectively) was employed. At pretreatment, 12.0 % of the total variance in patients’ overall QoL scores was explained by sex, age, preexisting comorbid conditions, prior cancer history, disease stage, time since diagnosis, previous chemotherapy experience, surgery prior to adjuvant chemotherapy ± radiotherapy (7.4 % cumulatively by confounding variables) and appetite loss (4.6 % alone). Age, disease stage and appetite loss (0.003 ≤ p ≤ 0.043) exhibited significant independent effects on overall QoL at pretreatment. After controlling for confounding variables, appetite loss accounted for a significant increase in explained variance of 4.6 % in overall QoL (p = 0.003), while nausea and appetite loss individually did not significantly influence overall QoL at pretreatment (p > 0.1) and were eliminated in the stepwise procedure.

Table 4 The impact of nausea, vomiting and appetite loss as part of a GI symptom cluster on patients’ overall QoL scores at each treatment time point after controlling for demographics, clinical characteristics and overall QoL at pre-/on-treatment in hierarchical multiple regression analyses
Full size table

At on-treatment, 32.3 % of the total variance in overall QoL was explained by confounding variables (27.7 % cumulatively) and vomiting (4.6 % alone). Age, preexisting comorbidities, previous chemotherapy experience, overall QoL at pretreatment and vomiting (0.001 ≤ p ≤ 0.03) exhibited significant independent effects on overall QoL at on-treatment. After controlling for confounding variables (including baseline QoL at pretreatment), vomiting accounted for a significant increase in explained variance of 4.6 % in the change in overall QoL (p = 0.021), while nausea and appetite loss individually did not significantly influence overall QoL at on-treatment relative to pretreatment (p > 0.1).

Finally, 47.7 % of the total variance in overall QoL at post-treatment was explained by confounding variables (41.9 % cumulatively) and nausea (5.8 % alone). Overall QoL at on-treatment (p < 0.001) and nausea (p = 0.008) exhibited significant independent effects on overall QoL at post-treatment. After controlling for confounding variables (including QoL at pretreatment and on-treatment), nausea accounted for a significant increase in explained variance of 5.8 % in overall QoL (p = 0.008), while vomiting and appetite loss individually did not significantly influence overall QoL at post-treatment relative to pretreatment or on-treatment (p > 0.1).

A LMM analysis (multiple regression with repeated measures) was performed to better examine the effect of GI cluster symptoms on overall QoL across time during treatment (Table 5). Controlling for other covariates (demographics, clinical characteristics), overall QoL impairment at the end of treatment (post-treatment) was significantly predicted by younger age (p = 0.028), previous chemotherapy experience (p = 0.019) and overall QoL impairment before treatment (pretreatment; p < 0.001). Additionally, the co-occurrence of nausea, vomiting and appetite loss significantly predicted overall QoL impairment at the end of treatment (post-treatment; p = 0.002, b = −4.53), and had a stronger synergistic impact than nausea + vomiting (p = 0.003, b = −3.62), nausea + appetite loss (p = 0.004, b = −4.04) or nausea alone (p = 0.002, b = −3.67).

Table 5 The contribution of nausea, vomiting and appetite loss as part of a GI symptom cluster to overall QoL impairment across treatment after adjusting for demographics, clinical variables and overall QoL at pretreatment (baseline) in a linear mixed model analysis (N = 200)
Full size table

Discussion

This study demonstrated that cancer patients experience multiple concurrent symptoms during treatment that interact to exert synergistic effects on patient outcomes such as QoL, in a manner that is different from that of individual symptoms alone. From a clinical perspective, recognition of a consistent symptom cluster involving nausea, vomiting and appetite loss is important given the great difficulties clinicians have experienced in treating nausea per se, and may suggest therapeutic strategies not previously considered.

Statistical validity and reliability of a GI symptom cluster containing nausea, vomiting and appetite loss

Patients with mixed diagnoses exhibited a GI cluster consisting exclusively of nausea, vomiting and appetite loss before, during and at the end of combined modality treatment. Unsurprisingly, patients who experience TIN+/−V tend to develop appetite loss due to the negative association between food intake and TIN+/−V [36, 37], and patients in the present study were no exception.

GI symptom clusters have commonly been identified in studies of both homogeneous and heterogeneous cancer patients, often including N&V alone (e.g. [38–40]) or with appetite loss, taste alteration, diarrhoea and/or bloating (e.g. [21, 40–42]). Nevertheless, only a few studies have found a GI symptom cluster comprising solely of nausea, vomiting and appetite loss as revealed in this study [38, 40–42], and none have demonstrated symptom stability over time. Outcomes of this study suggest, perhaps convincingly for the first time, that a stable GI symptom cluster may exist independently of cancer diagnosis, disease stage, treatment type/stage and other demographic/clinical factors (e.g. age, gender, preexisting comorbidities), as has been reported elsewhere for other symptom clusters (e.g. sickness behaviour cluster) [38, 39, 42]. Nausea, vomiting and appetite loss were identified in a GI cluster with remarkable consistency across the treatment trajectory (pre-, on- and post-treatment), accounting for approximately one tenth of the explained variance in QoL and demonstrating acceptable internal consistency (Cronbach’s alpha range: 0.73–0.74) at each time point. A comparable study involving 143 mixed cancer patients identified a relatively stable GI cluster over 12 months following diagnosis, but symptoms in the cluster exhibited some variability across time (N&V plus one to two other transient symptoms) and unsatisfactory internal consistency (Cronbach’s alpha < 0.7) on the majority of assessments [21]. The variability in GI cluster symptoms observed relative to the present study may have been due to several methodological differences (e.g. eligibility criteria, measures used, timing of assessments, chemotherapy emetogenicity, treatment duration, antiemetics prescribed, criteria used for statistical analyses) [21, 31].

Clinical validity of a GI symptom cluster containing nausea, vomiting and appetite loss

Statistically determined clusters may identify symptom clusters that are overlooked in clinical assessment, but are of little use to clinicians (and patients) unless they are shown to be clinically relevant. Few studies revealing statistically determined symptom clusters in cancer patients have attempted to demonstrate clinical significance also [43]. Clinical significance in this study was established by substantiating the hypothesis that previous chemotherapy experience would be coupled with higher scores on GI cluster symptoms (nausea, vomiting and/or appetite loss) across treatment. Similar results were also found in a cross-sectional study [39], where support was found for the hypothesis that advanced cancer patients receiving chemotherapy would have higher scores on GI cluster symptoms (nausea, vomiting) than those who were not. These findings suggest that higher GI symptom cluster distress may negatively impact on QoL and other patient outcomes [44].

Implications of a GI cluster (nausea, vomiting, appetite loss) and its individual symptoms for patients’ quality of life and psychological adjustment

Clinical relevance is best demonstrated though when the presence of a symptom cluster has an impact on patient outcomes, such as QoL, psychological adjustment or survival [43, 44]. Study results suggest that the GI cluster (nausea, vomiting and/or appetite loss) had a negative impact on QoL, and more so than nausea, vomiting and appetite loss individually or nausea and/or vomiting. As expected, patients who reported the co-occurrence of nausea, vomiting and appetite loss generally experienced greater QoL impairment. Patients with these symptoms experienced worse overall QoL, overall physical health and physical, role and social functioning across treatment than those unaffected, after adjusting for differences in clinical characteristics and baseline QoL/psychological functioning between patients in univariate analyses. Greater fatigue and cancer distress were also experienced across treatment by these patients, as well as higher levels of premorbid neuroticism.

Similar outcomes have been observed previously, albeit in less expansive studies examining the impact of acute and/or delayed CINV per se on QoL [10, 13, 36]. In the most comparable study, Osoba and colleagues [36] found in an antiemetic trial of 832 mixed cancer patients that patients with delayed CINV experienced significantly worse overall QoL, physical, cognitive and social functioning, fatigue, appetite loss and sleep disturbance after one cycle of M/HEC compared to those unaffected. Additionally, in an observational study of 151 mixed cancer patients, Cohen and colleagues [10] also found that CINV had cumulative effects on QoL and suggested that the experience of CINV in earlier cycles of chemotherapy affected QoL in subsequent chemotherapy cycles.

In the current study, approximately one tenth to over one third of the explained variance in overall QoL was accounted for by the GI cluster symptoms of nausea, vomiting or appetite loss at any of the assessed time points across treatment, after adjustment for demographics, clinical characteristics and pre-/on-treatment QoL in multivariate analyses. Of the GI symptoms, appetite loss was unsurprisingly the best independent predictor of QoL impairment at pretreatment in the absence of chemotherapy and radiotherapy (cf. N&V versus other GI symptoms during treatment as predictors of patient outcomes including symptom distress, which negatively affects QoL) [12, 21, 35, 44]. With the introduction of combined treatment, however, vomiting took precedence in the early stages (on-treatment), but by the end of treatment (post-treatment) nausea had assumed greatest importance in explaining the change in QoL for cancer patients across treatment. Consistent with previous studies, these results suggest that nausea (particularly delayed nausea over successive cycles of chemotherapy) is a more pervasive problem than vomiting or appetite loss across treatment, irrespective of antiemetic therapy, and has a significant impact on patients’ QoL [1, 6, 8, 10, 18, 23, 45]. Results also suggest that, although related, nausea, vomiting and appetite loss do not appear to share a common underlying mechanism, which would explain the minimal impact of antiemetics on nausea and, to a lesser extent, appetite loss for patients in this study. Large multicentre studies are needed to further confirm these results, however, and to determine the interactive/causal nature of the relationships among these symptoms in influencing patients’ QoL.

It was notable that nausea had the greatest impact of all the GI cluster symptoms on the QoL and psychological adjustment of cancer patients across treatment. Patients with nausea experienced worse overall QoL, overall physical health and physical, role and social functioning across treatment than those unaffected after adjusting for other variables. Greater fatigue, appetite loss and sleep disturbance were also experienced across treatment by these patients, as well as higher levels of premorbid neuroticism. Nevertheless, nausea was not merely additive in its impact on QoL but occurred with vomiting and appetite loss as a cluster of GI symptoms to exert a synergistic effect of multiplicative proportions on QoL. Indeed, results of the LMM analysis demonstrated that the GI symptom cluster of nausea, vomiting and appetite loss had a stronger negative impact on overall QoL across treatment than individual symptoms alone or any symptom pair within the cluster. To our knowledge, no longitudinal study has examined the synergistic effects of a GI cluster comprising N&V in this manner, although the results of two smaller studies employing cross-sectional analyses have been broadly consistent [21, 37]. Nonetheless, further research is needed to confirm these exploratory results.

Limitations

Some limitations of this study must be acknowledged. Heterogeneity of the sample in terms of demographics and clinical characteristics may limit the generalisability of results. However, all analyses (except EFA) controlled for many potential confounding variables. Moreover, it is important to identify symptom clusters that are applicable to a variety of cancer patients, rather than being disease- or treatment-specific, as this reflects routine clinical practice. Finally, symptom assessment was limited to individual items from the EORTC QLQ-C30, which is designed to measure QoL. Ideally, more comprehensive measures (e.g. patient diaries for GI symptoms, EORTC QLQ-C30 + cancer-specific modules) would have been utilised and may have resulted in the identification of a different GI symptom cluster [37, 39]. Multicentre studies employing larger samples (involving both homogeneous and heterogeneous patients), prospective longitudinal and cross-sectional designs and more rigorous measures are therefore required.

Clinical implications

Regardless of aetiology, uncontrolled N&V often results in significant appetite and weight loss during chemotherapy, leading to prolonged recovery between cycles and increased risk for anticipatory N&V and malnutrition [45–48]. Malnutrition and weight loss, in turn, are associated with poorer prognosis, treatment tolerance and response, performance status, QoL and survival [45–48]. Combined with the findings of the present study, we recommend a symptom cluster or multiple symptom intervention approach involving a central focus on nausea and vomiting as core symptoms, but featuring the adaptability of a modular approach in which supplementary strategies may be appended as needed to target associated cluster symptoms (e.g. appetite/weight loss, GI reflux). In particular, clinicians should genuinely consider combining essential antiemetic therapies with other evidence-based pharmacological (e.g. delayed/acute nausea: psychotropics, such as olanzapine, lorazepam and mirtazapine; dopamine antagonists) [49–51] and non-pharmacological approaches (e.g. N&V: relaxation techniques, music, hypnotherapy, acupressure) [52–55]. This multifaceted intervention would not only improve prevention and control of N&V for their patients, but also reduce the synergistic effects of cluster symptoms (e.g. N&V, appetite loss) as a whole and improve QoL outcomes. Where nausea and/or vomiting and, by extension, their associated symptoms are not adequately controlled by these antiemetic-based interventions, clinicians must then contemplate supplementing evidence-based strategies that specifically target the associated symptoms (for instance, appetite/weight loss: pharmacological approaches—e.g. progestins ± olanzapine, corticosteroids; non-pharmacological approaches—e.g. nutritional counselling) [56–58].

Control using this approach may be sub-optimal, however, unless greater recognition, communication/understanding and assessment of nausea, vomiting and associated symptoms, including their effects, are undertaken by both clinicians and patients [8, 18]. Patients (and their caregivers) require support and education about TIN+/−V and associated symptoms (present or not) [19], particularly regarding the insidious manifestation and effects of nausea per se and delayed symptoms that occur outside treatment settings. Patients also need instruction on common-sense self-management strategies (e.g. nausea: basic dietary habits, such as avoiding fried/fatty foods) to consolidate the support and education provided [48, 59].