It is not known whether symptoms cluster together after esophageal cancer surgery or whether such symptom clusters are associated with survival in patients with esophageal cancer who are treated surgically.
It is not known whether symptoms cluster together after esophageal cancer surgery or whether such symptom clusters are associated with survival in patients with esophageal cancer who are treated surgically.
Data from a prospective Swedish nationwide cohort study of surgically treated patients with esophageal cancer recruited between 2001 and 2005 were used. General and esophageal cancer-specific symptoms were assessed using the European Organization for Research and Treatment of Cancer QLQ-C30 quality of life questionnaire and the QLQ-OES18 module at 6 months after surgery. Associations between symptom clusters and survival were analyzed using Cox proportional hazards models, providing hazards ratios with 95% confidence intervals, adjusted for other known prognostic factors.
Among 402 patients reporting symptoms 6 months after surgery, 3 symptom clusters were identified. The first symptom cluster (“fatigue/pain”) was characterized by symptoms of pain, fatigue, insomnia, and dyspnea and was present in 30% of patients. The second symptom cluster (“reflux/cough”) was characterized by symptoms of dry mouth, problems with taste, coughing, and reflux and was present in 27% of patients. The third symptom cluster (“eating difficulties”) was characterized by appetite loss, dysphagia, eating difficulties, and nausea/vomiting and was present in 28% of patients. The presence of the reflux/cough and eating difficulties symptom clusters was associated with a statistically significantly increased risk of mortality (adjusted hazards ratio, 1.43 [95% confidence interval, 1.08-1.89] and adjusted HR, 1.41 [95% confidence interval, 1.06-1.87], respectively).
Symptoms experienced by surgically treated patients with esophageal cancer appear to cluster together, and the presence of these symptom clusters appears to have strong prognostic value. Cancer 2014;120:286–293. © 2013 American Cancer Society.
Esophageal cancer is an aggressive disease associated with an extremely poor prognosis. The overall 5-year survival rate is reported to be as low as 10% in Europe. Surgical resection is the most established potentially curative treatment. The prognosis is unfavorable even when the tumor is surgically removed, with a postoperative 5-year survival rate of 31%. The operation is more extensive than most other standard surgical procedures and is associated with a substantial negative impact on patients' health-related quality of life (HRQOL),[4-6] with patients experiencing persistent problems with physical function and specific symptoms even at 3 years postoperatively.[7, 8] Although the majority of patients recover in terms of HRQOL functions and symptoms, and report scores similar to those of the background population at 5 years, a substantial minority of long-term survivors appear to deteriorate further over time.
Cancer patients often experience multiple concurrent symptoms as a result of their disease and treatment, and these tend to cluster together in systematic ways. For example, symptoms of fatigue, insomnia, and depression are often experienced simultaneously.[10, 11] In the past decade, interest in the impact of symptom clusters on patient outcomes has increased, and symptom clusters may have a negative impact on survival in patients with advanced cancer.[12, 13] Patients with esophageal cancer experience numerous symptoms after surgical treatment and previous studies have evaluated the impact of individual HRQOL functions and symptoms on patient outcomes. Several aspects have been identified as prognostic indicators for survival in this group of patients with cancer. Better pretreatment physical and role functioning as well as global quality of life (QOL) are associated with better survival. Other studies have demonstrated an association between poor pretreatment appetite loss,[15-17] fatigue,[18-20] dyspnea,[20, 21] reflux, and an increased mortality risk. Moreover, poor postoperative global QOL, physical function, and social function, as well as symptoms of fatigue, pain, dyspnea, appetite loss, dysphagia, and esophageal pain have been shown to predict mortality independently of established prognostic factors.
Although studies indicate a prognostic role of individual HRQOL functions and symptoms, to our knowledge it is not known whether postsurgery symptoms cluster together in a particular manner, or what the implications are of the presence of any such symptom clusters on patient outcomes. Therefore, we aimed to establish whether the symptoms experienced by patients with esophageal cancer 6 months after surgery cluster together, and whether the presence of such symptom clusters is predictive of survival.
A Swedish nationwide, prospective, cohort study of patients with esophageal cancer was conducted between April 2, 2001 and December 31, 2005. All patients in Sweden diagnosed with esophageal or esophagogastric junction cancer who underwent curatively intended surgery during the study period were eligible for inclusion. The cohort included 90% of all surgically treated patients in Sweden during this time period. Patients were followed with measures of HRQOL 6 months after surgery. Ethical approval was granted by the Regional Ethical Review Board in Stockholm, Sweden, and informed consent was obtained from all patients.
The prospective data collection has been described in full elsewhere.[23, 24] In brief, data were collected via a nationwide network of hospital departments and physicians involved in the diagnosis and treatment of patients with esophageal or esophagogastric junction cancer in Sweden, and registered in the Swedish Esophageal and Cardia Cancer database. Based on a predefined study protocol, data regarding patient and tumor characteristics as well as information regarding hospital, surgical treatment, and complications were collected prospectively through manual review of medical records. Tumor stage was classified according to the Union Internationale Contre le Cancer. Information regarding mortality was obtained via linkage to the 100% complete Swedish Register of the Total Population .
Several sociodemographic and clinical characteristics were assessed in the study and categorized as follows: age (< 60 years, 60-74 years, or ≥ 75 years), sex (male or female), tumor stage (0-I, II, or III-IV), tumor histology (squamous cell carcinoma or adenocarcinoma), type of surgery (esophageal resection, cardia resection, extended total gastrectomy, or total gastrectomy and esophageal resection), number of postoperative complications (0, 1, or > 1), and number of comorbidities (0 or ≥ 1). Information regarding these characteristics was gathered from medical records. Postoperative complications were predefined and included perioperative bleeding, leakage or perforation of the anastomosis, stricture of the anastomosis, wound infection needing treatment, wound rupture, gastric perforation, intraabdominal or intrathoracic abscess, ileus, thoracic duct injury, reintubation or need for prolonged mechanical ventilation, paresis of the recurrent laryngeal nerve, pneumonia, pulmonary thrombosis, and myocardial infarction. Comorbidity included angina, heart failure, hypertension, diabetes, pulmonary disease, renal failure, liver failure, and other cancer.
To assess symptoms, 2 HRQOL questionnaires were used in the study. The European Organization for Research and Treatment of Cancer (EORTC) QLQ-C30 was used to assess symptoms common among cancer patients in general, whereas the esophageal-specific module QLQ-OES18 was used to measure symptoms common among patients with esophageal cancer in particular. The QLQ-C30 incorporates 5 functional scales; 1 global QOL scale; 1 item regarding financial difficulties; 3 symptom scales assessing pain, fatigue, and nausea/vomiting; and 5 single items addressing symptoms of dyspnea, sleeping difficulties, appetite loss, constipation, and diarrhea. The EORTC QLQ-OES18 contains 4 symptom scales addressing symptoms of dysphagia, eating difficulties, reflux, and esophageal pain, and 6 single items assessing trouble swallowing saliva; choking when swallowing; dry mouth; and trouble with taste, coughing, or talking. The 3 symptom scales and 5 single symptom items from the QLQ-C30 and all symptom scales and single items from the QLQ-OES18 were included in the analyses. Responses to the functional scales, global QOL, and the financial difficulties question were not included in the current analyses because the aim of the current study was to cluster symptoms not functions. Items were scored on a 4-point scale: 1) “not at all”; 2) “a little”; 3) “quite a bit”; and 4) “very much,” referring to symptoms experienced during the past week. For the purposes of the current analyses, symptoms were recoded as binary variables to represent the “presence” (for a response of 3 or 4) or “absence” (for a response of 1 or 2) of symptoms. This cutoff value was based on previous work and clinical experience, representing a level of symptoms that would likely be noticed by clinicians in practice.[6, 28]
The symptom clustering was performed in 2 steps. In the first step, symptoms were clustered using the principal component analysis (PCA) method. To address the sensitivity of the final clusters to selected clustering methods, the hierarchical cluster analysis (HCA) method was also used to compare the final clusters in this first step.[29, 30] These methods (PCA and HCA) group symptoms in clusters so that the symptoms are highly correlated within their own cluster and have very low correlations with symptoms in other clusters. The splitting criterion with a maximum eigenvalue of 1 was used to select the final cluster solution from the PCA method. Moreover, for the HCA method, the lowest maximum 1-R2 ratio value for a variable was used as the measure of clustering fit and to select the final cluster solution. Among 18 symptom scales and single items, 14 symptoms with at least 15% prevalence in the sample were selected for clustering (including symptoms of pain, fatigue, nausea/vomiting, dyspnea, sleeping difficulties, appetite loss, diarrhea, dysphagia, eating difficulties, reflux, esophageal pain, dry mouth, trouble with taste, cough). In the second step, patients were assigned cluster membership if they experienced at least 50% of symptoms from within a symptom cluster (ie, symptom cluster present). Hence a patient could belong to more than 1 cluster.
The chi-square test was used to examine within cluster differences on baseline sociodemographic and clinical characteristics. Multivariable logistic regression analysis, providing adjusted odds ratios (OR) with 95% confidence intervals (95% CI), was used to assess the baseline sociodemographic and clinical characteristics of final clusters. The symptom clusters were analyzed in relation to survival by the Kaplan-Meier method. Statistical significance was tested at the 5% level by the log-rank test. A Cox proportional hazards model, providing adjusted hazards ratios (HR) with 95% CI, was used to analyze the 5-year survival of final clusters, controlling for confounding by baseline sociodemographic and clinical characteristics (using the categorizations of variables as described in the measures section above). The logistic regression model was used to assess the discriminative power of these clusters using the area under the receiver operating characteristic curve. Due to the potential overlapping cluster membership, interactions between clusters were also tested at the 5% level of significance using the Wald chi-square test. The PROC VARCLUS statement (SAS 9.2; SAS Institute Inc, Cary, NC) was used for the PCA and HCA (with centroid option) clustering methods and SAS 9.2 software was used for other data analyses.
During the study period, 616 patients were included in the cohort. Of these, 110 (18%) patients died within 6 months of undergoing surgery. Therefore, 506 patients were eligible for the current study; of these, 402 patients (79%) responded to the 6-month follow-up symptom assessment. There were no statistically significant differences noted between responders and nonresponders with regard to any of the baseline clinical or sociodemographic characteristics. Baseline characteristics of the participating patients are detailed in Table 1. Based on the total group (n = 402), patients were predominantly male (81%) and aged between 60 years and 74 years (58%). The majority of patients were diagnosed with adenocarcinoma of the esophagus (76%) of tumor stage III (41%), and had undergone esophageal resection (76%). At least 1 complication was recorded in 133 patients (33%), and the majority of patients (59%) had ≥ 1 comorbidity.
|Characteristic, No. (%)||Total (n = 402)||Present (n = 122)||Absent (n = 280)||Present (n = 107)||Absent (n = 295)||Present (n = 112)||Absent (n = 290)|
|<60||101 (25)||40 (33)||61 (22)||19 (18)||82 (28)||28 (25)||73 (25)|
|60–74||232 (58)||60 (49)||172 (61)||71 (66)||161 (55)||64 (57)||168 (58)|
|>74||69 (17)||22 (18)||47 (17)||17 (16)||52 (18)||20 (18)||49 (17)|
|Female||75 (19)||24 (20)||51 (18)||24 (22)||51 (17)||23 (21)||52 (18)|
|Male||327 (81)||98 (80)||229 (82)||83 (78)||244 (83)||89 (79)||238 (82)|
|0-I||84 (21)||17 (14)||67 (24)||19 (18)||65 (22)||15 (13)||69 (24)|
|II||120 (30)||32 (27)||88 (32)||31 (29)||89 (31)||29 (26)||91 (32)|
|III||161 (41)||60 (50)||101 (36)||51 (48)||110 (38)||59 (53)||102 (36)|
|IV||33 (8)||11 (9)||22 (8)||6 (6)||27 (9)||9 (8)||24 (8)|
|Squamous cell carcinoma||98 (24)||42 (34)||56 (20)||38 (36)||60 (20)||34 (30)||64 (22)|
|Adenocarcinoma||304 (76)||80 (66)||224 (80)||69 (64)||235 (80)||78 (70)||226 (78)|
|Type of surgeryb|
|Esophageal resection||302 (76)||96 (79)||206 (74)||87 (81)||215 (73)||89 (79)||213 (74)|
|Cardia resection||17 (4)||3 (2)||14 (5)||2 (2)||15 (5)||1 (1)||16 (6)|
|Extended total gastrectomy||41 (10)||11 (9)||30 (11)||10 (9)||31 (11)||12 (11)||29 (10)|
|Total gastrectomy and esophageal resection||40 (10)||12 (10)||28 (10)||8 (7)||32 (11)||10 (9)||30 (10)|
|No. of complications|
|0||269 (67)||67 (55)||202 (72)||59 (55)||210 (71)||63 (56)||206 (71)|
|1||89 (22)||35 (29)||54 (19)||32 (30)||57 (19)||31 (28)||58 (20)|
|>1||44 (11)||29 (16)||24 (9)||16 (15)||28 (9)||18 (16)||26 (9)|
|No. of comorbidities|
|0||164 (41)||45 (37)||119 (43)||42 (39)||122 (41)||44 (39)||120 (41)|
|≥1||238 (59)||77 (63)||161 (57)||65 (61)||173 (59)||68 (61)||170 (59)|
The final cluster solutions from PCA and HCA were identical with the exception of the symptom of diarrhea, which formed a single symptom cluster when using the HCA method. All other symptoms clustered into identical clusters produced by both methods. Therefore, we selected the 3-cluster solution using the PCA method.
Baseline characteristics of the 3 symptom clusters are presented in Table 1. The first cluster was characterized by the symptoms of pain (general and esophageal), fatigue, insomnia, and dyspnea. This cluster was named the “fatigue/pain” (FP) symptom cluster. Among the 402 patients who responded to the symptom assessment at 6 months after surgery, this cluster was present (ie, meeting the criteria for cluster membership by reporting at least 50% of symptoms in the cluster) in 122 patients (30%) and absent in 280 patients (70%). The most prevalent symptom in this cluster was fatigue (97%). Cluster 2 was characterized by symptoms of dry mouth, problems with taste, coughing, and reflux. This cluster was named the “reflux/cough” (RC) symptom cluster. This symptom cluster was present in 107 patients (27%) and absent in 295 patients (73%). The prevalence of each of the 4 symptoms within this cluster was approximately 60%. The third cluster was defined predominantly by symptoms related to eating, such as appetite loss, dysphagia, eating difficulties and nausea/vomiting, and was named the “eating difficulties” (ED) symptom cluster. The ED cluster was present in 112 patients (28%) and absent in 290 patients (72%). The most prevalent symptoms in this cluster were eating difficulties (95%) and appetite loss (84%). The composition of the symptom clusters is presented in Table 2.
|Symptom, No. (%)||Fatigue/Pain Symptom Cluster (n = 122; 30%)||Symptom, No. (%)||Reflux/Cough Symptom Cluster (n = 107; 27%)||Symptom, No. (%)||Eating Difficulties Symptom Cluster (n = 112; 28%)|
|Fatigue||118 (97)||Reflux||68 (64)||Eating difficulties||106 (95)|
|Esophageal pain||96 (79)||Coughing||65 (61)||Appetite loss||94 (84)|
|Pain||94 (77)||Trouble with taste||65 (61)||Nausea/vomiting||71 (63)|
|Dyspnea||82 (67)||Dry mouth||65 (61)||Dysphagia||69 (62)|
|Insomnia||64 (52)||Diarrhea||59 (53)|
Univariable differences in baseline characteristics within the clusters were assessed with chi-square analyses. Younger age, higher tumor stage, squamous cell carcinoma, and more complications were associated with the presence of the FP symptom cluster (all at P < .05). Squamous cell carcinoma and more complications were associated with the presence of the RC symptom cluster, whereas with a higher tumor stage and more complications were associated with the presence of the ED symptom cluster (all at P < .05). No other statistically significant univariable differences were observed within the clusters. Subsequent multinomial logistic regression analysis, adjusting for all other baseline characteristics, demonstrated that compared with patients with squamous cell carcinoma, those with adenocarcinoma were less likely to be assigned cluster membership to the FP symptom cluster (adjusted OR, 0.49; 95% CI, 0.28-0.84) and the RC symptom cluster (adjusted OR, 0.54; 95% CI, 0.32-0.94). Compared with having > 1 complication, patients with no complications were less likely to be assigned cluster membership to the FP symptom cluster (adjusted OR, 0.42; 95% CI, 0.21-0.84) and the ED symptom cluster (adjusted OR, 0.46; 95% CI, 0.23-0.93).
The median survival times for the FP, RC, and ED symptom clusters were 18.13 months, 21.16 months, and 21.17 months, respectively. Of the 402 patients responding to the 6-month follow-up symptom assessment, 124 (31%) were alive at 5 years. Figure 1 shows the Kaplan-Meier estimates of survival, illustrating the significantly reduced 5-year survival estimates (shown in months) for the presence of symptom clusters (P < .001, log-rank test). Cox proportional hazards modeling demonstrated that symptom cluster was a significant predictor of mortality (Table 3). After adjustment for age, sex, tumor stage, tumor histology, type of surgery, postoperative complications, and number of comorbidities, the presence of the RC and ED symptom clusters were associated with a statistically significantly increased risk of mortality (HR, 1.43; 95% CI, 1.08-1.89 and HR, 1.41; 95% CI, 1.06-1.87, respectively). Moreover, the area under the receiver operating characteristic curve was 0.79, which indicates that the clusters discriminate well between 5-year survivors and nonsurvivors. Because patients could be assigned membership to more than 1 cluster (31 patients belonged to all 3 clusters, 65 belonged to both the PF and ED clusters, 48 belonged to both the RC and ED clusters, and 55 patients belonged to both the FP and RC clusters), interactions between the symptom clusters were also tested by introducing product terms into the Cox model (using the Wald chi-square test at the 5% level of significance). No interactions emerged as statistically significant.
|Symptom Cluster||HRa||95% CI||P|
|Fatigue/pain: present vs absent||1.20||0.90–1.60||.21|
|Reflux/cough: present vs absent||1.43||1.08–1.89||.01|
|Eating difficulties: present vs absent||1.41||1.06–1.87||.02|
Using PCA, 3 symptom clusters were identified based on 14 self-reported general and esophageal cancer-specific symptoms 6 months after patients underwent curatively intended surgery. The identified symptom clusters were named fatigue/pain, reflux/cough and eating difficulties based on the composition of symptoms forming each cluster. The presence of symptom clusters at 6 months after surgery was found to be a significant predictor of survival duration, with the RC and ED symptom clusters emerging as statistically significant predictors of 5-year mortality. The presence of these symptom clusters conferred an approximately 40% increased risk of mortality. It is important to note that the multivariable adjusted survival analysis demonstrated that these associations existed over and above what could be explained by other known prognostic factors, suggesting a significant independent prognostic role of self-reported symptom clusters 6 months after esophageal cancer surgery.
The strengths of the current study include the population-based design, the high participation rate and large sample size, the high quality of the clinical data, the use of well-validated questionnaires, and adjustment for other known prognostic factors. To the best of our knowledge, this is the first study to examine symptom clusters among surgically treated esophageal cancer patients. Nevertheless, the interpretation of results must be made cautiously because statistically derived clusters can be affected by several methodological considerations and clinical relevance can be difficult to determine. Symptom clusters may vary according to the assessment of symptoms, the prevalence of symptoms within the sample, sample size, and sample composition. In addition, the symptoms reported by patients were recoded into binary variables to estimate the extent of the presence or absence of a symptom based on the proportion of patients responding “quite a bit” and “very much” (ie, symptom present) versus “not at all” and “a little” (ie, symptom absent) to the symptom items in the questionnaires. Categorizing symptoms as present or absent addresses only 1 dimension of symptom experience (ie, symptom frequency). However, this categorization describing the proportion of patients experiencing a symptom based on responses of “very much” or “quite a bit” to items on the EORTC QLQ-C30 was used previously in symptom clusters research among patients with cancer. Furthermore, it is difficult to determine whether the symptom clusters identified in the current study are stable long-term clusters. However, the assessment of symptoms at 6 months after surgery was based on previous research suggesting that by this point the acute postoperative phase has subsided, and HRQOL and symptom levels have generally returned to a preoperative baseline value in the majority of patients.[8, 32] Nevertheless, it would be of value to test the cluster solution on data obtained at different time points.
There are 2 conceptual approaches to symptom clusters research, one of which involves the identification of clusters in patients and the other the identification of different subgroups of patients based on their similar symptom profiles. In line with the methodology of a previous study of symptom clusters in patients with advanced cancer, we clustered symptoms reported at 6 months; subsequently, the symptom cluster was considered present if patients reported at least 50% of symptoms within the symptom cluster. It is important to note that the use of these arbitrary cutoff values to determine cluster membership may have affected the clusters. However, the clinical importance of the symptom clusters identified in the current study is supported by the finding of an association between the presence of symptom clusters among patients and survival duration, independent of other prognostic factors.
Although to the best of our knowledge symptom clusters have not been assessed previously in this group of patients, findings from previous prognostic studies of individual symptoms and functions resonate well with the results of the current study. We identified a cluster characterized largely by symptoms related to aspects of eating (the ED symptom cluster). Both pretreatment and posttreatment appetite loss[15-17, 22] and dysphagia have previously been identified as independent predictors of increased mortality among surgically treated esophageal cancer patients. Similarly, reflux has been shown to predict mortality among these patients. We also found that the presence of the RC symptom cluster, characterized by reflux, dry mouth, problems with taste, and coughing, significantly predicted mortality. We identified a cluster largely composed of symptoms related to pain and fatigue (the FP symptom cluster) that was associated with an increased risk of mortality (HR, 1.2; 95% CI, 0.9-1.6). Although this association was not found to be statistically significant, it is in keeping with previous studies that have found that the individual symptoms of pain and fatigue predict survival independently of a range of known prognostic factors among these patients (HR range, 1.10-1.65).[18-20, 22]
Due to the poor postoperative prognosis for patients with esophageal cancer, it is of utmost importance to further the understanding of factors related to prognosis. Previous research regarding the prognostic impact of symptoms has focused largely on the adverse impact of individual symptoms on outcomes. However, symptoms seldom occur in isolation, and to the best of our knowledge the current study is the first published to date evaluating the prognostic role of symptom clusters in this patient population. Results from this study have potentially important implications for clinical practice. From a clinical perspective, targeting the symptoms in the ED and RC symptom clusters may be particularly important. It has been suggested that the most prevalent symptom within a symptom cluster may be an indicator of the presence of that particular symptom cluster. Thus, targeting the most prevalent symptom in a symptom cluster may contribute to the streamlined assessment and management of symptoms to improve outcomes.[33, 34] Should these results be confirmed in other studies, the findings may be used in clinical practice to direct additional investigations and symptom management strategies in patients who have undergone surgery for esophageal cancer.
Symptoms experienced by surgically treated patients with esophageal cancer at 6 months after surgery appear to cluster together, and these clusters have strong prognostic value. If confirmed in future studies, these findings may have clinical relevance for the prediction of prognosis and the planning of the follow-up care of these patients after surgery.
The funding for this project was provided by The Swedish Research Council, the Swedish Cancer Society, and the Cancer Society in Stockholm. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the article. The funders do not bear any responsibility for the analyses or interpretations presented herein.
The authors made no disclosures.