Influence of postoperative complications following esophagectomy for cancer on quality of life: A European multicenter study

Introduction: Postoperative complications following major surgery have been shown to be associated with reduced health-related quality of life (HRQL), and severe complications may have profound negative effects. This study aimed to examine whether long-term HRQL differs with the occurrence and severity of complications in a European multicenter prospective dataset of patients following esophagectomy for cancer. Methods: Disease-free patients following esophagectomy for cancer between 2010 and 2016 from the LASER study were included. Patients completed the LASER, EORTC QLQ-C30 and EORTC QLQ-OG25 questionnaires >1 year following treatment. Long-term HRQL was compared between patients with and without postoperative complications, subgroup analysis was performed for severity of complications (no, minor [Clavien-Dindo I-II], severe [Clavien-Dindo III]), using univariable and multivariable regression. Results: 645 patients were included: 283 patients with no, 207 with minor and 155 with severe complications. Significantly more dyspnea (QLQ-C30) was reported by patients with compared to patients without complications (difference in means 6.3). In subgroup analysis, patients with severe complications reported more dyspnea (difference in means 8.3) than patients with no complications. None of the differences were clinically relevant (difference in means 10 points). LASER-based low mood (OR2.3) was statistically different for minor versus severe complications. Conclusion: Comparable HRQL was found in patients with and without postoperative complications following esophagectomy for cancer, after a mean follow-up of 4.4 years. Furthermore, patients with different levels of severity of complications had comparable HRQL. The level of HRQL in esophageal cancer patients are more likely explained by the impact of the complex procedure of the esophagectomy itself. © 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The incidence of esophageal cancer is increasing globally, with over 600.000 esophageal cancer patients diagnosed in 2020 [1].After curative treatment, usually consisting of neoadjuvant chemo(radio)therapy and esophagectomy with lymphadenectomy, the current 5-year survival rate reaches almost 50% [2e6].Postoperative complications after esophageal cancer surgery remain a major concern, as in over 50% of the patients complications occur, of which over 17% of patients have severe complications, requiring surgical intervention under general anaesthesia (Clavien-Dindo !IIIb) [7].Postoperative complications are associated with increased postoperative mortality and decreased long-term survival [8e10].Studies investigating the relation between the occurrence of complications following major surgery and healthrelated quality of life (HRQL) show that complications are negatively associated with short-and long-term HRQL [10e16].A multicenter longitudinal study with a total of 785 patients following major elective gastrointestinal, cardiothoracic, or vascular surgery found severe complications to be associated with significantly impaired short-term HRQL [11].Several other studies also found postoperative complications following an esophagectomy to be negatively associated with HRQL [10,14,17,18].One of these studies found that severe complications are associated with a long-lasting poor HRQL.Patients who endured major postoperative complications reported significantly more problems with dyspnea and fatigue compared to patients without any major postoperative complications [10].However in a recent populationbased study including 486 Dutch patients after an esophagectomy, no significant differences were found in short-and long-term HRQL between patients with and without complications [19].None of these studies analysed HRQL according to the severity of postoperative complications, using a complication grading system, such as the Clavien-Dindo classification in an international group of patients [20,21].
Therefore the primary aim of this study was to examine whether long-term HRQL differs with the occurrence and severity of postoperative complications in a European multicenter prospective dataset of esophagectomy patients (LASER study) [22].We hypothesized that the severity of postoperative complications are proportionally negatively associated with long-term HRQL: patients with severe complications are likely to have significantly worse long-term HRQL compared to patients with no or minor postoperative complications.

The Lasting Symptoms After Esophageal Resection (LASER) study database
The LASER study is a prospective European multicenter crosssectional study, which aimed to identify symptoms with the highest incidence and the greatest impact on HRQL in disease-free patients following an esophagectomy for esophageal or gastroesophageal junction cancer [22].The primary study outcomes of the LASER study have been published before [22].Ethical approval to use the LASER study dataset for side studies were gained by each participating centre during the original LASER study, and written informed consent was signed by all participants to use their data in retrospective studies.
The STROBE guidelines were followed to ensure the correct structure in this article [23].
This current multicenter population-based prospective comparative cohort study was performed with the data from the Lasting Symptoms After Esophageal Resection (LASER) study database [22].Patients who underwent a curative intent esophagectomy in one of the 20 participating centers across Europe, between January 1, 2010 and June 30, 2016, and were disease-free and at least 1 year post completion of treatment were eligible for inclusion and invited to participate.Eligible patients completed three questionnaires, solely once (The European Organization for Research and Treatment of Cancer [EORTC] QLQ-C30 and EORTC QLQ-OG25 HRQL questionnaires and the LASER symptom questionnaire) [22,24].Exclusion criteria for this study were missing clinical and treatment data, and missing Clavien-Dindo categorization.
Complications were recorded according to the ECCG criteria (Supplementary Table S1) and graded using the Clavien-Dindo grading system [20,21,25].Patients were first divided into subgroups according to the presence of post-operative complications, i.e., those with or without postoperative complications.They were secondly divided according to the severity of post-operative complications, i.e., those with no, minor (Clavien-Dindo grade I-II) or severe (Clavien-Dindo grade !III postoperative complications) [7,20,21].

EORTC quality of life questionnaires
The EORCT QLQ-C30 is validated for all cancer patients and consist of 30 questions [24].Response options range from 1 to 7 ('very poor'-'excellent') for two questions and from 1 to 4 ('not at all', 'a little', 'quite a bit' and 'very much') in the remaining 28 questions.These items are combined to form 15 outcomes as follows: six multiple-item scales (global health, physical, role, emotional, cognitive and social functioning), three symptom scales (fatigue, nausea and vomiting, and pain), and five additional single items assessing symptoms often reported by cancer patients (dyspnea, insomnia, appetite loss, constipation and diarrhea) and finally financial difficulties.
The EORTC QLQ-OG25 is recommended to supplement the QLQ-C30 when assessing the HRQL in patients with esophageal, junctional or gastric cancer.It includes 25 questions with answers ranging from 1 to 4 ('not at all', 'a little', 'quite a bit' and 'very much') [26].From this questionnaire, a total of 16 HRQL domains are assessed.The domains are subdivided into six multi-item scales (dysphagia, eating restrictions, reflux, odynophagia, pain and discomfort, and anxiety) and 10 single-items (eating with others, dry mouth, trouble with taste, body image, trouble swallowing saliva, choked when swallowing, trouble with coughing, trouble talking, weight loss and hair loss).
The 31 HRQL domain scores derived from both questionnaires are linearly transformed into scores ranging from 0 to 100.Higher mean scores in global health and functioning domains represent better HRQL.In symptom domains, a higher mean score represents more symptomatology.
The LASER questionnaire was developed by the LASER study group, to appoint symptoms associated with esophagectomy experienced in the past 6 months [22] among other things 28 symptoms associated with esophagectomy.These LASER symptoms are graded according the impact upon quality of life ('none', 'some', and 'substantial') and frequency of symptoms, scoring from 0 to 5 ('never', 'rarely', 'weekly', 'daily' and 'multiple times per day').In the LASER study, the following three LASER symptoms were found to be associated with poor HRQL as measured by the EORTC QLQ-C30 and QLQ-OG25; pain on scars on chest, low mood and reduced energy or activity tolerance.These symptoms will be referred to as LASER key symptoms and will be investigated in this study [22].(Appendix A).

Statistical analysis
Descriptive statistics summarized the patient, tumor and treatment characteristics of the included patients.Continuous variables are noted as means with standard deviation or medians and interquartile ranges (IQR).Categorical variables were noted with percentage frequencies.
Univariable and multivariable linear regression analyses were performed for the analysis of the differences in HRQL domain scores (EORTC QLQ-C30 and QLQ-OG25) between patients 'with' and patients 'without' complications with adjustment for the confounders; age, sex, neoadjuvant treatment, years since surgery, surgical technique, surgical access, anastomotic site, pathological tumor stage and tumor subtype.HRQL domain scores were entered in the multivariable analysis if a p-value of <0.10 was reached in univariable analysis.Univariable and multivariable pairwise subgroup analyses were also performed according to severity of complications between patient with minor versus no, severe versus no, and minor versus severe postoperative complications.Mean scores with 95% confidence intervals (95%CI) were calculated and were compared between groups.Statistical significance was tested using the Student's t-test.A p-value of <0.01 was considered to be statistically significant.
For LASER key symptoms odds ratios adjusted for above mentioned confounders were calculated with a 95%CI using logistic regression models.To reduce the type-1 error due to multiple comparisons, statistical significance only tested if adjusted mean score difference was of clinical relevance [27e29].A p-value of <0.05 was considered to be statistically significant.
Additionally, subgroup analyses were performed to examine whether HRQL differs with the occurrence and severity of postoperative complications over time.These analyses and results can be found supplement I.
In this study, a mean score difference in HRQL score of 10 or more points was considered clinically relevant.Whereas the minimally important change in mean scores varies across HRQL domains, a cut-off point of 10 points is most likely the upper bound for most domains [24,27].
Data was handled anonymously and the statistical analysis were conducted by a biostatistician (A.J.) with SAS 9.4 software.

Patient, tumor and complication characteristics
A total of 645 patients from the LASER dataset were included for analysis in this study.The response rate in the LASER study was 81%: 876 of 1081 invited patients.Two hundred thirty-one patients were excluded from analyses in this study: 154 patients did not have available clinical and/or treatment data, and 77 patients had no available Clavien-Dindo categorization (Fig. 1).The majority of the remaining patients included in this study were male (78%), the mean age was 64 years (SD 9).Most patients (79%) received neoadjuvant therapy and 14% received adjuvant therapy.The most frequently performed surgical procedure was an Ivor-Lewis esophagectomy (54%), followed by a McKeown (25%).

Patients 'with' and 'without' postoperative complications
For HRQL domains global health, physical functioning, role functioning, social functioning, pain, dyspnea, appetite loss, reflux, pain and discomfort, and weight loss a p-value of <0.10 was found after univariable linear regression analysis.In multivariable analysis, significantly more dyspnea was found in patients with postoperative complications than in patients without postoperative complications (difference in means 6.26, p < 0.01) (Table 2, Fig. 2).However, this result was not clinically relevant (difference in means < 10 points) nor were the results of any other outcome of the EORTC QLQ-C30 and QLQ-OG25.
For the three LASER key symptoms, none of the odds ratios in the univariable analysis were statistically significant nor reached the threshold to perform multivariate analysis (Table 3).

Patients with 'no', 'minor' and 'severe' postoperative complications
After univariable linear regression analysis of all the HRQL domains, patients with minor postoperative complications reported higher levels of dyspnea (difference in means 6.8, p 0.006) and appetite loss (difference in means 7.5, p 0.003) than patients with no postoperative complications.
Patients with severe complications reported more dyspnea (difference in means 9.6, p < 0.001) compared to patients with no postoperative complications.
No statistically significant difference in mean scores was found for patients with minor postoperative complications compared to patients with severe postoperative complications (Table 4 [domains with statistically significant differences], Supplementary Table S2 and Supplementary Fig. S1 [all domains]).
For the domains entered in the multivariable analysis the patients with severe postoperative complications reported more dyspnea (difference in means 8.3, p < 0.01) compared to patients with no postoperative complications (Supplementary Table S3).
However, none of the differences for any of the 31 HRQL domains between any of the severity groups after univariable or multivariable regression analysis was found to reach the threshold for clinical relevance.
For patients with minor versus no postoperative complications the odds ratio for LASER key symptoms reduced energy and activity tolerance (OR 1.5, p 0.05) was statistically significant.
For patients with minor versus severe postoperative complications the odds ratio for LASER key symptoms low mood (OR 2.5, p 0.02) and reduced energy and activity tolerance (OR 1.6, p 0.04) were statistically significant.
For patient with severe versus no, none of the OR found for the LASER key symptoms differed significantly (Table 5).
After multivariable regression the found odds ratio for LASER key symptoms low mood (OR 2.3, p 0.04), for minor versus severe postoperative complications remained statistically significant (Supplementary Table S4).

Discussion
This study investigated whether long-term HRQL was associated with postoperative complications in disease-free patients following an esophagectomy for distal esophageal or gastroesophageal junction cancer after a follow-up of at least one year post completion of therapy.The results of this study show, that in general, long-term HRQL does not differ between patients with and without postoperative complications, nor between patients with different grades in severity of postoperative complications.Although the symptom dyspnea differed statistical significantly after multivariate analysis between patients with and without postoperative complications, and between patients with severe and no postoperative complications, this was not a clinically relevant difference (difference in means < 10 points).The other statistically significant finding in this study was that patients with minor postoperative complications reported more often the LASER key symptom low mood than patients with severe postoperative complications.These findings run counter to our hypothesis as we had expected that the presence and the severity of postoperative complications are proportionally negatively associated with longterm HRQL.A possible explanation could be, that in this current study, HRQL was measured at a mean of 4.4 years after treatment.Earlier studies have shown, that the postoperative reduction in HRQL in disease-free patients restores to baseline within one to two years after esophagectomy [19,30,31].However other studies report a negative effect of major postoperative complications to last more than 5 years.In the study by Kauppila et al. the impact of an esophagectomy with postoperative complications was found to be associated with an impaired HRQL up to 10 years after esophagectomy [14].In the study of Derogar et al. evaluating the influence of major postoperative complications on HRQL in 5-year survivors of esophageal cancer surgery, patients with major postoperative complications reported more problems, that are clinically relevant and statistically significant, for appetite loss, fatigue and dyspnea at 6-months, 3 years and 5 years compared to patients without postoperative complications [10].Comparable with our results is that significant more symptomology was reported for dyspnea by patients with severe postoperative complications, in our study this difference was not clinically relevant.In this current study, only disease-free patients, who were free from surgical complications at the time of assessment were included compared    to all patients who survived for 5 or more years after curative surgery in the prospective study by Derogar [10].Therefore the found differences could be less pronounced in our study, since patients with more severe complications may not have been included, due to ongoing postoperative complications or due to early death, which may lead to more fit patients in our study cohort with better HRQL.Several limitations of our study merit attention.This study has a cross-sectional design.We therefore do not know the baseline HRQL nor the baseline patient characteristics.Moreover, we cannot report changes in these scores over time.The follow-up varies from 1 year to 8 years, and there might be intra-individual difference over time but due to the cross-sectional design of this study and the one-time assessment of the questionnaires, this could not be investigated.The study may have been prone to non-response bias and selection bias, as it only includes patients who were diseasefree and at least one year following surgery.Patients who died, had recurrence of the disease, and patient who refused to participate in the study were not included in the analysis, which may have led to a bias towards patients with relatively more positive HRQL levels.The reason for patients declining participation was not recorded.Furthermore, in the LASER questionnaire, patients were asked to appoint symptoms experienced in the last 6 months, this might have led to recall bias.The EORTC QLQ-C30 and QLQ-OG25 focuses on symptoms experienced during the past week, this might have led to a snapshot of the HRQL.The number of statistical test performed was relatively high in this study, therefore to reduce the statistical probability of finding significant difference by chance, the statistical significance was only tested if adjusted mean score difference was of clinical relevance and a probability value of less than 0.01 was used.
This study also has several strengths.It employed a large sample size of patients who underwent an esophagectomy in 20 European centers.Most of the included patients were treated with neoadjuvant therapy.Therefore, our study reflects the current treatment practices of esophageal cancer patients in Europe and the results may be generalizable to esophageal cancer patients in other countries of Europe that adopt the same treatment practices.
There is a need of future studies employing a longitudinal design, (e.g. as in the PACAP, POCOP and PLCRC studies) [32] to enable the investigation of patterns of HRQL over time, i.e., deterioration, recovery and improvement in patients based on severity of postoperative complications.Additionally, what can be done to reduce the postoperative decline in HRQL, and what steps we can take to restore this HRQL in esophagectomy patients faster.
In conclusion, in this study comparable HRQL was found in patients with and without postoperative complications following esophagectomy for distal esophageal or gastro-esophageal cancer, after a mean of follow-up of 4.4 years.Patients with different grades of severity of complications had comparable long-term HRQL.The level of HRQL in esophageal cancer patients are more likely explained by the impact of the complex procedure of the esophagectomy itself.

Fig. 2 .
Fig. 2. Spider plots showing domain outcome scores for patients with (green line) and patients without (blue line) postoperative complications.Higher scores in global health and functioning domains represent better HRQL.In symptom domains, a higher score represents more symptomatology.The asterisk indicates a significant difference (p < 0.01) between patient with and patients without postoperative complications within a HRQL domain.A: EORTC QLQ-C30 Global health and functioning domains, B: EORTC QLQ-C30 symptom and financial domains, C: EORTC QLQ-OG25 Multi-item domains, D: EORTC QLQ-OG25 Single-item domains.
. The questionnaire assesses

Table 1
Baseline characteristics of patients with 'no', 'minor' and 'severe' postoperative complications following and esophagectomy for esophageal or junctional cancer.
Data are presented as n (%) unless otherwise indicated.IQR ¼ interquartile range.SD ¼ standard deviation.y ¼ year.Kg ¼ kilograms.MIE ¼ minimal invasive esophagectomy.c/(y)pTNM tumor staging classification.Minor postoperative complications defined as Clavien-Dindo grade I-II, severe postoperative complications defined as Clavien-Dindo !III.Percentages may not add up due to rounding.

Table 2
Univariable and multivariable linear regression analysis of HRQL comparing patients 'with' and patients 'without' postoperative complications.Health related quality of life (HRQL) domains with p-value <0.1 in univariable analysis were entered in multivariable analysis.In bold values that were statistically significant.(p-value<0.01).Patients without postoperative complications defined as not having experienced any complications postoperatively.Patient with complications defined as having experienced any postoperative complications(Clavien-dindo !I).N. Schuring, E. Jezerskyte, M.I. van Berge Henegouwen et al.European Journal of Surgical Oncology 49 (2023) 97e105 ∧ ¼ corrected for confounders.* ¼

Table 3
Univariable analysis of the LASER key symptoms comparing patients 'with' and patients with 'no' postoperative complications.CI) are shown for univariable analysis.None of odds ratios had a p-value <0.1 in univariable analysis therefore no multivariable analysis were performed.Patients without postoperative complications defined as not having experienced any complications postoperatively.Patient with complications defined as having experienced any postoperative complications (Clavien-dindo !I).

Table 4
Statistically significant differences in HRQL after univariable analysis comparing patients with minor versus no, severe versus no and severe versus minor postoperative complications.Difference in between means with 95% confidence interval (CI) are shown for univariable analysis.In bold values that were statistically significant (p-value < 0.01).* ¼ domains with p-value <0.1 in univariable analysis were entered in multivariable analysis.