Two-Year Quality of Life Outcomes After Robotic-Assisted Minimally Invasive and Open Esophagectomy

Abstract

Background:

Robotic-assisted minimally invasive esophagectomy (RAMIE) is a safe alternative to open esophagectomy (OE). However, differences in quality of life (QOL) after these procedures remain unclear. We previously reported on short-term QOL outcomes after RAMIE and OE and describe here our results from two years of follow-up.

Methods:

We conducted a prospective, nonrandomized trial of patients with esophageal cancer undergoing transthoracic resection via RAMIE or OE at a single institution. The primary outcomes were patient-reported QOL, measured by the Functional Assessment of Cancer Therapy–Esophageal (FACT-E), and pain, measured by the Brief Pain Inventory (BPI). Generalized linear models were used to assess the relationship between QOL outcomes and surgery cohort. P values were adjusted (p-adj) within each model using the false discovery rate correction.

Results:

In total, 170 patients underwent esophagectomy (106 OE and 64 RAMIE). The groups did not differ significantly by any measured clinicopathologic variables. After covariates were controlled for, FACT-E scores were higher in the RAMIE cohort than in the OE cohort (parameter estimate [PE], 6.13; p-adj=0.051). RAMIE was associated with higher esophageal cancer subscale (PE, 2.72; p-adj=0.022) and emotional well-being (PE, 1.25; p-adj=0.016) scores. BPI pain severity scores were lower in the RAMIE cohort than in the OE cohort (PE, −0.56; p-adj=0.005), but pain interference scores did not differ significantly between groups (p-adj=0.11).

Conclusions:

During 2 years of follow-up, RAMIE was associated with improved patient-reported QOL, including esophageal symptoms and emotional well-being, and decreased pain, compared with OE.

Esophagectomy is an important component of multimodality treatment for patients with resectable esophageal cancer.¹ Although open esophagectomy (OE) is the traditional curative standard, during the last decade, minimally invasive esophagectomy (MIE),² including robotic-assisted MIE (RAMIE),³ has emerged as a viable alternative approach. Previous reports have suggested that MIE results in decreased morbidity while achieving equivalent oncologic outcomes compared with OE.^4–8 MIE may also be associated with superior quality of life (QOL).^9,10 Multiple studies have compared MIE and OE, but few have made direct comparisons between RAMIE and OE, particularly with QOL as a primary endpoint. In the recently reported ROBOT trial,⁸ the only randomized controlled trial of RAMIE versus OE to date, QOL was assessed longitudinally as a secondary endpoint. However, only short-term QOL data have been reported from this trial.

We performed a prospective comparison of QOL and pain between patients undergoing RAMIE and those undergoing OE at a high-volume esophageal cancer center. We hypothesized that the minimally invasive approach would confer superior QOL and decreased pain. Early results from this study, including perioperative and short-term QOL outcomes, have been reported.¹¹ Here, we describe long-term QOL outcomes after two years of follow-up.

Patients and Methods

The design and methods for this study have been previously published.¹¹ In brief, we conducted a prospective, nonrandomized trial of consecutive, unselected patients with esophageal cancer undergoing transthoracic resection via RAMIE or OE at a single institution (ClinicalTrials.gov: NCT01558648). Patients requiring laryngectomy or colon interposition were excluded. The primary outcomes were (1) patient-reported QOL, measured by the Functional Assessment of Cancer Therapy–Esophageal (FACT-E),^12,13 and (2) pain, measured by the Brief Pain Inventory (BPI).¹⁴ All QOL data were collected prospectively by research study staff using the described instruments in direct patient interviews and/or correspondence. Secondary outcomes from the study, including complications and perioperative outcomes, have been reported.¹¹ The surgical approach (RAMIE versus OE) was determined by the surgeon to whom each patient presented, with two of eight surgeons performing RAMIE³ exclusively and all others performing OE only. Ivor Lewis was the predominant type of esophagectomy in both the robotic group and the open group (RAMIE, 62/64; OE, 103/106). The esophagogastric anastomosis was created with a circular stapler in the majority of cases. Preoperative and postoperative treatment algorithms were standardized. The study was approved by the Institutional Review Board at Memorial Sloan Kettering Cancer Center. Written informed consent was obtained from all enrolled patients.

The study was powered to detect differences in QOL outcomes at 4 months, as described in our previous manuscript.¹¹ Data were analyzed on a per-protocol basis. Thus, 2 patients who underwent conversion from RAMIE to OE were analyzed with the OE cohort.

FACT-E

FACT-E was scored according to guidelines from FACIT.org. FACT-E is the sum of the Functional Assessment of Cancer Therapy–General (FACT-G), a general QOL instrument,¹² and an esophageal cancer subscale (ECS).¹³ FACT-G consists of 4 subscales: (1) physical well-being, (2) functional well-being, (3) social/family well-being, and (4) emotional well-being. The ECS addresses specific concerns of patients with esophageal cancer, such as swallowing, appetite, weight loss, difficulty breathing, and voice quality. Questions from each subscale were answered on a 5-point Likert scale.

BPI

BPI was scored according to guidelines from the IMMPACT panel.¹⁴ The BPI includes 2 subscales: (1) pain severity and (2) pain interference. For pain severity, patients were asked to rate their worst pain, least pain, average pain, and current pain on a scale from 0 to 10. The scores from these items were averaged to calculate the pain severity score. For pain interference, patients were asked to assess how much pain has interfered with their general activity, walking, work, mood, enjoyment of life, relationships, and sleep, also on a scale from 0 to 10. These scores were averaged to calculate the pain interference score.

Statistical Methods

Comparison of clinical characteristics between surgery cohorts was performed with the Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for categorical variables. Two-year overall survival (OS) from surgery until death or last follow-up was estimated using Kaplan-Meier methods. QOL scores were visualized with boxplots and summarized with medians and ranges stratified by surgery cohort. We performed a sensitivity analysis using univariable linear models to assess differences in baseline scores between surgery cohorts. We also checked for interactions between surgery cohort and time for each QOL scale.

Generalized linear models with linear link function and unstructured covariance matrices were used to assess the relationship between QOL and surgery cohort over time. A multivariable model was created for each QOL scale,, including the following potential confounding factors: age, sex, BMI, ASA class, comorbidities (cardiac, diabetes, pulmonary, and renal failure), induction treatment, pathologic stage, and impending death. To account for informative missingness, we incorporated death while on study into the multivariable models. A patient’s death status was recorded at a given time point if death occurred during the following 6 months. Within the generalized linear models, the parameter estimate (PE) with 95% confidence interval (CI) represents the average expected difference in score between comparison groups. Individual hypothesis tests were constructed to compare RAMIE to OE and all time points to both baseline and 1 month after surgery. For each QOL scale, the overall difference in score between the RAMIE and OE cohorts was visualized as a forest plot of the least-squares means from the corresponding multivariable model.

To account for multiple testing, P values were adjusted using the false discovery rate correction in the multivariable models for each QOL scale and in our comparison of clinicopathologic variables between surgical cohorts. Two-sided adjusted P values (p-adj) <0.05 were considered significant. All analyses were performed using SAS 9.4 TS1M4 (SAS Institute, Cary, NC).

Results

Patient Characteristics

In total, 170 patients underwent esophagectomy (106 OE and 64 RAMIE) between March 2012 and August 2014. Patient characteristics are summarized in Table 1. There were no significant differences between surgical cohorts for any of the measured clinicopathologic variables (p-adj=0.31 to >0.95). Two-year OS was 78% (95% CI, 65%−86%) for RAMIE and 65% (95% CI, 55%−74%) for OE.

Table 1:

Patient and Tumor Characteristics

Characteristic		OE (N=106)	RAMIE (N=64)	P-adj a
Age, years		63 (28–83)	61 (45–82)	>0.95
BMI		28.4 (16.9–49.5)	29.1 (15.6–47.8)	>0.95
Sex
Male		91 (85.8)	53 (82.8)	>0.95
Female		15 (14.2)	11 (17.2)
ASA class
	2	15 (14.2)	9 (14.1)	>0.95
	3	84 (79.2)	51 (79.7)
	4	7 (6.6)	4 (6.3)
Histologic subtype
Adenocarcinoma		98 (92.5)	59 (93.7)	>0.95
Squamous cell carcinoma		7 (6.6)	4 (6.3)
Other		1 (0.9)	0 (0)
Clinical stage
0		2 (1.9)	1 (1.6)	>0.95
I		14 (13.2)	11 (17.5)
II		26 (24.5)	17 (27)
III		63 (59.4)	34 (54)
IV		1 (0.9)	0 (0)
Pathologic stage
0		20 (19)	13 (20.3)	>0.95
I		25 (23.8)	22 (34.4)
II		33 (31.4)	15 (23.4)
III		27 (25.7)	14 (21.9)
Induction treatment
No		19 (17.9)	16 (25)	>0.95
Yes		87 (82.1)	48 (75)
Cardiac comorbidity
No		43 (40.6)	29 (45.3)	>0.95
Yes		63 (59.4)	35 (54.7)
Pulmonary comorbidity
No		81 (76.4)	58 (90.6)	0.31
Yes		25 (23.6)	6 (9.4)
Renal failure
No		103 (97.2)	63 (98.4)	>0.95
Yes		3 (2.8)	1 (1.6)
Diabetes
No		87 (82.1)	50 (78.1)	>0.95
Yes		19 (17.9)	14 (21.9)

Data are no. (%) or median (range). ASA, American Society of Anesthesiologists; OE, open esophagectomy; RAMIE, robotic-assisted minimally invasive esophagectomy.

^aP values were adjusted using the false discovery rate correction.

FACT-E

FACT-E data are summarized in Figure 1 and Supplemental Table 1. After 24 months of follow-up, FACT-E scores were available for 39% of the OE cohort (41/106) and 56% of the RAMIE cohort (36/64). The OE and RAMIE cohorts started with similar baseline scores (Supplemental Table 2) and experienced a similar pattern of score shifts over time. In both groups, QOL scores decreased from baseline to 1 month postoperative, increased substantially between 1 and 4 months, and increased slowly or remained unchanged through 24 months. There were no significant interactions between surgery cohort and time (p=0.21–0.49). Thus, we focused on the difference in QOL scores between surgery cohorts and the difference in QOL scores over time as separate main effects in our models.

Figure 1:

Box plots with overlay lines for mean of Functional Assessment of Cancer Therapy–Esophageal (FACT-E), Functional Assessment of Cancer Therapy–General (FACT-G), and esophageal cancer subscale (ECS) scores. OE, open esophagectomy; PO, postoperative; RAMIE, robotic-assisted minimally invasive esophagectomy.

After covariates were controlled for, FACT-E scores were higher in the RAMIE cohort than in the OE cohort (parameter estimate [PE], 6.13; 95% CI, 0.92–11.34; p-adj=0.051) (Table 2). In the generalized linear models, the PE represents the average expected difference in score between comparison groups. In this instance, the PE of 6.13 means that, on average, FACT-E score was 6.13 points higher for RAMIE than for OE (Figure 2). FACT-G scores were not significantly different between groups (p-adj=0.18) (Table 2). Of note, RAMIE was associated with higher ECS (PE, 2.72; 95% CI, 0.72–4.72; p-adj=0.022) and emotional well-being (PE, 1.25; 95% CI, 0.37–2.12; p-adj=0.016) scores, compared with OE (Table 2 and Supplemental Table 3). No significant differences were seen with respect to physical (padj=0.10), functional (p-adj=0.41), and social (p-adj=0.91) well-being (Supplemental Tables 4-6).

Table 2:

Generalized Linear Models for FACT-E, FACT-G, and ECS

	Multivariable Model
	PE	95% CI	P-adj a	Overall P-adja
FACT-E
RAMIE	6.13	0.92–11.34		0.051
OE	REF
Time point				<.001
1 month PO vs base	−23.2	−26.80 to − 19.57	<.001
24 months PO vs base	−3.22	−6.97 to 0.52	0.18
24 months PO vs 1 month PO	19.96	15.72−24.20	<.001
FACT-G
RAMIE	3.15	−0.52 to 6.81		0.18
OE	REF
Time point
1 month PO vs base	−11.8	−14.10 to − 9.57	<.001	<.001
24 months PO vs base	0.93	−1.61 to 3.47	0.60
24 months PO vs 1 month PO	12.76	9.90–15.63	<.001
ECS
RAMIE	2.72	0.72–4.72		0.022
OE	REF
Time point
1 month PO vs base	−11.6	−13.36 to − 9.92	<.001
24 months PO vs base	−4.21	−6.05 to −2.36	<.001
24 months PO vs 1 month PO	7.43	5.56–9.30	<.001

Multivariable model is adjusted for age, sex, body mass index, American Society of Anesthesiologists class, cardiac comorbidity, diabetes, pulmonary comorbidity, renal failure, induction treatment, pathologic stage, and death.

ECS, esophageal cancer subscale; FACT-E, Functional Assessment of Cancer Therapy–Esophageal; FACT-G, Functional Assessment of Cancer Therapy–General; OE, open esophagectomy; Overall P, P value from global F test; P, individual hypothesis test P value; PE, parameter estimate; PO, postoperative; RAMIE, robotic-assisted minimally invasive esophagectomy.

^aP values were adjusted using the false discovery rate correction.

Figure 2:

Forest plot of least-squares mean of quality of life scores by surgery cohort. CI, confidence interval; ECS, esophageal cancer subscale; FACT-E, Functional Assessment of Cancer Therapy–Esophageal; FACT-G, Functional Assessment of Cancer Therapy–General; LS, least-squares; OE, open esophagectomy; PE, parameter estimate; RAMIE, robotic-assisted minimally invasive esophagectomy.

FACT-E scores for both procedures were lower at 1 month than at baseline (PE, −23.2; 95% CI, −26.8 to −19.57; p-adj<0.001), higher at 24 months than at 1 month (PE, 19.96; 95% CI, 15.72–24.20; p<0.001), and not significantly different between 24 months and baseline (PE, −3.22; p-adj=0.18) (Table 2). FACT-G, physical well-being, and functional well-being time-point comparisons followed a similar pattern (Table 2 and Supplemental Tables 4 and 5). In contrast, ECS scores remained depressed at 24 months compared with baseline (PE, −4.21; 95% CI, −6.05 to −2.36; p-adj<0.001). Emotional well-being scores were higher at 1 month than at baseline (PE, 1.10; 95% CI, 0.43–1.77; p-adj=0.005) and remained elevated through 24 months of follow-up (PE, 1.48; 95% CI, 0.68–2.28; p-adj=0.001) (Supplemental Table 3).

Age was positively associated with scores for FACT-E (PE, 0.47; 95% CI, 0.17–0.76; p-adj=0.006), FACT-G (PE, 0.27; 95% CI, 0.06–0.48; p-adj=0.032), ECS (PE, 0.22; 95% CI, 0.1–0.33; p-adj<0.001), emotional well-being (PE, 0.07; 95% CI, 0.02–0.12; p-adj=0.021), and social well-being (PE, 0.06; 95% CI, 0.01–0.11; p-adj=0.030) (Supplemental Tables 3, 6-9). Additionally, patients who died had lower scores just before death, compared with patients who did not die, for FACT-E (PE, −11.3; 95% CI-, 18.39 to −4.27; p-adj=0.005), FACT-G (PE, −8.09; 95% CI, −12.97 to −3.22; p-adj=0.004), ECS (PE, −3.29; 95% CI, −5.95 to −0.62; p-adj=0.041), emotional well-being (PE, −2.15; 95% CI, −3.40 to −0.89; padj=0.003), physical well-being (PE, −3.21; 95% CI, −4.68 to −1.74; p-adj<0.001), and functional well-being (PE, −3.23; 95% CI, −4.84 to −1.61; p-adj<0.001) (Supplemental Tables 3-5, 7-9).

BPI

BPI data are summarized in Figure 3 and Supplemental Table 10. After 24 months of follow-up, BPI scores were available for 39% of the OE cohort (41/106) and 56% of the RAMIE cohort (36/64). The OE and RAMIE cohorts started with similar baseline scores (Supplemental Table 2) and experienced a similar pattern of score shifts over time. As previously noted in our analysis of short-term outcomes,¹¹ scores for both BPI scales peaked at the first inpatient visit and decreased substantially over the course of 4 months. On long-term follow-up, these scores decreased slowly or remained unchanged between 4 and 24 months.

Figure 3:

Box plots with overlay lines for mean of Brief Pain Inventory (BPI) pain severity and pain interference scores. OE, open esophagectomy; PO, postoperative; RAMIE, robotic-assisted minimally invasive esophagectomy.

After covariates were controlled for, pain severity scores were lower in the RAMIE cohort than in the OE cohort (PE, −0.56; 95% CI, −0.92 to −0.20; p-adj=0.005) (Table 3, Figure 2). However, pain interference scores did not differ significantly between groups (PE, −0.35; p-adj=0.11).

Table 3:

Generalized Linear Models for BPI Pain Severity and Pain Interference

	Multivariable Model
	PE	95% CI	P-adj a	Overall P-adja
Pain severity
RAMIE	−0.56	−0.92 to −0.20		0.005
OE	REF
Time point				<.001
1 month PO vs base	1.02	0.71−1.33	<.001
24 months PO vs base	0.06	−0.26 to 0.39	0.75
24 months PO vs 1 month PO	−0.96	−1.30 to −0.62	<.001
Pain interference
RAMIE	−0.35	−0.72 to 0.03		0.11
OE	REF
Time point				<.001
1 month PO vs base	2.16	1.78–2.53	<.001
24 months PO vs base	0.30	−0.05 to 0.65	0.14
24 months PO vs 1 month PO	−1.85	−2.31 to −1.40	<.001

Multivariable model is adjusted for age, sex, body mass index, American Society of Anesthesiologists class, cardiac comorbidity, diabetes, pulmonary comorbidity, renal failure, induction treatment, pathologic stage, and death.

BPI, Brief Pain Inventory; OE, open esophagectomy; Overall P, P value from global F test; P, individual hypothesis test P value; PE, parameter estimate; PO, postoperative; RAMIE, robotic-assisted minimally invasive esophagectomy.

^aP values were adjusted using the false discovery rate correction.

Pain severity scores for both procedures were higher at 1 month than at baseline (PE, 1.02; 95% CI, 0.71–1.33; p-adj<0.001), lower at 24 months than at 1 month (PE, −0.96; 95% CI, −1.30 to −0.62; p-adj<0.001), and not significantly different between 24 months and baseline (p-adj=0.75) (Table 3). Of note, pain severity remained elevated, compared with baseline, up to 4 months after surgery (PE, 0.35; 95% CI, 0.05–0.65; p-adj=0.043) (Supplemental Table 11). Pain interference scores followed a similar pattern and likewise remained elevated, compared with baseline, up to 4 months after surgery (PE, 0.59; 95% CI, 0.20–0.98; p-adj=0.007) (Supplemental Table 12).

Age was negatively associated with pain severity (PE, −0.04; 95% CI, −0.05 to −0.02; p-adj<0.001) and pain interference (PE, −0.02; 95% CI, −0.04 to 0.0; p-adj=0.035) (Supplemental Tables 11 and 12). Additionally, patients who died had higher scores just before death, compared with patients who did not die, for both pain severity (PE, 1.24; 95% CI, 0.68–1.80; p-adj<0.001) and pain interference (PE, 1.15; 95% CI, 0.47–1.83; p-adj=0.002) (Supplemental Tables 11 and 12).

Comment

Esophagectomy is associated with significant morbidity and mortality and carries important implications for health-related QOL (HRQOL) in patients undergoing treatment for esophageal cancer.¹⁵ In a systematic review of HRQOL after esophagectomy, Parameswaran et al. noted a significant decline in most aspects of HRQOL in the immediate postoperative period, with many deficits lasting up to 1 year after surgery.¹⁶ Other investigators have described longer-lasting effects.¹⁷ Clinical predictors of postoperative HRQOL have also been identified, including comorbidity, tumor stage, and surgical complications.^18,19

Several randomized trials comparing MIE and OE have evaluated HRQOL as a secondary outcome.^8–10 In the TIME trial, the first randomized controlled trial comparing MIE and OE, MIE was associated with improved 6-week postoperative scores for physical component summary (SF-36), global health (EORTC C30), and talking and pain (OES18).⁵ Physical component summary, global health, and pain continued to favor MIE after 1 year of follow-up.⁹ In the ROBOT trial, the only randomized controlled trial of RAMIE versus OE to date, van der Sluis et al.⁸ assessed QOL at regular intervals for up to 5 years. However, only short-term results were presented in their initial report. Compared with OE, RAMIE was associated with superior global health and physical function 6 weeks after surgery.

In agreement with other studies, we observed a significant decline in global QOL (FACT-E and FACT-G) for all patients after esophagectomy, with substantial improvement between 1 and 4 months postoperatively and complete recovery at 24 months of follow-up. Physical and functional well-being followed a similar temporal trend. However, despite substantial improvement between 1 and 4 months, esophageal well-being did not fully recover to baseline, even 24 months after surgery. Emotional well-being improved after esophagectomy and remained elevated for the duration of the study. Similar long-term effects of esophagectomy on emotional measures have been observed by other investigators and may speak to the optimism of patients undergoing potentially curative treatment for an otherwise harrowing disease.²⁰ As expected, pain increased significantly for all patients after esophagectomy, with substantial improvement between the first inpatient time point and 1 month of follow-up. It is worth noting that, although complete recovery to baseline was achieved at 24 months, a small but discernable increase in pain scores persisted for up to 4 months after esophagectomy.

When the two surgical cohorts were compared, our primary QOL measure, FACT-E, favored RAMIE over OE. This result was largely driven by underlying differences in emotional well-being and ECS scores, which clearly favored RAMIE. The ECS addresses issues that are particularly relevant to patients with esophageal cancer, such as swallowing, appetite, weight loss, difficulty breathing, and voice quality. The superior performance of RAMIE with regard to esophageal symptoms is an important finding from our study and should be confirmed in future investigations. We did not find a significant difference between RAMIE and OE in terms of FACT-G, a general QOL instrument. A large body of literature suggests that MIE is associated with less postoperative pain than OE.^5,8 Indeed, in our study, RAMIE was associated with decreased pain severity, compared with OE. Pain interference, however, did not differ between surgical approaches. Thus, although we previously observed a significant difference in pain interference, favoring RAMIE, after limited follow-up,¹¹ this finding was not borne out after 24 months.

Although not surprising, it is worth noting that patients nearing death had worse scores for essentially all QOL scales evaluated in this study. This finding is consistent with previous reports²¹ and underscores the importance of controlling for impending death in the long-term analysis of our data. Age was associated with marginally increased QOL scores and marginally decreased pain scores. However, the mechanism for this protective effect is unclear.

This study has several important limitations. Foremost, the nonrandomized design introduces an inherent risk of selection bias. Although patient characteristics appeared remarkably similar between the RAMIE and OE groups, it is possible that unmeasured patient variables may have influenced the decision to undergo robotic or open surgery. In this study, all robotic procedures were performed by only two of eight surgeons. However, these were experienced robotic surgeons, both well beyond the learning curve for RAMIE, thus enabling us to make valid comparisons between surgical techniques. To minimize the effects of recall bias, both QOL and pain instruments were collected in a rigorous, prospective manner. However, because of the unblinded nature of the study, it is possible that patient perception of the operation performed could have had an effect on subjective, self-reported outcome measures. Use of opioid pain medication, a potential confounder for pain assessment, was not evaluated. Regarding statistical power, our study was specifically powered to detect differences in QOL outcomes at 4 months and may have been underpowered to detect differences at later time points up to 24 months. Finally, this study reflects the experience of a single high-volume esophageal cancer center with considerable experience with esophagectomy and well-established preoperative and postoperative protocols for management of these patients, potentially limiting the generalizability of our results.

In conclusion, during 2 years of follow-up, RAMIE was associated with improved patient-reported QOL, compared with OE, especially in regard to esophageal symptoms and emotional well-being. RAMIE was also associated with decreased postoperative pain. However, pain interference did not differ between surgical groups. Taken together, these findings suggest that RAMIE may offer HRQOL benefits to patients undergoing curative resection for esophageal cancer and should receive consideration as a minimally invasive alternative to OE.

Abbreviations

BPI

Brief Pain Inventory

CI

confidence interval

ECS

esophageal cancer subscale

FACT-E

Functional Assessment of Cancer Therapy–Esophageal

FACT-G

Functional Assessment of Cancer Therapy–General

HRQOL

health-related quality of life

MIE

minimally invasive esophagectomy

OE

open esophagectomy

OS

overall survival

p-adj

adjusted P

PE

parameter estimate

QOL

quality of life

RAMIE

robotic-assisted minimally invasive esophagectomy