Study population and eligibility criteria
We retrospectively analyzed the case data of 699 patients (aged ≥ 65 years) who underwent radical surgery for esophageal cancer from January 1, 2012 to January 31, 2017 in the perioperative database of geriatric thoracic surgery in the First Medical Center of Chinese PLA General Hospital. Esophageal cancer diagnoses were verified via manual chart review of electronic medical records. Ethics Committee Board of the First Medical Center of Chinese PLA General Hospital (No. S2021-342–01) waived the need for informed consent due to the retrospective nature of the study.
In this study, patients included: 1) If they were elderly (age ≥ 65 years), tissue confirmed diagnosis of esophageal cancer and underwent an esophagectomy (comprising partial and total esophagectomy. 2) The surgical time takes more than 2 h, and the postoperative hospital stay is longer than 3 days. Patients were excluded: 1) patients with ASA classification V. 2) Patients with delirium, coma or disturbance of consciousness before the operation. 3) Pneumonia and pulmonary infection in the patient's preoperative medical records. 4) Missing covariate data information for patients or accurate outcome information during follow-up. Figure 1 shows the inclusion and exclusion criteria for the study population of 699 individuals.
Data Sources and Covariates
Electronic data is derived from patient records systems, clinical examination systems, medical imaging management systems, radio-information management systems, transfusion management systems and nursing workstations. Access to data in electronic medical record systems using SQL servers (Microsoft, United States). From the patient record integrated management system (PRIDE 2.1.2.193, Heren Health, China), we extracted the preoperative data of eligible patients, including gender, age, American society of Anesthesiologists (ASA) classification, BMI, smoking history, drinking history, comorbidities (coronary heart disease, myocardial infarction, cerebrovascular disease, asthma, malignant tumor, renal insufficiency) and mFI-5 variables (hypertension, type Ⅱ diabetes, congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), dependent functional status). Intraoperative data obtained from anesthesia and medical records included anesthesia method, estimated blood loss, operative duration, anesthesia duration, surgical procedure and pathological diagnosis type. Laboratory testing includes hemoglobin, red blood cell, serum albumin, leukocyte count, Serum potassium, blood sodium, total bilirubin, preoperative mean artery pressure.
We manually checked electronic anesthesia records for all procedures and compared the results with electronic data extractions for these cases to ensure data accuracy. All data (100%) was then reviewed and confirmed by at least 6 clinical professionals, and conflicting data was double-checked to ensure data quality.
5-index Modified Frailty Index
The deficits model considers frailty to be a burden of risk factors leading to adverse events. The more deficits that a person has, the frailer they are. Data development from the Canadian Health and Aging Study (CSHA) forms the basis of the mFI-5 scale[15]. Unlike previous frailty index, mFI-5 uses a small number of variables readily available in a patient's history, including functional status (partial or complete dependence), history of diabetes, COPD, congestive heart failure, and hypertension requiring medication[16]. 1 point is assigned to each variable. Functional status refers to needing some or all of the assistance of others in daily activities, including bathing, eating, dressing, going to the toilet, moving, traveling, and more[17]. The mFI-5 score was calculated by increasing the number of variables per patient[18]. Patients were divided into 3 groups based on their mFI-5: frail group (mFI-5, 2–5), prefrail group (mFI-5, 1) and robust group (mFI-5, 0)[19]. The range of the mFI-5 is from 0 to 5 with increments of 1, and increasing the mFI-5 implies increasing frailty. We compared characteristics including patient demographics, intraoperative information, comorbidities, and laboratory test factors between robust, prefrail, and frail groups. Covariates were included based on clinical relevance, previous studies of surgical outcomes for esophageal cancer surgical outcomes, and availability in the database (Table 1).
Table 1
Demographic and clinical characteristics of patients in different frailty groups (n = 699)
Variables
|
Total (n = 699)
|
Robust (n = 342)
|
Prefrail (n = 184)
|
Frail (n = 173)
|
P ¶
|
Demographics
|
|
|
|
|
|
Gender, Male
|
577 (82.5)
|
280 (81.9)
|
139 (75.5)
|
158 (91.3)
|
< 0.001
|
Age (years)
|
|
|
|
|
|
65–75
|
638 (91.3)
|
310 (90.6)
|
169 (91.8)
|
159 (91.9)
|
0.846
|
> 75
|
61 (8.7)
|
32 (9.4)
|
15 (8.2)
|
14 (8.1)
|
|
ASA classification
|
|
|
|
|
|
≤ Ⅱ
|
427 (61.1)
|
295 (86.3)
|
73 (39.7)
|
59 (34.1)
|
< 0.001
|
Ⅲ
|
231 (33.0)
|
40 (11.7)
|
98 (53.3)
|
93 (53.8)
|
|
Ⅳ
|
41 (5.9)
|
7 (2.0)
|
13 (7.1)
|
21 (12.1)
|
|
BMI (kg/m2)
|
|
|
|
|
|
≤ 25
|
490 (70.1)
|
257 (75.1)
|
108 (58.7)
|
125 (72.3)
|
0.001
|
25–30
|
190 (27.2)
|
80 (23.4)
|
66 (35.9)
|
44 (25.4)
|
|
> 30
|
19 (2.7)
|
5 (1.5)
|
10 (5.4)
|
4 (2.3)
|
|
Smoking history
|
161 (23.0)
|
68 (19.9)
|
40 (21.7)
|
53 (30.6)
|
0.021
|
Drinking history
|
181 (25.9)
|
82 (24.0)
|
47 (25.5)
|
52 (30.1)
|
0.328
|
Intraoperative information
|
|
|
|
|
|
Anesthesia method
|
|
|
|
|
|
Intravenous general anesthesia
|
18 (2.6)
|
10 (2.9)
|
5 (2.7)
|
3 (1.7)
|
0.716
|
Intravenous inhalation anesthesia
|
681 (97.4)
|
332 (97.1)
|
179 (97.3)
|
170 (98.3)
|
|
Estimated blood loss, ml
|
150.0 (100.0-200.0)
|
200.0 (100.0-200.0)
|
175.0 (100.0-200.0)
|
150.0 (100.0-200.0)
|
0.143
|
Operative duration, min
|
240.0 (193.0-310.0)
|
238.0 (190.0-306.3)
|
250.0 (205.3–304.0)
|
243.0 (186.0-314.0)
|
0.450
|
Anesthesia duration, min
|
290.0 (240.0-367.0)
|
285.0 (239.3-364.75)
|
300.0 (252.0-357.5)
|
291.0 (230.0-370.0)
|
0.647
|
Surgical procedure
|
|
|
|
|
|
Thoracotomy
|
485 (69.4)
|
234 (68.4)
|
130 (70.7)
|
121 (69.9)
|
0.855
|
Minimally invasive surgery
|
214 (30.6)
|
108 (31.6)
|
54 (29.3)
|
52 (30.1)
|
|
Pathological diagnosis type
|
|
|
|
|
|
Squamous cell carcinoma
|
455 (65.1)
|
223 (65.2)
|
113 (61.4)
|
119 (68.8)
|
0.587
|
Adenocarcinoma
|
222 (31.8)
|
109 (31.9)
|
63 (34.2)
|
50 (28.9)
|
|
Other
|
22 (3.1)
|
10 (2.9)
|
8 (4.3)
|
4 (2.3)
|
|
Surgical approaches
|
|
|
|
|
0.596
|
McKeow
|
74(10.6)
|
33(9.6)
|
23(12.5)
|
18(10.4)
|
|
Ivor Lewis
|
625(89.4)
|
309(90.4)
|
161(87.5)
|
155(89.6)
|
|
Extent of lymphadenectomy
|
|
|
|
|
0.902
|
Three-field
|
41(5.9)
|
19(5.6)
|
12(6.5)
|
10(5.8)
|
|
Two field
|
658(94.1)
|
323(94.4)
|
172(93.5)
|
163(94.2)
|
|
Resection margin
|
|
|
|
|
0.362
|
R0
|
657(94.0)
|
324(94.7)
|
169(91.8)
|
164(94.8)
|
|
R1
|
42(6.0)
|
18(5.3)
|
15(8.2)
|
9(5.2)
|
|
Location of cancer
|
|
|
|
|
0.566
|
Upper thoracic esophagus
|
31(4.4)
|
14(4.1)
|
7(3.8)
|
10(5.8)
|
|
Middle thoracic esophagus
|
316(45.3)
|
147(43.1)
|
91(49.5)
|
78(45.1)
|
|
Lower thoracic esophagus
|
351(50.3)
|
180(52.8)
|
86(46.7)
|
85(49.1)
|
|
mFI-5
|
|
|
|
|
|
Hypertension
|
217 (31.0)
|
0 (0.0)
|
138 (75.0)
|
79 (45.7)
|
< 0.001
|
Type Ⅱ diabetes
|
94 (13.4)
|
0 (0.0)
|
36 (19.6)
|
58 (33.5)
|
< 0.001
|
CHF
|
8 (1.1)
|
0 (0.0)
|
0 (0.0)
|
8 (4.6)
|
< 0.001
|
COPD
|
24 (3.4)
|
0 (0.0)
|
0 (0.0)
|
24(13.9)
|
< 0.001
|
Dependent functional status
|
58 (8.3)
|
0 (0.0)
|
10 (5.4)
|
48 (27.7)
|
< 0.001
|
Comorbidity
|
|
|
|
|
|
Coronary heart disease
|
58 (8.3)
|
11 (3.2)
|
24 (13.0)
|
23 (13.3)
|
< 0.001
|
Myocardial infarction
|
6 (0.9)
|
1 (0.3)
|
1 (0.5)
|
4 (2.3)
|
0.055
|
Cerebrovascular disease
|
71 (10.2)
|
24 (7.0)
|
23 (12.5)
|
24 (13.9)
|
0.024
|
Asthma
|
12 (1.7)
|
3 (0.9)
|
6 (3.3)
|
3 (1.7)
|
0.133
|
Malignant tumor
|
17 (2.4)
|
10 (2.9)
|
4 (2.2)
|
3 (1.7)
|
0.685
|
Renal insufficiency
|
5 (0.7)
|
1 (0.3)
|
2 (1.1)
|
2 (1.2)
|
0.429
|
Laboratory testing
|
|
|
|
|
|
Hemoglobin, g/L
|
137.0 (127.0-147.0)
|
138.0 (127.0-147.0)
|
137.0 (124.0-149.0)
|
136.0 (128.0-145.0)
|
0.600
|
RBC, 1012/L
|
4.5 (4.2–4.8)
|
4.5 (4.2–4.7)
|
4.5 (4.2–4.8)
|
4.5 (4.1–4.7)
|
0.964
|
Serum albumin, g/L
|
40.5 (38.2–43.0)
|
40.6 (38.0-43.1)
|
40.8 (38.6–43.7)
|
40.2 (38.2–42.0)
|
0.091
|
Leukocyte count, 109/L
|
6.0 (5.0-7.2)
|
5.8 (4.8-7.0)
|
6.0 (5.1–7.2)
|
6.3 (5.4–7.7)
|
0.002
|
Serum potassium, mmol/L
|
4.1 (3.9–4.4)
|
4.1 (3.9–4.4)
|
4.1 (3.9–4.3)
|
4.2 (3.9–4.4)
|
0.009
|
Blood sodium, mmol/L
|
142 (140.3-143.5)
|
142.3 (140.5-143.7)
|
141.8 (140.0-143.4)
|
141.7 (140.2-143.2)
|
0.017
|
TBIL, µmol/L
|
10.8(8.6–13.8)
|
10.5 (8.6–13.4)
|
11.3 (9.0-14.8)
|
10.7 (8.3–13.6)
|
0.173
|
Preoperative MAP, mmHg
|
94.5 (11.0)
|
93.1 (10.3)
|
97.8 (11.4)
|
93.7 (11.3)
|
< 0.001
|
Outcomes
|
|
|
|
|
|
Postoperative delirium
|
76 (10.9)
|
10 (2.9)
|
27 (14.7)
|
39 (22.5)
|
< 0.001
|
Postoperative pneumonia
|
54 (7.7)
|
13 (3.8)
|
18 (9.8)
|
23 (13.3)
|
< 0.001
|
Postoperative 30-day mortality
|
17 (2.4)
|
4 (1.2)
|
3 (1.6)
|
10 (5.8)
|
0.004
|
Abbreviations: ASA, American Society of Anesthesiologists; mFI, modified Frailty Index; CHF, congestive heart failure; RBC, red blood cell; BMI, body mass index; TBIL, total bilirubin; COPD, chronic obstructive pulmonary disease; MAP, mean artery pressure; SD, standard deviation; IQR, interquartile range. Note: Data are presented as n (%) or mean (SD) or median (IQR); Bold text hinted that these variables were statistically significant; ¶, comparison of patients in different frailty groups. |
Primary and Secondary Outcomes
Primary outcome was 30-day mortality. Secondary outcomes were postoperative pneumonia and postoperative delirium. Delirium is defined as the acute dysfunction of attention and cognition, which seriously affects the long-term quality of life of patients[20]. Delirium is recorded in medical and nursing records by describing text, including: altered mental state, confusion, disorientation, agitation, delirium, inappropriate behavior, attention deficit, hallucinations and aggressive behavior, disorganized thinking, altered level of consciousness, memory impairment, cognitive impairment, psychomotor disorder, awakening disorders and sleep cycle disorders[21]. Second, the patients preliminarily diagnosed by a computer were rechecked by neurologists using the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) criteria[22].
Clinical suspicion of pneumonia-infiltrated chest radiography and at least two clinical standards below[23, 24]: (1) patients had a vital sign of fever > 38.3°C; (2) leucocytosis > 12 × 10 9/ml; (3) patients had new or progressive respiratory symptoms, such as coughing and expectoration; (4) isolated pathogen from blood culture or sputum; (5) the presence of purulent tracheobronchial secretions. 30-day mortality outcomes and long-term survival data were obtained through telephone interviews with clinical professionals. Follow-up was bimonthly and survival data were recorded in the database by professionals.
Statistical Analysis
Continuous variables conforming to normal distribution were expressed in mean standard deviation (SD) and compared with the student's t test. While continuous variables that did not fit normal distribution were expressed as median (interquartile range, IQR), and compared using the Mann-Whitney U test. Categorical variables were expressed as frequency and percentage (%) and compared using the Chi-squared test or Fisher’s exact test. Survival curves were generated using the Kaplan-Meier (KM) method and compared using the log-rank test. Since there are too few people in ASA classification Ⅰ, the combination of Ⅰ and Ⅱ is expressed as ≤ Ⅱ. Univariable and multivariable logistic regression analyzes were used to identify independent risk factors for postoperative delirium, pneumonia, and 1-year mortality. The receiver operating characteristic (ROC) curve analysis, and decision curve analysis (DCA) were used to evaluate the predictive efficacy and clinical net benefit of different variables in predicting postoperative delirium, pneumonia, and 1-year mortality. All tests were double-tailed and p < 0.05 was considered statistically significant. Data analyzing was performed using the SPSS software (version 26.0, IBM, Armonk, New York, USA), and R program (version 3.6.3, R Foundation for Statistical Computing, Vienna, Austria). Packages used in the R environment included “rmda”, “survival”, “rms”, “survminer”, “pROC”.