Cancer Epidemiology 52 (2018) 91–98

Contents lists available at ScienceDirect

Cancer Epidemiology
journal homepage: www.elsevier.com/locate/canep

Patient education-level affects treatment allocation and prognosis in T

esophageal- and gastroesophageal junctional cancer in Sweden
⁎
Gustav Lindera, , Fredrik Sandinb, Jan Johanssonc, Mats Lindbladd, Lars Lundelld,
Jakob Hedberga
a
Department of Surgical Sciences, Uppsala University, Uppsala, Sweden
b
Regional Cancer Center Sweden, Uppsala, Sweden
c
Department of Surgery, Skåne University Hospital, Lund, Sweden
d
Division of Surgery, Center for Digestive Diseases, Karolinska University Hospital and Department of Clinical Science, Intervention and Technology, Karolinska Institutet,
Stockholm, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Background: Low socioeconomic status and poor education elevate the risk of developing esophageal- and
Education level junctional cancer. High education level also increases survival after curative surgery. The present study aimed to
Esophageal cancer investigate associations, if any, between patient education-level and treatment allocation after diagnosis of
Inequality in cancer treatment esophageal- and junctional cancer and its subsequent impact on survival.
Curative treatment
Methods: A nation-wide cohort study was undertaken. Data from a Swedish national quality register for eso-
Multi-disciplinary conference
phageal cancer (NREV) was linked to the National Cancer Register, National Patient Register, Prescribed Drug
Register, Cause of Death Register and educational data from Statistics Sweden. The effect of education level (low;
≤9 years, intermediate; 10–12 years and high > 12 years) on the probability of allocation to curative treatment
was analyzed with logistic regression. The Kaplan-Meier-method and Cox proportional hazard models were used
to assess the effect of education on survival.
Results: A total of 4112 patients were included. In a multivariate logistic regression model, high education level
was associated with greater probability of allocation to curative treatment (adjusted OR: 1.48, 95% CI:
1.08–2.03, p = 0,014) as was adherence to a multidisciplinary treatment-conference (adjusted OR: 3.13, 95% CI:
2.40–4.08, p < 0,001). High education level was associated with improved survival in the patients allocated to
curative treatment (HR: 0.82, 95% CI: 0.69–0.99, p = 0,036).
Discussion: In this nation-wide cohort of esophageal- and junctional cancer patients, including data regarding
many confounders, high education level was associated with greater probability of being offered curative
treatment and improved survival.

1. Introduction alternative. The above treatment regimes introduce a significant burden

Cancer of the esophagus and gastro-esophageal junction is the eight
most common type and the sixth most lethal type of cancer in the world
[1]. There are several life-style associated risk factors where low so-
cioeconomic and educational status elevate the risk of developing these
cancers [2,3]. Patients with higher education are known to have better
survival after curative esophageal cancer surgery [4].
In esophageal- and gastroesophageal junctional cancers, standard
curative treatment regime is multimodal with neoadjuvant chemor-
adiotherapy (CRT) followed by one of the most extensive oncological
surgical procedures [5]. In patients declining surgery or not physically
fit to undergo surgical treatment, curative CRT can be offered as an
on each individual patient’s physical and mental capacities [6,7]. Ide-
ally the choice of treatment should be guided by the preoperative tumor
staging, the patient’s health and stamina and the recommendation of a
multidisciplinary conference [8,9]. Such a recommendation is the base
for a final treatment decision made after a discussion with the patient.
Equality of healthcare for each individual patient is strived for in
many publicly funded health care systems, like in Sweden. Accordingly
it is stated in Swedish law that care should be given equally regardless
of socioeconomic status. Socioeconomic status is a wide concept con-
stituting many factors where education, in addition to income, marital
status, place of residence and occupation, play an important role. Since
poor education is associated with decreased survival after surgery for

⁎
Corresponding author at: Department of Surgical Sciences, Uppsala University, SE-75185, Uppsala, Sweden.
E-mail address: gustav.linder@surgsci.uu.se (G. Linder).

https://doi.org/10.1016/j.canep.2017.12.008
Received 10 October 2017; Received in revised form 14 December 2017; Accepted 16 December 2017
Available online 04 January 2018
1877-7821/ © 2017 Elsevier Ltd. All rights reserved.
G. Linder et al.
esophageal cancer, and patients with low socioeconomic status have an
increased risk of the disease, it is of essence to explore mechanisms of
how education, and other, corresponding factors influence these risks.
National registries have certain strengths in giving population based
data. Moreover, Sweden offers many nationwide data bases and by
linking the data, through a unique ten digit personal identification
number given to all Swedes at birth, it is possible to explore and
strengthen epidemiological findings.
On these grounds the present study aimed to investigate the re-
lationship between the patient’s educational level and respective
treatment allocation after the diagnosis of esophageal- and gastro-
esophageal junctional cancer and its subsequent impact on survival.
2. Materials and methods
2.1. Study design
This cohort study recruited data from 2006 to 2012 in the Swedish
National Register for Esophageal and Gastric Cancer (NREV). All
Swedish patients diagnosed with esophageal- and gastroesophageal
junctional cancer (ICD10; C15.*, C16.0A, C16.0B and C16.0X) during
the study years were included. Data from NREV was linked to the
Swedish Cancer Register, National Patient Register, Prescribed Drug
Register and the Cause of Death Register. All the above registries are
well described, researched and validated [10–14].
All data from the above-mentioned registers was extracted in 2014.
The study was approved by the local Ethics board in Stockholm, Dnr
2013/596-31/3.
2.2. Educational level – exposure
Data on education level was extracted from the Education Register,
administered by Statistics Sweden [15]. The educational level was di-
vided into three well-defined classes, as advocated by previous studies
[16], based on the highest attained number of years of schooling at the
time of data extraction. Low education level, ≤9years, corresponds to
compulsory primary school. Intermediate education level, 10–12 years
corresponds to completed primary school and partly or fully competed
secondary school. High education level, > 12 years, corresponds to
post-secondary education such as university or other academic studies.
2.3. Treatment allocation – outcome
In Sweden, patients diagnosed with esophageal- or gastro-
esophageal junctional cancers are reported to the NREV. The accuracy
of this register was recently evaluated and was in 91.1% of entries in
exact agreement with reabstracted data. The coverage of NREV was, in
the same validation study, 95.5% when compared to the Swedish
Cancer Register [12]. The intended choice of treatment (treatment al-
location), determined at a multidisciplinary treatment conference
(MDC) or by the treating physician, is a mandatory variable reported to
the register concluding the diagnostic workup. This variable has, in the
present register, been determined to be of high validity [12].
2.4. Covariates and confounding factors
A directed acyclic graph (DAG) model [17] was used to identify
possible factors governing outcome as well as confounding factors.
Covariates included well-established prognostic factors such as age, sex,
TNM-stage, histopathology and ASA-class (American Society of An-
esthesiologists) as well as possibly influencing comorbidities such as;
myocardial infarction (ICD10; J42-44) or chronic obstructive pul-
monary disease (COPD)-diagnosis (ICD10; I21-22) within 10 years be-
fore cancer diagnose. Other covariates included geographic region and
whether or not the patient was presented at a multidisciplinary treat-
ment conference.
92
Cancer Epidemiology 52 (2018) 91–98
2.5. Statistical methods
The effect of education level on the probability of allocation to
curative treatment was analyzed with univariate and multivariate lo-
gistic regression. In the survival analysis, patients were followed until
death, emigration or December 31 st 2013, whichever came first.
Survival probability was displayed according to the Kaplan- Meier
method and Cox proportional hazard models were used to assess the
effect of educational level when adjusting for multiple variables.
Separate models were fitted for each allocated treatment. Missing data
on histology, ASA class, TNM, multidisciplinary conference and edu-
cational level were handled by use of the Multivariate Imputation by
Chained Equations (MICE) algorithm [18]. Data were in these cases
imputed forty times using all variables as predictors in the MICE al-
gorithm. Separate models were fitted to the dataset in each iteration
and the results were then pooled using Rubin’s rules [19]. A p-value of
0.05 was considered significant. All statistical analyses were performed
with the R statistical software package [20].
3. Results
3.1. Patients
In total, 4112 patients were included in the analysis. Patient char-
acteristics as well as missing and imputed, data are summarized in
Table 1. Out of all patients, the duration of education was ≤9 years
(Low) in 1686 (41.0%), 10–12 years (Intermediate) in 1522 (37.0%)
and > 12 years (High) in 638 (15.5%). Data on the duration of educa-
tion were missing in 266 patients (6.5%). Patients with higher educa-
tion were slightly younger and geographically concentrated to the ca-
pital region and presented with less comorbidity. In all other aspects the
characteristics were similar in the three study groups.
3.2. Educational level and treatment allocation
Curative treatment was planned in 1587 patients (38.7%). In the
group with low educational level 36.0% were offered curative treat-
ment. For the intermediate and high educational level groups the per-
centages were 42.6% and 46.4% respectively. In a univariate model
both intermediate education, (OR 1.35, 95% CI 1.17–1.55), and high
education level, (OR 1.58, 95% CI 1.31–1.90) were associated with
increased probability of being offered curative treatment. When ap-
plying a multivariate model adjusting for region, sex, age, ASA-class,
TNM-stage, histopathology, myocardial infarction, COPD and multi-
disciplinary conference, the association between high educational level
and curative treatment still remained, with an adjusted OR of 1.48
(95% CI 1.08–2.03). The adherence to a multidisciplinary conference
for clinical decision-making, emerged as a robust positive predictor of
allocation to curative treatment, (adjusted OR 3.13, 95% CI 2.40–4.08)
as were also geographical region South and West. Negative predictors
for curative treatment allocation in the multivariate model were geo-
graphical region Central, positive N- and M-stage- disease, higher ASA-
class, squamous histopathology, COPD, and age > 70 years.
Corresponding data are presented in Table 2 and graphically in Fig. 1.
In an additional multivariate logistic regression model, stratified by sex,
age and histological subtype, the association between high education
level and allocation to curative treatment remained for men, (adjusted
OR 1.47, 95% CI 1.02–2.12), was strengthened in the elderly (> 70y),
(adjusted OR 1.61, 95% CI 1.06–2.44) and in patients with adeno-
carcinoma, (adjusted OR 1.73, 95% CI 1.13–2.64). The multi-
disciplinary conference as a predictor for curative treatment allocation
was confirmed for the three subgroups in the stratified analysis. Sup-
plementary Table 1.
G. Linder et al. Cancer Epidemiology 52 (2018) 91–98

Table 1
Characteristics of patients diagnosed with esophageal- and gastroesophageal juctional cancer in Sweden, 2006–2012, by level of education.

Education level

Low Intermediate High Missing Total

N = 1 686 N = 1 522 N = 638 N = 266 N = 4 112

Region, No (%)
Capital 252 (14.9) 301 (19.8) 194 (30.4) 46 (17.3) 793 (19.3)
Central 418 (24.8) 360 (23.7) 113 (17.7) 61 (22.9) 952 (23.2)
Southeast 202 (12.0) 138 (9.1) 57 (8.9) 25 (9.4) 422 (10.3)
South 354 (21.0) 303 (19.9) 130 (20.4) 51 (19.2) 838 (20.4)
West 286 (17.0) 246 (16.2) 97 (15.2) 50 (18.8) 679 (16.5)
North 174 (10.3) 174 (11.4) 47 (7.4) 33 (12.4) 428 (10.4)

Sex, No (%)
Male 1262 (74.9) 1131 (74.3) 473 (74.1) 189 (71.1) 3055 (74.3)
Female 424 (25.1) 391 (25.7) 165 (25.9) 77 (28.9) 1057 (25.7)

Age at diagnosis
Median (IQR) 73 (66–81) 67 (60–76) 68 (61–76) 74 (65–82) 70 (63–79)

ASA class, No (%)

1 244 (14.5) 344 (22.6) 173 (27.1) 30 (11.3) 791 (19.2)
2 674 (40.0) 611 (40.1) 257 (40.3) 78 (29.3) 1620 (39.4)
3 558 (33.1) 392 (25.8) 157 (24.6) 71 (26.7) 1178 (28.6)
4 121 (7.2) 98 (6.4) 31 (4.9) 30 (11.3) 280 (6.8)
5 1 (0.1) 3 (0.2) 2 (0.3) 4 (1.5) 10 (0.2)
Missing 88 (5.2) 74 (4.9) 18 (2.8) 53 (19.9) 233 (5.7)

Clinical T-stage, No (%)

Tis 75 (4.4) 73 (4.8) 24 (3.8) 6 (2.3) 178 (4.3)
T0 4 (0.2) 1 (0.1) 2 (0.3) 0 (0.0) 7 (0.2)
T1 72 (4.3) 68 (4.5) 33 (5.2) 7 (2.6) 180 (4.4)
T2 228 (13.5) 236 (15.5) 111 (17.4) 24 (9.0) 599 (14.6)
T3 641 (38.0) 595 (39.1) 275 (43.1) 76 (28.6) 1587 (38.6)
T4 213 (12.6) 193 (12.7) 85 (13.3) 53 (19.9) 544 (13.2)
TX 434 (25.7) 346 (22.7) 105 (16.5) 84 (31.6) 969 (23.6)
Missing 19 (1.1) 10 (0.7) 3 (0.5) 16 (6.0) 48 (1.2)

Clinical N-stage, No (%)

N0 642 (38.1) 563 (37.0) 233 (36.5) 57 (21.4) 1495 (36.4)
N1 568 (33.7) 560 (36.8) 260 (40.8) 105 (39.5) 1493 (36.3)
N2 75 (4.4) 110 (7.2) 46 (7.2) 12 (4.5) 243 (5.9)
N3 47 (2.8) 44 (2.9) 16 (2.5) 8 (3.0) 115 (2.8)
NX 331 (19.6) 230 (15.1) 78 (12.2) 66 (24.8) 705 (17.1)
Missing 23 (1.4) 15 (1.0) 5 (0.8) 18 (6.8) 61 (1.5)

Clinical M-stage, No (%)

M0 1012 (60.0) 943 (62.0) 395 (61.9) 97 (36.5) 2447 (59.5)
M1 461 (27.3) 437 (28.7) 201 (31.5) 107 (40.2) 1206 (29.3)
MX 188 (11.2) 124 (8.1) 35 (5.5) 42 (15.8) 389 (9.5)
Missing 25 (1.5) 18 (1.2) 7 (1.1) 20 (7.5) 70 (1.7)

Histopathology, No (%)
Adenocarcinoma 1039 (61.6) 953 (62.6) 413 (64.7) 168 (63.2) 2573 (62.6)
Squamous 484 (28.7) 418 (27.5) 169 (26.5) 76 (28.6) 1147 (27.9)
Other 147 (8.7) 140 (9.2) 51 (8.0) 21 (7.9) 359 (8.7)
Missing 16 (0.9) 11 (0.7) 5 (0.8) 1 (0.4) 33 (0.8)

Comorbidities – within 10 years of cancer, No (%)

Myocardial infarction 136 (8.1) 94 (6.2) 18 (2.8) 19 (7.1) 267 (6.5)
COPD 115 (6.8) 79 (5.2) 28 (4.4) 12 (4.5) 234 (5.7)

Multidisciplinary conference
Yes 1077 (63.9) 1051 (69.1) 474 (74.3) 95 (35.7) 2697 (65.6)
No 539 (32.0) 409 (26.9) 148 (23.2) 114 (42.9) 1210 (29.4)
Missing 70 (4.2) 62 (4.1) 16 (2.5) 57 (21.4) 205 (5.0)

Planned primary treatment, No (%)

Curative surgery 475 (28.2) 572 (37.6) 252 (39.5) 29 (10.9) 1328 (32.3)
Curative chemo + radiotherapy 97 (5.8) 55 (3.6) 35 (5.5) 5 (1.9) 192 (4.7)
Other curative treatment 33 (2.0) 21 (1.4) 9 (1.4) 4 (1.5) 67 (1.6)
Palliative oncological treatment 297 (17.6) 304 (20.0) 140 (21.9) 35 (13.2) 776 (18.9)
Other palliative treatment 453 (26.9) 357 (23.5) 114 (17.9) 88 (33.1) 1012 (24.6)
No tumour treatment 291 (17.3) 172 (11.3) 74 (11.6) 64 (24.1) 601 (14.6)
Missing 40 (2.4) 41 (2.7) 14 (2.2) 41 (15.4) 136 (3.3)

ASA-class, American society of anesthesiologists classification; COPD, chronic obstructive pulmonary disease.

93
G. Linder et al. Cancer Epidemiology 52 (2018) 91–98

Table 2
Odds ratios (OR) with 95% confidence limits (CI) from logistic regression models for the outcome planned curative treatment.

Proportion planned curative treatment in subgroup (%) Univariate Multivariate

OR (95% CI) p OR (95% CI) p

Education level
Low (36.8) 1.00 (ref.) – 1.00 (ref.) –
Intermediate (43.8) 1.35 (1.17–1.55) < 0.001 1.05 (0.83–1.33) 0.666
High (47.4) 1.58 (1.31–1.90) < 0.001 1.48 (1.08–2.03) 0.014

Sex
Male (41.4) 1.00 (ref.) – 1.00 (ref.) –
Female (35.7) 0.79 (0.68–0.91) 0.002 0.82 (0.65–1.05) 0.114

Age at diagnosis
< 60 (56.3) 1.00 (ref.) – 1.00 (ref.) –
60–64 (50.9) 0.80 (0.64–1.01) 0.061 0.84 (0.57–1.26) 0.406
65–69 (49.4) 0.76 (0.61–0.94) 0.010 0.69 (0.48–1.00) 0.053
70–74 (46.5) 0.68 (0.54–0.84) < 0.001 0.54 (0.37–0.79) 0.001
75–80 (35.1) 0.42 (0.33–0.53) < 0.001 0.35 (0.23–0.51) < 0.001
80+ (12.9) 0.12 (0.09–0.15) < 0.001 0.06 (0.04–0.09) < 0.001

Multidisciplinary conference
No (21.7) 1.00 (ref.) – 1.00 (ref.) –
Yes (48.5) 3.42 (2.93–4.00) < 0.001 3.13 (2.40–4.08) < 0.001

Multivariate analysis is adjusted for Geographical region, Clinical TNM-stage, ASA-class, Histopathological subtype and Myocardial infarction and/or COPD (10 years prior to diagnosis of
cancer), in addition to variables presented above.
Complete analysis is available in Supplementary Table 2.
ASA-class, American society of anesthesiologists classification; COPD, chronic obstructive pulmonary disease.

3.3. Education level and survival
In the 1587 patients allocated to curative treatment, a multivariate
Cox regression model utilizing as outcome death from any cause, re-
vealed that patients with high education level had better survival (HR
0.82, 95% CI 0.69–0.99) as compared to patients with short education
(≤9 years). Early TNM-stage, lower age and lower ASA-class as well as
no COPD diagnosis were all independently associated with improved
survival.
The presentation of the case at a multidisciplinary conference was a
solid predictor of improved survival in both the 1788 patients allocated
to palliative treatment (HR 0.73, 95% CI 0.66–0.82) as well as in those
737 patients allocated to no treatment at all (HR 0.69, 95% CI
0.56–0.85). Table 3.
3.4. Overall survival
Among all patients in the cohort with known educational status
(n = 3846), the unadjusted 5-year survival was 13.3% (95% CI
11.6–15.4) for patients with short education, 17.7% (95% CI
15.6–20.0) for patients with intermediate and 19.7% (95% CI
16.4–23.8) for patients with long duration of education, log rank test, p-
value < 0.001, as visualized in Fig. 2.
4. Discussion
It can be assumed that the educational level covaries with the so-
cioeconomic status. However, solely educational level has not, until
now, been focused on in the context of predicting the treatment choice
in patients with esophageal- or gastroesophageal junctional cancer. A
subgroup analysis of all patients allocated to curative treatment also
demonstrated improved survival for those patients with longer educa-
tion. Moreover, patients only amenable for palliative treatment, or no
treatment at all, still fared better than their counterparts if presented
and discussed at a multidisciplinary conference.
Allocation of esophageal- or gastroesophageal junctional cancer to
curative or palliative treatment is based on tumor and patient related
factors; most importantly tumor stage, comorbidity and physical fitness.
Fit patients with localized disease are most commonly offered a
combination of neoadjuvant CRT followed by esophagectomy. Patients
that are not so fit but with a localized tumor may still be able to receive
curative treatment by surgery alone or definite CRT. Unfit patients, or
patients with severe comorbidity or very advanced tumor are normally
offered palliative treatment. Treatment allocation also depends on the
decision of the physician, a decision that may be better balanced in a
multidisciplinary setting, and on the preference of the patient.
If the results from the present study are interpreted correctly they
should raise questions on the equality of healthcare provided at hos-
pitals throughout Sweden. In the Netherlands, with a different health
care system than in Sweden, it seems as if curative treatment is offered
in greater extent to esophageal cancer patients with high socioeconomic
status [21]. Launay et al. studied patients with esophageal cancer living
in socioeconomically deprived areas and found shorter survival in those
residing in the most deprived areas [22]. Likewise data in patients with
pancreatic cancer, living in areas characterized by high socioeconomic
standards, revealed that those patients were more likely to undergo
surgical resection at University hospitals [23]. Similarly, results have
been presented to suggest a difference in the quality and standard of
treatment as well as in outcomes, in those patients with low income and
low socioeconomic status suffering from breast cancer [24] or prostate
cancer [25].
In an effort to separate the key elements confined to low socio-
economic status, the present study focuses on the patient’s educational
level (as reflected by the length of the well-defined education), rather
than on the patient’s place of residence or income, and these factor’s
possible impact on the intended treatment. The following mechanisms
are proposed to explain the impact of education on the selection of
treatment and outcomes. First, there is the patients readiness to detect,
accept and understand the signs and symptoms of the disease and
submit to the consequences thereof, potentially leading to effects on the
tumor stage at the time of the diagnosis. Attempts were made to adjust
for other corresponding imbalances in established prognostic risk fac-
tors, already present at the time of the diagnosis.
Second, the issue of communication between patient and doctor
plays a pivotal role [26] where focus must be to guide the patient to-
wards an understanding of the underlying disease and possible treat-
ments. It is unprofessional to administer CRT and/or to perform de-
manding surgery, altering the patient’s ability to ingest food, on an ill

94
G. Linder et al.
Fig. 1. Graphic presentation of Odds ratios (OR) with 95% confidence limits (CI) from multivariate logistic regression model for the outcome planned curative treatment.
informed patient. The decisions regarding treatment must be made in
mutual agreement between the physician in charge and the patient. A
recent interview-study by Lineback et al., covering 80 patients with
esophageal cancer, concluded that communication difficulties and lack
of understanding of treatment were more commonly reported in the
lower socioeconomic status groups. They also reported that patients
with low socioeconomic status were infrequently offered surgical
treatment [27].
Patients with a higher educational level are possibly more receptive
to information and psychologically and intellectually more ready to
understand and grasp the concept of the multimodal treatment, the
risks of not adhering to plans and integrate programs and treatment
ordinations. This is in line with recent findings regarding an increased
compliance in patients with high education level to screening programs
for breast- and cervical cancer [28]. On the other hand, in patients with
95
Cancer Epidemiology 52 (2018) 91–98
limited education, the need for well-tailored information might be even
larger, emphasizing the necessity for the physician in charge to explain
the disease-characteristics and management concepts together with
treatment options before any decision on future treatment can be made.
Another finding in the present study was the beneficial effects of the
presentation of each individual case to a multidisciplinary conference.
In this cohort of 4112 patients, it was apparent that patients who had
their case presented at a multidisciplinary conference had an associated
higher probability of being offered curative treatment. This is probably
largely due to selection bias, in the sense that patients amenable for
surgery are more likely to be presented at a multidisciplinary con-
ference than patients with metastatic disease. Another mechanism for
selection bias could be diagnosis and treatment of patients at lower
surgical volume county hospitals, not utilizing multidisciplinary con-
ferences, before the implementation of centralization of esophageal-
 G. Linder et al.

Table 3
Hazard ratios (HR) with 95% confidence limits (CI) from Cox proportional hazards models for the outcome death from any cause. Separate models were fitted for each planned treatment.

Curative treatment planned Palliative treatment planned No tumor treatment planned

Univariate Multivariate Univariate Multivariate Univariate Multivariate

HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p

Education level
Low 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) –
Intermediate 0.88 (0.76–1.01) 0.064 0.95 (0.82–1.10) 0.497 0.97 (0.88–1.08) 0.630 1.01 (0.90–1.13) 0.855 0.93 (0.76–1.12) 0.433 1.06 (0.84–1.34) 0.600
High 0.83 (0.69–0.99) 0.039 0.82 (0.69–0.99) 0.036 0.97 (0.84–1.12) 0.693 1.05 (0.90–1.23) 0.498 0.98 (0.76–1.28) 0.895 1.10 (0.81–1.49) 0.539

Sex
Male 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) –
Female 0.92 (0.79–1.07) 0.294 0.95 (0.81–1.11) 0.496 1.00 (0.90–1.12) 0.985 0.95 (0.85–1.07) 0.381 1.03 (0.86–1.24) 0.751 1.12 (0.91–1.37) 0.295

96
Age at diagnosis
< 60 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) –
60–64 1.27 (1.04–1.55) 0.018 1.20 (0.98–1.47) 0.081 0.88 (0.73–1.06) 0.184 0.85 (0.70–1.03) 0.089 1.98 (1.25–3.14) 0.004 1.17 (0.71–1.92) 0.546
65–69 1.18 (0.97–1.43) 0.094 1.16 (0.95–1.42) 0.155 0.96 (0.80–1.14) 0.608 0.93 (0.78–1.12) 0.439 1.45 (0.93–2.24) 0.100 1.03 (0.63–1.67) 0.910
70–74 1.35 (1.10–1.64) 0.004 1.30 (1.05–1.61) 0.016 1.18 (0.99–1.42) 0.069 1.16 (0.95–1.41) 0.137 1.50 (0.99–2.28) 0.056 1.08 (0.68–1.71) 0.751
75–80 1.35 (1.08–1.69) 0.008 1.31 (1.04–1.65) 0.024 1.14 (0.96–1.37) 0.140 1.12 (0.92–1.35) 0.265 1.32 (0.88–1.98) 0.176 0.93 (0.58–1.48) 0.752
80+ 1.57 (1.22–2.01) < 0.001 1.78 (1.36–2.34) < 0.001 1.16 (0.99–1.35) 0.069 1.24 (1.04–1.49) 0.016 1.61 (1.12–2.32) 0.011 0.91 (0.58–1.40) 0.656

Multidisciplinary conference
No 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) – 1.00 (ref.) –
Yes 1.27 (1.07–1.51) 0.007 0.97 (0.80–1.18) 0.795 0.74 (0.67–0.82) < 0.001 0.73 (0.66–0.82) < 0.001 1.06 (0.88–1.27) 0.556 0.69 (0.56–0.85) < 0.001

Multivariate analysis is adjusted for Geographical region, Clinical TNM-stage, ASA-class, Histopathological subtype and Myocardial infarction and/or COPD (10 years prior to diagnosis of cancer) in addition to variables presented above. Complete
analysis is available in Supplementary Table 3.
ASA-class, American society of anesthesiologists classification; COPD, chronic obstructive pulmonary disease.
Cancer Epidemiology 52 (2018) 91–98
G. Linder et al.
and gastroesophageal junctional cancers to university hospitals was in
effect. The solid result in the multivariate model of the present study,
however, merits attention. The advantages of a multidisciplinary con-
ference are vigorously advocated but less is published in support of its
positive effects on improved clinical outcomes. One previous study
have shown improved timeliness to surgery, adherence to guidelines
and increased probability of a complete staging as a direct effect of the
introduction of a multidisciplinary conference [29]. Furthermore, the
present study demonstrated that, in patients allocated to palliative
treatment or no treatment at all, a multidisciplinary conference was
associated with increased survival. It could of course be argued that
palliative patients discussed multidisciplinary having better outcomes
reflects the benefits of having multiple disciplines available for these
patients, rather than true treatment changes because of a multi-
disciplinary team meeting. Taking that into consideration, these find-
ings still add to the support, and usefulness, of the multidisciplinary
conference.
As in all observational studies the risk of confounding has to be
carefully appreciated. In order to minimize the risk of this error a DAG
analysis was carried out beforehand with the ambition to address all
possible confounders and adjust for those. Among possible confounders
both alcohol consumption and tobacco use increase the risk of devel-
oping esophageal- and gastroesophageal junctional cancer [30,31] and
are more common in those representing low socioeconomic status [32].
Data on smoking is unfortunately incomplete for the included years
with accurate information on tobacco use missing in 63.6% of all pa-
tients in the cohort. The amount of missing information on tobacco use
was determined too high for meaningful imputation according to the
MICE-algorithm. However, in an earnest attempt to address the issue,
the registered diagnoses of COPD and myocardial infarction within 10
years of diagnosis of esophageal- or gastroesophageal junctional cancer
was utilized as a proxy in the multivariable models. There was no
collected data on use of alcohol, physical fitness, or dietary habits in the
97
Cancer Epidemiology 52 (2018) 91–98
Fig. 2. Kaplan-Meier curve depicting overall survival by education level.
Log rank test, p < 0.001.
cohort, exposures that may be associated with both educational level
and treatment allocation. Thus there might be some residual con-
founding.
In determining the exposure, duration of education, there is a risk of
underestimating the impact of alternative forms of education. Informal
education may not be reported as regular schooling why some patients
may in fact harbor a higher level of education than registered in our
database. This implies a situation where some patients would have
longer education than officially reported. However, this would only
dilute the results in the lower educational group further, thus driving
the results towards the null.
Some patients, although reported in the register as allocated to
curative treatment, never actually receive curative treatment but are
handled as “intention-to-treat” in this study. This is a strength when
drawing conclusions from the survival analysis.
The follow-up time, forming the base for survival analysis, ranged
from 1 to 8 years. There were no significant changes in proportions of
the exposure, education level, over the study period (data not shown).
The choice to include patients in this cohort on the basis criteria that
they have been diagnosed with esophageal- or gastroesophageal junc-
tional cancers, not limiting the study only to patients that have un-
dergone treatment is a strength of the present study, differing it from
other cohort studies. Other mentionable strengths of the study, con-
stituting more than 95% of all patients diagnosed with esophageal- or
gastroesophageal junctional cancers in Sweden during the study period,
are the population-based design, great number of patients included and
the robust data from well-validated registers minimizing selection bias
and misclassification.
In conclusion, this large nationwide cohort study demonstrate that
high education level, as reflected by long duration of schooling, is as-
sociated with a greater probability of being offered curative treatment
for esophageal- and gastroesophageal junctional cancers. High educa-
tion level is also associated with improved survival for these patients. In
G. Linder et al.
patients ineligible for curative treatment and instead planned for pal-
liative treatment or no treatment at all, the present findings support
that they still fare better if their case is presented at a multidisciplinary
conference.
Authorship contribution statement
All authors have made substantial contributions to conception and
design, acquisition of data as well as analysis and interpretation of data
for this manuscript.
All authors have made substantial contributions in drafting and
critically revising the article as well as having read and approved the
final version of the manuscript to be published.
Role of the funding source
Partial financial support for this study was provided by the Bengt
Ihre research fund, author JH, and by the Lennanders research fund,
author GL.
The funding sources had no role in study design, concept, analysis
or interpretation.
Conflict of interest statement
The authors declare no conflict of interests.
Acknowledgements
Because of the sensitive nature of the data collected for this study,
requests to access the dataset from qualified researchers trained in
human subject confidentiality protocols may be sent to the corre-
sponding author.
The authors would like to thank all individuals reporting patient
data to the Swedish National Register for Esophageal and Gastric
Cancer.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.canep.2017.12.008.
References
[1] L.A. Torre, F. Bray, R.L. Siegel, J. Ferlay, J. Lortet-Tieulent, A. Jemal, Global cancer
statistics, 2012, CA. Cancer J. Clin. 65 (2) (2015) 87–108.
[2] R. Baastrup, M. Sorensen, J. Hansen, R.D. Hansen, H. Wurtzen, J.F. Winther, Social
inequality and incidence of and survival from cancers of the oesophagus, stomach
and pancreas in a population-based study in Denmark, 1994–2003, Eur. J. Cancer
44 (14) (2008) 1962–1977.
[3] J. Lagergren, G. Andersson, M. Talback, et al., Marital status, education, and income
in relation to the risk of esophageal and gastric cancer by histological type and site,
Cancer 122 (2) (2016) 207–212.
[4] N. Brusselaers, F. Mattsson, M. Lindblad, J. Lagergren, Association between edu-
cation level and prognosis after esophageal cancer surgery: a Swedish population-
based cohort study, PLoS One 10 (3) (2015) e0121928.
[5] P. van Hagen, M.C. Hulshof, J.J. van Lanschot, et al., Preoperative chemor-
adiotherapy for esophageal or junctional cancer, N. Engl. J. Med. 366 (22) (2012)
2074–2084.
[6] M. Scarpa, S. Valente, R. Alfieri, et al., Systematic review of health-related quality
of life after esophagectomy for esophageal cancer, World J. Gastroenterol. 17 (42)
(2011) 4660–4674.
98
Cancer Epidemiology 52 (2018) 91–98
[7] M. Jacobs, R.C. Macefield, R.G. Elbers, et al., Meta-analysis shows clinically re-
levant and long-lasting deterioration in health-related quality of life after esopha-
geal cancer surgery, Qual. Life Res.: Int. J. Qual. Life Asp. Treat. Care Rehabil. 23
(4) (2014) 1097–1115.
[8] M.M. Boniface, S.B. Wani, T.E. Schefter, et al., Multidisciplinary management for
esophageal and gastric cancer, Cancer Manage. Res. 8 (2016) 39–44.
[9] R.M. Quiros, C.L. Bui, Multidisciplinary approach to esophageal and gastric cancer,
Surg. Clin. North Am. 89 (1) (2009) 79–96 viii.
[10] L. Barlow, K. Westergren, L. Holmberg, M. Talback, The completeness of the
Swedish Cancer Register: a sample survey for year 1998, Acta Oncol. 48 (1) (2009)
27–33.
[11] B. Wettermark, N. Hammar, C.M. Fored, et al., The new Swedish Prescribed Drug
Register-opportunities for pharmacoepidemiological research and experience from
the first six months, Pharmacoepidemiol. Drug Saf. 16 (7) (2007) 726–735.
[12] G. Linder, M. Lindblad, P. Djerf, et al., Validation of data quality in the Swedish
national register for oesophageal and gastric cancer, Br. J. Surg. (2016).
[13] J.F. Ludvigsson, E. Andersson, A. Ekbom, et al., External review and validation of
the Swedish national inpatient register, BMC Public Health 11 (2011) 450.
[14] L.A. Johansson, R. Westerling, Comparing Swedish hospital discharge records with
death certificates: implications for mortality statistics, Int. J. Epidemiol. 29 (3)
(2000) 495–502.
[15] LISA-database. Statistics Sweden, 2012. http://www.scb.se/sv_/Vara-tjanster/
Bestalla-mikrodata/Vilka-mikrodata-finns/Longitudinell-integrationsdatabas-for-
sjukforsakrings-och-arbetsmarknadsstudier-LISA/.
[16] E. Backlund, P.D. Sorlie, N.J. Johnson, A comparison of the relationships of edu-
cation and income with mortality: the National Longitudinal Mortality Study, Soc.
Sci. Med. (1982) 49 (10) (1999) 1373–1384.
[17] T.J. VanderWeele, M.A. Hernan, J.M. Robins, Causal directed acyclic graphs and
the direction of unmeasured confounding bias, Epidemiology (Cambridge, Mass.)
19 (5) (2008) 720–728.
[18] S. Van Buuren, Flexible Imputation of Missing Data, CRC press, 2012.
[19] D.B. Rubin, Multiple Imputation for Nonresponse in Surveys, Wiley, New York,
1987.
[20] R. A Language and Environment for Statistical Computing. [computer Program], R
Foundation for Statistical Computing, 2014.
[21] P. Bus, M.J. Aarts, V.E. Lemmens, et al., The effect of socioeconomic status on
staging and treatment decisions in esophageal cancer, J. Clin. Gastroenterol. 46 (10)
(2012) 833–839.
[22] L. Launay, O. Dejardin, C. Pornet, et al., Influence of socioeconomic environment on
survival in patients diagnosed with esophageal cancer: a population-based study,
Dis. Esophagus 25 (8) (2012) 723–730.
[23] M.H. van Roest, M.A. van der Aa, L.G. van der Geest, K.P. de Jong, The impact of
socioeconomic status, surgical resection and type of hospital on survival in patients
with pancreatic cancer. a population-based study in The Netherlands, PLoS One 11
(11) (2016) e0166449.
[24] C.J. Bradley, C.W. Given, C. Roberts, Race, socioeconomic status, and breast cancer
treatment and survival, J. Natl. Cancer Inst. 94 (7) (2002) 490–496.
[25] I. Cheng, J.S. Witte, L.A. McClure, et al., Socioeconomic status and prostate cancer
incidence and mortality rates among the diverse population of California, Cancer
Causes Control 20 (8) (2009) 1431–1440.
[26] R. Ruiz-Moral, E. Perez Rodriguez, L.A. Perula de Torres, J. de la Torre, Physician-
patient communication: a study on the observed behaviours of specialty physicians
and the ways their patients perceive them, Patient Educ. Couns. 64 (1–3) (2006)
242–248.
[27] C.M. Lineback, C.M. Mervak, S.L. Revels, M.T. Kemp, R.M. Reddy, Barriers to ac-
cessing optimal esophageal cancer care for socioeconomically disadvantaged pa-
tients, Ann. Thorac. Surg. 103 (2) (2016) 416–421.
[28] G. Damiani, D. Basso, A. Acampora, et al., The impact of level of education on
adherence to breast and cervical cancer screening: evidence from a systematic re-
view and meta-analysis, Prev. Med. 81 (2015) 281–289.
[29] R.K. Freeman, J.M. Van Woerkom, A. Vyverberg, A.J. Ascioti, The effect of a
multidisciplinary thoracic malignancy conference on the treatment of patients with
esophageal cancer, Ann. Thorac. Surg. 92 (4) (2011) 1239–1242 discussion 1243.
[30] J. Lagergren, R. Bergstrom, A. Lindgren, O. Nyren, The role of tobacco, snuff and
alcohol use in the aetiology of cancer of the oesophagus and gastric cardia, Int. J.
Cancer 85 (3) (2000) 340–346.
[31] M. Morita, R. Kumashiro, N. Kubo, et al., Alcohol drinking, cigarette smoking, and
the development of squamous cell carcinoma of the esophagus: epidemiology,
clinical findings, and prevention, Int. J. Clin. Oncol. 15 (2) (2010) 126–134.
[32] G.E. Nagelhout, D. de Korte-de Boer, A.E. Kunst, et al., Trends in socioeconomic
inequalities in smoking prevalence, consumption, initiation, and cessation between
2001 and 2008 in the Netherlands. Findings from a national population survey,
BMC Public Health 12 (2012) 303.