Estimating excess mortality in people with cancer and multimorbidity in the COVID-19 emergency

Abstract

Predicted estimate of excess deaths in cancer patients related to the Covid-19 emergency. Data from England, Northern Ireland and US.

Full text

Estimating excess mortality in people with
cancer and multimorbidity in the COVID-19
emergency
Alvina G. Lai, Ph.D.1,2, , Laura Pasea, Ph.D.1,2* , Amitava Banerjee, DPhil1,2,3*, Spiros Denaxas, Ph.D.1,2,6,7, Michail Katsoulis,
Ph.D.1,2, Wai Hoong Chang, MSc1,2, Bryan Williams, Ph.D.4,5,6, Deenan Pillay, Ph.D.8, Mahdad Noursadeghi, Ph.D.8, David
Linch, FMedSci6,9, Derralynn Hughes, FRCPath10,11, Martin D. Forster, Ph.D.4,10, Clare Turnbull, Ph.D.12, Natalie K. Fitzpatrick,
MSc1,2, Kathryn Boyd, MD13 , Graham R. Foster, Ph.D.14 , DATA-CAN15, Matt Cooper, Ph.D.15, Monica Jones, PGDip15, Kathy
Pritchard-Jones, FMedSci15,16,17,18, Richard Sullivan, Ph.D.19, Geoff Hall, Ph.D.15,20,21, Charlie Davie, FRCP11,15,16, Mark
Lawler, Ph.D.15,22 , and Harry Hemingway, FMedSci1,2,6,
1Institute of Health Informatics, University College London, London, UK
2Health Data Research UK, University College London, London, UK
3Barts Health NHS Trust, The Royal London Hospital, London, UK
4University College London Hospitals NHS Trust, London, UK
5Institute of Cardiovascular Science, University College London, London, UK
6University College London Hospitals NIHR Biomedical Research Centre, London, UK
7The Alan Turing Institute, London, UK
8Division of Infection and Immunity, University College London, London, UK
9Department of Hematology, University College London Cancer Institute, London, UK
10University College London Cancer Institute, London, UK
11Royal Free NHS Foundation Trust, London, UK
12Division of Genetics and Epidemiology, Institute of Cancer Research, London, UK
13Northern Ireland Cancer Network, UK
14Barts Liver Centre, Blizard Institute, Queen Mary University of London, London, UK
15DATA-CAN, Health Data Research UK hub for cancer hosted by UCLPartners, London, UK
16UCLPartners Academic Health Science Partnership, London UK
17Centre for Cancer Outcomes, University College London Hospitals NHS Foundation Trust, London, UK
18UCL Great Ormond Street Institute for Child Health, University College London, London, UK
19Conflict and Health Research Group, Institute of Cancer Policy, King’s College London, London, UK
20Leeds Institute of Medical Research, University of Leeds, Leeds, UK
21School of Medicine, University of Leeds, Leeds, UK
22Patrick G Johnston Centre for Cancer Research, Queen’s University Belfast, UK
*Joint second authors
Background: Cancer and multiple non-cancer conditions are
considered by the Centers for Disease Control and Prevention
(CDC) as high risk conditions in the COVID-19 emergency.
Professional societies have recommended changes in cancer
service provision to minimize COVID-19 risks to cancer
patients and health care workers. However, we do not know
the extent to which cancer patients, in whom multi-morbidity
is common, may be at higher overall risk of mortality as
a net result of multiple factors including COVID-19 infec-
tion, changes in health services, and socioeconomic factors.
Methods: We report multi-center, weekly cancer diagnostic
referrals and chemotherapy treatments until April 2020 in
England and Northern Ireland. We analyzed population-
based health records from 3,862,012 adults in England to
estimate 1-year mortality in 24 cancer sites and 15 non-
cancer comorbidity clusters (40 conditions) recognized by
CDC as high-risk. We estimated overall (direct and indirect)
effects of COVID-19 emergency on mortality under differ-
ent Relative Impact of the Emergency (RIE) and different
Proportions of the population Affected by the Emergency
(PAE). We applied the same model to the US, using Surveil-
lance, Epidemiology, and End Results (SEER) program data.
Results: Weekly data until April 2020 demonstrate significant
falls in admissions for chemotherapy (45-66% reduction)
and urgent referrals for early cancer diagnosis (70-89%
reduction), compared to pre-emergency levels. Under con-
servative assumptions of the emergency affecting only people
with newly diagnosed cancer (incident cases) at COVID-
19 PAE of 40%, and an RIE of 1.5, the model estimated
6,270 excess deaths at 1 year in England and 33,890 excess
deaths in the US. In England, the proportion of patients
with incident cancer with ≥1 comorbidity was 65.2%. The
number of comorbidities was strongly associated with can-
cer mortality risk. Across a range of model assumptions,
and across incident and prevalent cancer cases, 78% of
excess deaths occur in cancer patients with ≥1 comorbidity.
Conclusion: We provide the first estimates of potential excess
mortality among people with cancer and multimorbidity due to
the COVID-19 emergency and demonstrate dramatic changes
in cancer services. To better inform prioritization of cancer care
and guide policy change, there is an urgent need for weekly data
on cause-specific excess mortality, cancer diagnosis and treat-
ment provision and better intelligence on the use of effective
treatments for comorbidities.
Correspondence: alvina.lai@ucl.ac.uk | h.hemingway@ucl.ac.uk
Introduction
The excess risk of death in people living with cancer
during the COVID-19 emergency may be due not only
Lai et al. | April 28, 2020 | 1–10

to COVID-19 infection, but also to the unintended health
consequences of changes in health service provision, the
physical or psychological effects of social distancing, and
economic upheaval. Recent national evidence suggests that
there is an excess of deaths, both in those infected with
SARS-CoV-2, and in those with no infection(1). Optimal
cancer care must balance the need to protect patients from
COVID-19 infection, with continued access to early diag-
nosis and potentially curative treatment(2,3). Professional
associations in the US(4–6), UK(7–9) and Europe(9) have
recommended revising cancer care activities for triaging of
systemic anti-cancer treatment, surgery and risk-adapted
radiotherapy(10). All elective surgery has been postponed in
the UK(8,9). In the US, more than a quarter of patients with
cancer reported a delay in their cancer treatment in April
2020 because of COVID-19(4). Despite these recommen-
dations for changes in healthcare services, there is a lack of
near real-time data quantifying the extent of disruption due
to reconfiguration in service delivery for cancer patients.
Existing publications on cancer and COVID-19 have been
limited to small case series(11,12). The US Centers
for Disease Control and Prevention (CDC) and Pub-
lic Health England (PHE) have identified patients with
specific malignant and non-malignant conditions as at
greater risk of developing severe illness from COVID-19
exposure(13–15). Patients with cancer commonly have other
conditions considered to further increase their COVID-19
risk; multimorbidity in cancer is an increasing clinical
concern(16,17). Oncologists, internists and family physi-
cians need to balance the requirement to encourage people
to socially isolate, with their need to access hospital services
to ensure the most effective cancer care. There is a lack
of evidence on pan-cancer estimation of mortality risks
according to type and number of multimorbid conditions.
Our objectives were to: 1) quantify changes in cancer care
activities in near real-time from weekly multi-center hospital
data; 2) estimate the prevalence, incidence and background
(pre-COVID-19) 1-year mortality risk across 24 site-specific
cancers; 3) estimate the prevalence of 15 CDC- and PHE-
relevant co-occurring condition clusters (multimorbidity) by
cancer site and their association with excess mortality and
4) estimate overall excess deaths due to the COVID-19 pan-
demic over a 1-year period, based on different Proportion of
the population Affected by the Emergency (PAE) and Rela-
tive Impact of the Emergency (RIE), using a previously pub-
lished model(18).
Methods
Weekly information on cancer care. Through DATA-
CAN, the UK National Health Data Research Hub
for Cancer(19), we obtained weekly returns for urgent
cancer referrals for early diagnosis and chemother-
apy attendances (from 2018 to most current data)
from hospitals in Leeds, London and Northern Ireland.
Patient cohort. We used population-based electronic health
records in England from primary care data linked to the
Office for National Statistics(ONS) death registration, using
the open-access CALIBER resource(20,21). The study
population consisted of 3,862,012 adults aged ≥30 years,
registered with a general practice from 1 January 1997 to
1 January 2017 with at least one year of follow-up data.
Electronic health record phenotype definitions of diseases
and risk factors relevant to COVID-19 are available at
https://caliberresearch.org/portal and have previously been
validated(22–25). Further details are in supplementary
methods. In brief, phenotypes are based on hospital and
primary care information recorded in primary care, using
the Read clinical terminology(version 2). We defined
non-fatal, incident and prevalent cases of cancer across
24 primary cancer sites according to previously validated
CALIBER electronic health record phenotypes, which
included: biliary tract, bladder, bone, brain, breast, cervix,
colon-rectum, Hodgkin’s lymphoma, kidney, leukemia, liver,
lung, melanoma, multiple myeloma, non-Hodgkin’s lym-
phoma, esophagus, oropharynx, ovary, pancreas, prostate,
stomach, testis, thyroid and uterus(26). We defined cancers
as prevalent (diagnosed at any time prior to baseline) and
incident if they occurred during follow-up. We defined
15 comorbid conditions or condition clusters involving 40
individual conditions defined by the Centers for Disease
Control (CDC) or Public Health England (PHE) as associated
with poor outcomes caused by severe COVID-19 infection.
Analyses. For full details of the analyses, see supplementary
methods. Briefly, we estimated incidence rates per 100,000
person years and 1-year mortality in our study population.
Estimates were used to calculate the excess deaths in cancer
alone and cancer plus comorbidities, based on three levels
of PAE (10%, 40% and 80%) and four RIE scenarios asso-
ciated with COVID-19(1.2, 1.5, 2.0 and 3.0) (additional de-
tails on plausible choice of PAE and RIE are presented in
the discussion). We presented all results in Figures 2, 5, S4
and S8, but in order to represent a plausible, likely conser-
vative, scenario for the current emergency, we chose to high-
light data for a PAE of 40% and a RIE of 1.5, with a range
of 1.2–2 in the results text section of the manuscript. We
applied our model to US population using publicly available
data from the US from the Surveillance, Epidemiology, and
End Results (SEER) program on incidence and 1 year mor-
tality (taken as 1 minus the reported survival(27)). We con-
sidered mortality data from SEER for individuals aged 40-64,
65-74 and 75+.
Results
Real-time weekly hospital data for urgent cancer re-
ferrals and chemotherapy attendances in England and
Northern Ireland. We observed that a majority of patients
with cancer or suspected cancer are not accessing healthcare
services. Major declines in chemotherapy attendances (45%,
66%, 66% and 63% reduction; average=60%) and urgent
2 Lai et al. | Excess deaths in cancer and multimorbidity

Royal Free (England)
25
50
75
100
125
Month
% Relative to the 2019 average for each hospital
Chemotherapy
Leeds (England)
25
50
75
100
125
Urgent referral for early diagnosis
25
50
75
100
125
Five Northern Ireland HSCs
28-Feb-2020
25
50
75
100
125
2018 2019 2020 2018 2019 2020
J F M A M J J A S O N D J F M A M J J A S O N D J F M J F M A M J J A S O N D J F M A M J J A S O N D J F M
Christmas First case of COVID-19
UCLH (England)
28-Feb-2020
11-Feb-2020
11-Feb-2020
11-Feb-2020
11-Feb-2020
Easter
29-Feb-202029-Feb-2020
Fig. 1. Weekly hospital data on changes in urgent referrals and chemotherapy clinic attendance from eight hospitals in the UK. The data for Norther n Ireland includes
five Health and Social Care Trusts (HSCs) that cover all health service provision in Northern Ireland: Belfast HSC, Northern HSC, South Eastern HSC, Southern HSC and
Western HSC.
cancer referrals for early diagnosis (70%, 74%, 89% and
71% reduction; average=76%) were observed in eight hospi-
tals across the UK: England (Leeds Teaching Hospitals NHS
Trust, Royal Free Hospital and University College London
Hospitals) and Northern Ireland (all five Health and Social
Care Trusts) respectively, compared to pre-emergency levels
(Figure 1). These data signal the need to identify patient-
specific risk factors (cancer and multimorbidity) to facilitate
prioritization of service provision during the pandemic.
Prevalence, incidence and background (pre-COVID-19)
1-year mortality for 24 cancers. The overall prevalence of
any of the 24 cancers was 3.1% (117,978/3,862,012) in CAL-
IBER (England), (Table S1). Prevalence of the 24 cancers by
age, sex and year is shown in Figure S1. The age-adjusted
incidence rates of the 24 cancers estimated in CALIBER
was 635 per 100 000 (compared with estimates from In-
ternational Agency for Research on Cancer (IARC, UK) of
590 per 100 000) (Figure S2A). The 1-year mortality across
24 cancer sites was, as anticipated, higher among incident
cases of cancer, compared to prevalent cases (Figure S3).
Estimates of excess COVID-19-related deaths by cancer
site in England. When considering incident cancers in
England, at COVID-19 PAE of 40%, we estimated 2,509,
6270 and 12,543 excess deaths at RIE of 1.2, 1.5, and 2
respectively (Figure 2A). Higher numbers were seen among
prevalent cancers than incident cancers when estimating
absolute excess deaths, since the number of prevalent cases
were higher than the number of incident cases observed over
1 year, e.g., for PAE 40%, we observed 11,645 excess deaths
(range 4,655-23,287) at RIE of 1.5 (range 1.2–2) (Figure
S4). When considering both prevalent and incident cancers
together at COVID-19 PAE of 40%, we estimated 17,915 ex-
cess deaths (range 7,164-35,830) at RIE of 1.5 (range 1.2–2).
Estimates of excess COVID-19-related deaths by cancer
site for incident cases in US. We found broadly comparable
Lai et al. | Excess deaths in cancer and multimorbidity 3

Excess deaths across all 24 cancers
6270
3137
1567
628
25083
12543
6270
2509
50163
25083
12543
5018
Proportion of the population who are affected (PAE) by the COVID-19 emergency
A
Relative impact of the emergency (RAE)
14
22
3
63
11
2
76
2
9
24
14
186
11
10
21
53
10
6
52
14
20
0
2
3
35
54
9
157
28
5
189
5
22
59
34
466
27
26
54
131
26
15
130
35
49
0
4
7
70
109
17
315
56
10
378
10
45
118
69
932
53
51
107
263
52
30
261
70
99
0
8
14
140
218
34
629
113
19
756
20
90
236
137
1865
106
102
214
525
105
59
521
140
197
1
15
28
56
87
14
252
45
8
302
8
36
94
55
746
43
41
86
210
42
24
208
56
79
0
6
11
140
218
34
629
113
19
756
20
90
236
137
1865
106
102
214
525
105
59
521
140
197
1
15
28
280
436
68
1259
226
38
1512
40
179
471
274
3729
213
204
429
1050
210
118
1042
280
395
2
31
57
559
871
137
2518
451
76
3023
81
358
942
548
7458
426
408
858
2101
419
236
2084
560
790
3
62
114
112
174
27
504
90
15
605
16
72
188
110
1492
85
82
172
420
84
47
417
112
158
1
12
23
280
436
68
1259
226
38
1512
40
179
471
274
3729
213
204
429
1050
210
118
1042
280
395
2
31
57
559
871
137
2518
451
76
3023
81
358
942
548
7458
426
408
858
2101
419
236
2084
560
790
3
62
114
1118
1742
274
5035
902
152
6046
162
717
1885
1096
14917
851
816
1715
4202
838
472
4168
1120
1579
6
123
227
10% 40% 80%
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
Testis
Cervix
Thyroid
Hodgkin's lymphoma
Uterus
Bone
Ovary
Kidney
Melanoma
Multiple myeloma
Oropharynx
Breast
Prostate
Biliary tract
Liver
Stomach
Bladder
Non−Hodgkin's lymphoma
Leukaemia
Oesophagus
Pancreas
Brain
Colorectal
Lung
1
10
100
1000
10000
Absolute excess England deaths
24
72
5
91
48
19
161
8
64
89
180
1030
15
32
87
87
29
296
7
110
4
11
37
60
181
12
227
120
47
402
20
159
223
449
2574
37
81
218
218
72
741
17
276
10
26
93
120
362
24
453
241
94
804
40
318
446
898
5148
74
162
437
435
144
1482
34
552
21
53
185
239
725
47
906
482
188
1607
80
635
893
1796
10296
148
324
874
871
287
2963
69
1104
42
106
371
96
290
19
363
193
75
643
32
254
357
718
4118
59
129
349
348
115
1185
27
442
17
42
148
239
725
47
906
482
188
1607
80
635
893
1796
10296
148
324
874
871
287
2963
69
1104
42
106
371
478
1450
95
1813
964
376
3214
161
1270
1786
3592
20591
296
647
1747
1741
574
5926
137
2208
84
212
741
956
2899
190
3626
1927
752
6429
322
2541
3571
7183
41182
593
1294
3494
3482
1148
11853
274
4415
167
423
1482
191
580
38
725
385
150
1286
64
508
714
1437
8236
119
259
699
696
230
2371
55
883
33
85
296
478
1450
95
1813
964
376
3214
161
1270
1786
3592
20591
296
647
1747
1741
574
5926
137
2208
84
212
741
956
2899
190
3626
1927
752
6429
322
2541
3571
7183
41182
593
1294
3494
3482
1148
11853
274
4415
167
423
1482
1912
5798
379
7251
3854
1504
12858
643
5082
7142
14366
82365
1186
2589
6989
6965
2296
23706
549
8830
334
846
2965
10% 40% 80%
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
Testis
Bone
Prostate
Hodgkins lymphoma
Thyroid
Melanoma
Cervix
Biliary tract
Ovary
Multiple myeloma
Uterus
Breast
Kidney
Bladder
Oesophagus
Non−Hodgkins lymphoma
Leukaemia
Brain
Stomach
Colon
Liver
Pancreas
Lung
10
100
1000
10000
Absolute excess US deaths
age 40-64
Proportion of the population who are affected (PAE) by the COVID-19 emergency
Excess deaths across all 24 cancers
25053
12527
6263
2506
100203
50103
25053
10019
200409
100203
50103
20040
Relative impact of the emergency (RAE)
8
26
3
48
16
11
71
6
22
43
54
315
5
13
37
24
16
101
2
34
6
9
14
19
64
7
121
39
27
177
15
56
107
135
788
14
32
92
61
40
252
5
85
15
23
35
38
128
14
242
79
55
354
31
112
214
269
1576
27
65
184
122
79
503
10
171
30
47
71
76
255
28
484
158
110
708
61
223
427
539
3152
55
130
368
243
158
1006
20
341
59
94
142
30
102
11
194
63
44
283
24
89
171
216
1261
22
52
147
97
63
402
8
136
24
37
57
76
255
28
484
158
110
708
61
223
427
539
3152
55
130
368
243
158
1006
20
341
59
94
142
151
510
56
968
315
220
1416
122
446
855
1078
6304
109
259
736
486
316
2012
40
682
119
187
283
302
1021
111
1935
630
439
2831
244
892
1710
2155
12608
218
518
1472
973
632
4025
81
1365
238
374
566
60
204
22
387
126
88
566
49
178
342
431
2522
44
104
294
195
126
805
16
273
48
75
113
151
510
56
968
315
220
1416
122
446
855
1078
6304
109
259
736
486
316
2012
40
682
119
187
283
302
1021
111
1935
630
439
2831
244
892
1710
2155
12608
218
518
1472
973
632
4025
81
1365
238
374
566
605
2042
222
3870
1261
878
5662
488
1784
3419
4310
25216
437
1037
2944
1946
1264
8050
162
2730
475
749
1133
10% 40% 80%
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
Prostate
Bone
Melanoma
Testis
Hodgkins lymphoma
Biliary tract
Thyroid
Cervix
Multiple myeloma
Uterus
Breast
Ovary
Kidney
Oesophagus
Bladder
Stomach
Non−Hodgkins lymphoma
Leukaemia
Brain
Liver
Colon
Pancreas
Lung
10
100
1000
10000
Absolute excess US deaths
age 65-74
Proportion of the population who are affected (PAE) by the COVID-19 emergency
Excess deaths across all 24 cancers
8837
4421
2209
884
35340
17670
8837
3533
70684
35340
17670
7068
Relative impact of the emergency (RAE)
B C
Fig. 2. Estimated number of excess deaths at 1 year due to the COVID-19 emergency by cancer site for incident cases (A) scaled up to the population of England aged
30+ consisting of 35 million individuals using England (CALIBER) mortality estimates, (B) scaled up to the population of US aged 40-64 consisting of 208 million individuals
using US (SEER) mortality estimates for this age range and (C) scaled up to the population of US aged 65-74 consisting of 61 million individuals using US (SEER) mortality
estimates for this age range.
1-year mortality among incident cancers across cancer sites
in England and the US (Figure S2B). For the US, based on
SEER incidence and 1-year mortality for individuals aged
40-64, at COVID-19 PAE of 40%, we estimated 10,019,
25,053 and 50,103 excess deaths at RIE of 1.2, 1.5 and
2 respectively, when extrapolating to the US population
of this age group(208,284,801 individuals) (Figure 2B).
Using mortality estimates for individuals aged 65-74, at
COVID-19 PAE of 40%, we estimated 3,533, 8,837 and
17,679 excess deaths at RIE of 1.2, 1.5 and 2 respectively,
when extrapolating to the US population of this age group
(61,346,445 individuals) (Figure 2C). For a PAE of 40% and
RIE of 1.5 (range 1.2–2), we estimated 33,890 excess deaths
(range 13,552–67,782) in individuals older than age 40.
Proportion of 15 comorbidity clusters relevant to COVID-
19 risk. Across all cancers, the proportions of 0, 1, 2 and 3+
comorbidities were 51.8%, 25.2%, 14.2% and 8.8% respec-
tively for prevalent cancers. Comorbidities were common
in people with cancer; e.g., hypertension (34,696 [19.0%]),
CVD (23,532 [12.9%]), CKD (9,530 [5.2%]) and obesity
(9,491 [5.2%]) (Figure 3). Prevalence of comorbidities in
some cancers differed from the condition’s prevalence in the
overall population. For example, COPD (3.0%) compared to
CVD (12.9%) is a relatively uncommon comorbidity in pa-
tients with cancer, except for lung cancer where patients had
a higher than background COPD prevalence (25.7%) (Figure
3). Conversely, a ‘metabolic phenotype’ was particularly
common in uterine cancer, where a relatively large propor-
tion of patients had hypertension (48.8%), CVD (26.2%),
obesity (14.1%) and type-2 diabetes (9.3%) (Figure 3).
4 Lai et al. | Excess deaths in cancer and multimorbidity

Stomach Testis Thyroid Uterus
Oesophagus Oropharynx Ovary Pancreas Prostate
Liver Lung Melanoma Multiple myeloma Non−Hodgkin’s lymphoma
Cervix Colorectal Hodgkin’s lymphoma Kidney Leukaemia
Biliary tract Bladder Bone Brain Breast
CVD
Hypertension
Obesity
Type 2 diabetes
CKD
COPD
Immunosuppressive drugs
HIV or corticosteroid
Crohns disease
Chronic liver disease
Spleenic disorder
Cystic fibrosis
Rheumatoid arthritis
Multiple sclerosis
Chronic neurological disorder
CVD
Hypertension
Obesity
Type 2 diabetes
CKD
COPD
Immunosuppressive drugs
HIV or corticosteroid
Crohns disease
Chronic liver disease
Spleenic disorder
Cystic fibrosis
Rheumatoid arthritis
Multiple sclerosis
Chronic neurological disorder
CVD
Hypertension
Obesity
Type 2 diabetes
CKD
COPD
Immunosuppressive drugs
HIV or corticosteroid
Crohns disease
Chronic liver disease
Spleenic disorder
Cystic fibrosis
Rheumatoid arthritis
Multiple sclerosis
Chronic neurological disorder
CVD
Hypertension
Obesity
Type 2 diabetes
CKD
COPD
Immunosuppressive drugs
HIV or corticosteroid
Crohns disease
Chronic liver disease
Spleenic disorder
Cystic fibrosis
Rheumatoid arthritis
Multiple sclerosis
Chronic neurological disorder
CVD
Hypertension
Obesity
Type 2 diabetes
CKD
COPD
Immunosuppressive drugs
HIV or corticosteroid
Crohns disease
Chronic liver disease
Spleenic disorder
Cystic fibrosis
Rheumatoid arthritis
Multiple sclerosis
Chronic neurological disorder
0
10
20
30
40
50
0
10
20
30
40
50
0
10
20
30
40
50
0
10
20
30
40
50
0
10
20
30
40
50
Proportion (%)
N = 114 N = 4,983 N = 505 N = 925 N = 28,652
N = 25,515 N = 10,027 N = 2,327 N = 1,093 N = 3,745
N = 138 N = 2,338 N = 10,164 N = 1,109 N = 4,622
N = 770 N = 1,775 N = 2,130 N = 283 N = 10,244
N = 695 N = 2,236 N = 953 N = 2,306
Age = 71.5 Age = 71.3 Age = 51.9 Age = 55.5 Age = 61.9
Age = 41.6 Age = 70.0 Age = 44.4 Age = 66.3 Age = 65.9
Age = 69.3 Age = 71.4 Age = 50.0 Age = 70.3 Age = 63.6
Age = 71.0 Age = 62.9 Age = 62.2 Age = 72.0 Age = 73.3
Age = 72.9 Age = 37.7 Age = 51.4 Age = 65.6
Fig. 3. Propor tion of patients with any of the 15 comorbidity clusters by cancer site for prevalent cases (N=117,978) in a population of 3,862,012 adults in England (CALIBER).
Age indicates mean age at diagnosis.
Number of comorbid conditions and 1-year mortality in
incident cancers. The proportions of individuals with 0,
1, 2 and 3+ comorbidities for incident cancers were 34.8%,
25.6%, 20.5% and 19.2% respectively (Figure 4A). We found
in incident cases that multimorbidity (≥3 vs 0 conditions)
was associated with a further increase in 1-year mortality.
Cancers of the pancreas (65.7% vs. 80.1%), biliary tract
(58.6% vs. 64.8%), lung (51.9% vs. 60.9%), brain (46.4%
vs. 80.9%) and stomach (42.4% vs. 48.3%) exhibited the
most pronounced effects (Figure 4B, Figure S6). Consistent
findings were found for prevalent cancer cases (Figure S5,
Figure S7)
Excess 1-year COVID-19-related deaths by cancer site
and number of comorbid conditions in England. For inci-
dent cancers, when considering COVID-19 PAE of 40% and
RIE of 1.5, we observed 1,210, 1,509, 1,572 and 1,983 excess
deaths in individuals with 0, 1, 2 and 3+ non-cancer comor-
bidities; 5,064 (80.7%) of these deaths occur in patients with
≥1 comorbidities (Figure 5). When considering COVID-19
PAE of 40% and RIE of 1.5 for prevalent cancers, we ob-
served 2,724, 3,480, 2,955 and 2,558 excess deaths in indi-
viduals with 0, 1, 2 and 3+ non-cancer comorbidities; 8,993
(76.8%) of these deaths occur in patients with ≥1 comor-
bidities (Figure S8). When considering both prevalent and
Lai et al. | Excess deaths in cancer and multimorbidity 5

A B
%
0
1
2
3+
Testis
Cervix
Breast
Melanoma
Prostate
Hodgkin's lymphoma
Thyroid
Uterus
Bladder
Non−Hodgkin's lymphoma
Oropharynx
Multiple myeloma
Leukaemia
Ovary
Colorectal
Kidney
Bone
Oesophagus
Stomach
Brain
Liver
Biliary tract
Lung
Pancreas
0
20
40
60
80
0
1
2
3+
Testis
Cervix
Breast
Melanoma
Prostate
Hodgkin's lymphoma
Thyroid
Uterus
Bladder
Non−Hodgkin's lymphoma
Oropharynx
Multiple myeloma
Leukaemia
Ovary
Colorectal
Kidney
Bone
Oesophagus
Stomach
Brain
Liver
Biliary tract
Lung
Pancreas
Number of COVID-19 relevant comorbidities
20
40
60
80
Number of comorbidities Mortality
%
Number of COVID-19 relevant comorbidities
Fig. 4. Comorbidity clusters relevant for COVID-19 risk in incident cases in England (CALIBER). (A) Proportion of individuals with 0, 1, 2 and 3+ comorbidities by cancer site.
(B) 1-year mortality in individuals with 0, 1, 2 and 3+ comorbidities by cancer site.
incident cancers together at COVID-19 PAE of 40%, we es-
timated 17,991 excess deaths at RIE of 1.5; 78.1% of these
deaths occur in patients with ≥1 comorbidities.
Discussion
This is the first study demonstrating profound recent
changes in cancer care delivery in multiple centers and
the first study modelling overall excess deaths across
incident and prevalent cases of cancer (24 sites) where
comorbidities were present in 78.1% of excess deaths.
Near real-time intelligence. There has been a lack of em-
pirical insights and cancer-specific models to counterbalance
infectious disease modeling in informing evolving policy
responses to the emergency. We propose that weekly na-
tional indicators and warnings for vulnerable patient groups
such as cancer patients with multimorbidity are essential.
For example, weekly cause-specific mortality reporting
of registered deaths and linking of those death records to
primary and secondary care records for cancer and other
services would allow rapid ‘root cause analysis’ of the extent
to which service changes have a net positive or adverse con-
tribution to excess mortality and other non-fatal outcomes.
Changes in cancer services in response to the emergency.
We demonstrated major changes in the delivery of cancer
services in the UK, which are also likely to impact on cancer
patient survival. Two metrics were chosen to reflect both
active treatment for patients diagnosed with cancer and the
urgent referral of patients who may have a new diagnosis
of cancer. The large declines observed in chemotherapy
attendance may reflect workforce/capacity or resources being
redirected to care for infected patients (e.g., to intensive care)
and the desire of clinicians and patients to minimize the risks
of COVID-19 infection(8). We note the importance of con-
sidering multimorbidity in chemotherapy prioritization, as
the risk associated with COVID-19 increases in patients with
multimorbidity. However, some of these patients may benefit
from chemotherapy avoidance to mitigate immunosuppres-
sion. Additionally, we observed large declines in urgent
referrals for early cancer diagnosis. Delays in urgent cancer
referrals may be caused by patients struggling to secure
appointments due to reprioritized health systems, or patients
deciding not to seek care due to perceived risk of COVID-19
infection(17). An unintended consequence of these service
changes may be excess deaths via increases in emergency
admissions, cancer stage shift at presentation and the
undermining of curative/life prolonging surgical resection.
Modeling impact of the emergency: the ‘untold toll’. We
modeled a range of estimates of PAE because the true value
is not known. PAE is a summary measure of exposure to the
adverse health consequences of the emergency, combining
four parameters; the proportion of the population with
1) ill-health in those infected, 2) net adverse health due
to changes in health services designed to protect cancer
patients from infection, 3) net adverse health consequences
of physical distancing, 4) adverse health consequences of
economic downturn. Since empirical estimates of each of
these four parameters for PAE are not yet available, we chose
a PAE range of 10%-80%, with 40% as a plausible estimate.
RIE is a summary measure of the combined impact on
mortality of infection, health service change, physical
distancing and economic downturn. We modeled a range
of estimates of RIE. The only direct evidence we have to
date of the magnitude of the RIE comes from national death
reporting in England(1). Here, an RIE of 1.5 is consistent
with ONS estimates of excess deaths during the COVID-19
emergency in England, which reported almost 8,000 more
deaths in the week of 10th April 2020, than the 5-year
average of 10,520(1). A higher RIE is possible, given
the following evidence: 1) patients with cancer may have
6 Lai et al. | Excess deaths in cancer and multimorbidity

Proportion of the population who are affected (PAE) by the COVID-19 emergency
10% 40% 80%
01
23+
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
Testis
Cervix
Thyroid
Hodgkin's lymphoma
Uterus
Bone
Ovary
Kidney
Melanoma
Oropharynx
Multiple myeloma
Breast
Prostate
Biliary tract
Liver
Stomach
Bladder
Non−Hodgkin's lymphoma
Leukaemia
Oesophagus
Pancreas
Brain
Colorectal
Lung
Testis
Cervix
Thyroid
Hodgkin's lymphoma
Uterus
Bone
Ovary
Kidney
Melanoma
Oropharynx
Multiple myeloma
Breast
Prostate
Biliary tract
Liver
Stomach
Bladder
Non−Hodgkin's lymphoma
Leukaemia
Oesophagus
Pancreas
Brain
Colorectal
Lung
1
10
100
1000
Absolute excess England deaths
10% 40% 80%
10% 40% 80% 10% 40% 80%
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
1.2
1.5
2
3
Number of individuals with 0 comorbidity
45569 (34.8%)
Number of individuals with 2 comorbidities
26876 (20.5%)
Number of individuals with 1 comorbidity
33543 (25.6%)
Number of individuals with 3+ comorbidities
25108 (19.2%)
Excess deaths across all 24 cancers
1210
608
303
120
4838
2419
1210
482
9677
4838
2419
965
Excess deaths across all 24 cancers
1509
754
375
150
6022
3013
1509
602
12049
6022
3013
1203
Excess deaths across all 24 cancers
1572
787
393
157
6294
3146
1572
629
12586
6294
3146
1258
Number of comorbidities
Relative impact of the emergency (RAE)
Number of comorbidities
Excess deaths across all 24 cancers
1983
993
495
201
7928
3963
1983
792
15859
7928
3963
1585
Fig. 5. Estimated number of excess deaths at 1 year due to the COVID-19 emergency by cancer site and number of non-cancer comorbidities for incident cases, scaled up
to the population of England aged 30+ consisting of 35 million individuals.
twice the odds of developing COVID-19 infection (odds
ratio=2.31) compared to the rest of the population(28),
2) COVID-19 causes patients with cancer to deteriorate
more rapidly (hazard ratio=3.5) compared to those without
cancer(11) and 3) COVID-19 case fatality rates are higher
in patients with comorbid conditions including CVD, di-
abetes, hypertension and cancer(29). For cancer patients
of working age, unemployment may affect mortality, as
we have previously demonstrated across 75 countries(30).
Social isolation is also known to represent a mortality risk
in cancer(31,32). Our findings relate to predicted excess
mortality in the next 12 months. However, the socio-
economic effects on health from the current epidemic are
likely to last for a considerable period beyond one year(33),
meaning our figures are likely a significant underestimate.
Excess deaths. Based on the available evidence, we estimate
an excess of 6,270 excess deaths at 1 year in patients with
incident cancers in England. The recorded underlying
cause of these excess deaths may be cancer, COVID-19,
or comorbidity (such as myocardial infarction), and it is
likely that in the COVID-19 emergency, there may be
changes in cause-of-death recording. This is why our model
uses death from any cause – for which there can be no
changes in recording - and which is relevant to multimorbid
cancer patients. Our conservative estimate nonetheless
represents a significant additional burden of 20%, the total
number of cancer deaths annually among incident cases
in England is 31,354(34). We present estimates of excess
deaths in incident cases, most of whom will be under active
surgical, adjuvant or palliative treatments. We also demon-
strate excess deaths in those with prevalent cancers, many
of whom have survived the initial high-risk period; these
data are particularly relevant to ongoing patient management.
Model application in the US. Using SEER estimates
of incidence and 1-year mortality, we modeled excess
deaths in the US. For example, with an RIE 1.5 and
PAE 40% we estimate 33,890 excess 1-year deaths.
Like other cancer registries, SEER does not obtain
information on the range of (COVID-19 relevant) co-
morbidities. We found good agreement between Eng-
Lai et al. | Excess deaths in cancer and multimorbidity 7

land and US estimates for incidence and mortality(35).
Importance of multimorbidity. We show that most pa-
tients with cancer have non-cancer comorbid conditions
which confer additional mortality risk; many of these
comorbidities are treatable by non-cancer services. The
COVID-19 emergency has prompted new questions about
which cancer patients are most vulnerable and how best
to mitigate risk. Effective management of hypertension,
for example, may avert death and major adverse events,
but it is possible that blood pressure control may worsen
during the emergency and contribute to excess deaths.
Policy implications. Our study can inform policy in four
areas. First, there is a need to mobilize access to real-time
national data on mortality (to determine which disease
combinations pose the greatest risk) and on cancer health
services activity (to monitor the effects of system change
during the emergency on care delivery and future health
outcomes). This health services intelligence should include
both data from cancer services and from those services
(e.g. internists and family physicians) who manage the
treatable comorbidities of cancer patients. Second, the
triaging of chemotherapy prioritization would usefully
incorporate patient-specific risk/benefit assessments to
include multimorbidity, particularly in situations where
chemotherapy outweighs the benefits of non-treatment and
safety issues(17) related to COVID-19. Third, there is a
need to enable access to life-saving early diagnosis and
to apply a triage system to prioritize urgent referrals for
citizens with worrying symptoms, especially those with
comorbidities, as multimorbidity may mask an underlying
cancer. Fourth, the policy of ‘shielding’ vulnerable patients
under active treatment for cancer, or with one of a range of
other non-malignant conditions has been implemented in
1.5 million individuals in England, involving home delivery
of food parcels and medicines for an initial period of 12
weeks. We propose that estimates of background (pre
COVID-19) 1-year mortality risks provides a transparent
rationale, readily executable in health records, for iden-
tifying priority patient groups including cancer patients
with multimorbidities to receive shielding interventions.
Limitations. This study has important limitations. First,
there is a lack of near real-time clinical data in multiple
digitally-mature hospitals with large numbers of COVID-19
infected patients. Second, the health records we used
may have missed cases of cancer and underestimated
incidence(36); if so, our estimates of excess deaths may
be conservative. The National Health Service has na-
tional linked hospital admissions and cancer registration
data with information on stage and details of surgical,
chemotherapeutic and radiotherapy treatment of cancer.
However, information governance for such data can take
months to secure, making data-enabled research and time-
sensitive responsive service improvement difficult. Third,
we did not have access to multimorbidity data for the
US. Fourth, we did not have access to data on children.
Conclusion
Our data have highlighted how cancer patients with multi-
morbidity are a particularly at-risk group during the current
pandemic. In order to ensure effective cancer policy and
avoid excess deaths, both during and after the COVID-19
emergency, it is critical to ensure near-real time reporting of
cause-specific excess mortality, urgent cancer referrals and
treatment statistics, so as to inform the most optimal delivery
of care in this extremely vulnerable group of patients.
Authors’ contribution
Research question: AL, HH. Funding: AL, AB, HH,
DATA-CAN. Study design and analysis plan: AL, LP, AB,
MK, WHC, HH. Preparation of data, including electronic
health record phenotyping in the CALIBER open portal:
AL, LP, SD. Provision of weekly hospital data: GH, KPJ,
MDF, DH, ML, KB, CD. Statistical analysis: AL, LP, MK,
WHC. Drafting initial versions of manuscript: AL, HH.
Drafting final versions of manuscript: AL, ML, HH. Crit-
ical review of early and final versions of manuscript: All
authors.
Declaration of interests
ML has received honoraria from Pfizer, EMD Serono and
Roche for presentations unrelated to this research. ML has
received an unrestricted educational grant from Pfizer for re-
search unrelated to the research presented in this paper. MDF
has received research funding from AstraZeneca, Boehringer
Ingelheim, Merck and MSD and honoraria from Achilles,
AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Meyers
Squibb, Celgene, Guardant Health, Merck, MSD, Nanobi-
otix, Novartis, Pharmamar, Roche and Takeda for advisory
roles or presentations unrelated to this research. GRF re-
ceives funding from companies that manufacture drugs for
hepatitis C virus (AbbVie, Gilead, MSD) and consult for
GSK and Arbutus in areas unrelated to this research.
Acknowledgements
We thank Tony Hagger, Shiva Thapa, Mohammed Emran,
Cara Anderson, Louise Herron, Philip Melling and Lee Cog-
ger for their help on collating data on urgent cancer referrals
and chemotherapy attendances. We thank Charles Swanton
for his valuable comments on the manuscript. This work uses
data provided by patients and collected by the NHS as part of
their care and support.
Funding statement
We acknowledge Health Data Research UK (HDR UK) sup-
port for the HDR UK substantive sites involved in this re-
search (HDR London, HDR Wales and Northern Ireland) and
DATA-CAN. DATA-CAN is part of the Digital Innovation
8 Lai et al. | Excess deaths in cancer and multimorbidity

Hub Programme, delivered by HDR UK and funded by UK
Research and Innovation through the government’s Industrial
Strategy Challenge Fund (ISCF). AGL is supported by fund-
ing from the Wellcome Trust, National Institute for Health
Research (NIHR) University College London Hospitals and
NIHR Great Ormond Street Hospital Biomedical Research
Centers. AB is supported by research funding from NIHR,
British Medical Association, Astra-Zeneca and UK Research
and Innovation. KPJ is supported by the NIHR Great Or-
mond Street Hospital Biomedical Research Centre. CD is
funded by UCLPartners. HH is an NIHR Senior Investiga-
tor and is funded by the NIHR University College London
Hospitals Biomedical Research Centre, supported by Health
Data Research UK (grant No. LOND1), which is funded by
the UK Medical Research Council, Engineering and Physical
Sciences Research Council, Economic and Social Research
Council, Department of Health and Social Care (England),
Chief Scientist Office of the Scottish Government Health
and Social Care Directorates, Health and Social Care Re-
search and Development Division (Welsh Government), Pub-
lic Health Agency (Northern Ireland), British Heart Foun-
dation, Wellcome Trust, The BigData@Heart Consortium,
funded by the Innovative Medicines Initiative-2 Joint Under-
taking under grant agreement No. 116074.
Supplementary methods
This study was performed as part of the CALIBER
program (https://caliberresearch.org/portal), which is a
research resource consisting of anonymized, coded vari-
ables extracted from linked electronic health records,
methods, tools and specialized infrastructure. Ethical
approval was granted (20_074R2) by the MHRA (UK)
Independent Scientific Advisory Committee of the Clinical
Practice Research Data Link under Section 251 (NHS
Social Care Act 2006). The interpretation and conclu-
sions contained in this study are those of the authors’ alone.
15 comorbid conditions or condition clusters involv-
ing 40 individual conditions defined by the Centers
for Disease Control (CDC) or Public Health England
(PHE) as associated risk factors for poor outcomes caused
by severe COVID-19 infection. Both lists from CDC
and PHE included chronic respiratory disease; chronic
heart disease; immunocompromised; HIV or use of cor-
ticosteroids; obesity; diabetes; chronic kidney disease;
chronic liver disease. The PHE list included additional
condition clusters (chronic neurological disorders; splenic
disorders). We have performed analyses for all the above
conditions and have additionally considered hypertension,
Crohn’s disease, cystic fibrosis and rheumatoid arthritis.
Given that condition clusters such as (i) chronic heart
disease would involve a range of conditions, we have derived
composite variables to include 15 conditions considered as
cardiovascular disease (CVD) that included acute myocar-
dial infarction, unstable angina, chronic stable angina, heart
failure, cardiac arrest or sudden coronary death, transient
ischemic attack, intracerebral hemorrhage, subarachnoid
hemorrhage, ischemic stroke, abdominal aortic aneurysm,
peripheral arterial disease, atrial fibrillation, congenital
heart disease, hypertrophic and dilated cardiomyopathy
and valve disease (multiple, mitral and aortic)(23). We
also considered (ii) Hypertension, defined as ≥140 mmHg
systolic blood pressure (or ≥150 mmHg for people aged
≥60 years without diabetes and chronic kidney disease)
and/or ≥90 mmHg diastolic blood pressure(22), (iii) type
2 diabetes, (iv) obesity, defined as a body mass index of
≥30kg/m2, (v) chronic kidney disease (CKD), (vi) chronic
obstructive pulmonary disease (COPD)(25), (vii) patients
on immunosuppressive drugs (not cancer chemotherapy),
(viii) patients with HIV or corticosteroid prescription, (ix)
chronic neurological disorders, defined as a composite of
Parkinson’s disease, motor neuron disease, learning disabil-
ity and cerebral palsy, (x) multiple sclerosis separately, (xi)
splenic disorders, defined as a composite of splenomegaly,
splenectomy and hyposplenism, (xii) chronic liver diseases,
defined as a composite of chronic viral hepatitis B or C,
primary biliary cholangitis, liver fibrosis, liver cirrhosis
and non-alcoholic fatty liver disease, (xiii) Crohn’s dis-
ease, (xiv) cystic fibrosis and (xv) rheumatoid arthritis(24).
Analyses. In CALIBER (England), we estimated the fre-
quency (%) of comorbid conditions in prevalent cancers (di-
agnosed at or any time prior to baseline) and incident can-
cers as occurring any time over follow-up. We estimated
incidence rates for as the number of new cancers by can-
cer site per 100,000 person years. We compared CALIBER
incidence rates for England with those from the UK from
the International Agency for Research on Cancer (IARC).
In CALIBER, we generated Kaplan-Meier estimates on 1-
year mortality for incident cancers observed from 2012-2016
(by cancer cites and by number of comorbid conditions) and
prevalent cancers (by cancer sites and by number of comor-
bid conditions). We only considered non-fatal incident cases,
i.e., alive for at least 30 days following cancer diagnosis to ac-
count for potential time lag in death recording. Excess deaths
for incident cancers were estimated based on 1-year mortal-
ity and 1-year incidence rates, scaled up to a population of
35,407,313 individuals aged 30 and above in England based
on population estimates from 2018(37). We used publicly
available data from the US from the Surveillance, Epidemi-
ology, and End Results (SEER) program to estimate 1 year
mortality as 1 minus the reported survival(27). We consid-
ered mortality data from SEER for individuals aged 40-64,
65-74 and 75+. We applied our model (RIE and PAE) to
SEER estimates of incidence and 1-year mortality (accord-
ing to the corresponding age groups) scaled up to the US
population of individuals aged 40-64 (208,284,801 individu-
als) and 65-74 (61,346,445 individuals) based on population
estimates from 2018(38).
Bibliography
1. Office for National Statistics. Deaths registered weekly in England and Wales, provisional:
week ending 10 April 2020. Accessed 22 April 2020. .
Lai et al. | Excess deaths in cancer and multimorbidity 9

2. Lisa Rosenbaum. The Untold Toll — The Pandemic’s Effects on Patients without Covid-19.
New England Journal of Medicine. doi: 10.1056/NEJMms2009984.
3. Eduard Vrdoljak, Richard Sullivan, and Mar k Lawler. Cancer and COVID-19; how do we
manage cancer optimally through a public health crisis? European Journal of Cancer, 132:
P98–99, 2020.
4. American Cancer Society. Survey: COVID-19 Affecting Patients’ Access to Cancer Care.
Accessed 19 April 2020.
5. American College of Surgeons. COVID-19 Guidelines for Triage of Breast Cancer Patients.
Accessed 19 April 2020.
6. American Society for Radiation Oncology. COVID-19 Coding Guidance. Accessed 19 April
2020.
7. British Society of Gastroenterology. Endoscopy activity and COVID-19: BSG and JAG guid-
ance. Accessed 19 April 2020.
8. National Institute for Health and Care Excellence. COVID-19 rapid guideline: delivery of
systemic anticancer treatments. Accessed 19 April 2020.
9. Talha Khan Burki. Cancer guidelines during the COVID-19 pandemic. The Lancet Oncology,
2020.
10. David J Thomson, David Palma, Matthias Guckenberger, Panagiotis Balermpas, Jonathan J
Beitler, Pierre Blanchard, David Brizel, Wilfred Budach, Jimmy Caudell, June Corry, and
Others. Practice recommendations for risk-adapted head and neck cancer radiotherapy
during the COVID-19 pandemic: an ASTRO-ESTRO consensus statement. International
Journal of Radiation Oncology* Biology* Physics, 2020.
11. Wenhua Liang, Weijie Guan, Ruchong Chen, Wei Wang, Jianfu Li, Ke Xu, Caichen Li, Qing
Ai, Weixiang Lu, Hengrui Liang, and Others. Cancer patients in SARS-CoV-2 infection: a
nationwide analysis in China. The Lancet Oncology, 21(3):335–337, 2020.
12. Hanping Wang and Li Zhang. Risk of COVID-19 for patients with cancer. The Lancet
Oncology, 21(4):e181, 2020.
13. Centers for Disease Control and Prevention. Groups at Higher Risk for Severe Illness. Ac-
cessed 19 April 2020.
14. Public Health England. Guidance on shielding and protecting people defined on medical
grounds as extremely vulnerable from COVID-19. Accessed 19 April 2020. .
15. Public Health England. Guidance on social distancing for everyone in the UK. Accessed 19
April 2020. .
16. Cristina Renzi, Aradhna Kaushal, Jon Emery, Willie Hamilton, Richard D. Neal, Bernard
Rachet, Greg Rubin, Hardeep Singh, Fiona M. Walter, Niek J. de Wit, and Georgios Lyrat-
zopoulos. Comorbid chronic diseases and cancer diagnosis: disease-specific effects and
underlying mechanisms. Nature Reviews Clinical Oncology, 2019. ISSN 1759-4774. doi:
10.1038/s41571-019- 0249-6.
17. Timothy P Hanna, Gerald A Evans, and Christopher M Booth. Cancer, COVID-19 and the
precautionary principle: pr ioritizing treatment during a global pandemic. Nature Reviews
Clinical Oncology, pages 1–3, 2020.
18. Amitava Banerjee, Laura Pasea, Steve Harris, Arturo Gonzalez-Izquierdo, Ana Torralbo,
Laura Shallcross, Mahdad Noursadeghi, Deenan Pillay, Christina Pagel, Wai Keong Wong,
and Others. Estimating excess 1-year mortality from COVID-19 according to underlying
conditions and age in England: a rapid analysis using NHS health records in 3.8 million
adults. Lancet (in press), 2020.
19. Health Data Research UK. DATA-CAN - The Health Data Research Hub for Cancer. Ac-
cessed 20 April 2020.
20. Spiros C Denaxas, Julie George, Emily Herrett, Anoop D Shah, Dipak Kalra, Aroon D Hingo-
rani, Mika Kivimaki, Adam D Timmis, Liam Smeeth, and Harry Hemingway. Data resource
profile: cardiovascular disease research using linked bespoke studies and electronic health
records (CALIBER). Inter national journal of epidemiology, 41(6):1625–1638, 2012.
21. Spiros Denaxas, Ar turo Gonzalez-Izquierdo, Kenan Direk, Natalie K Fitzpatrick, Ghaza-
leh Fatemifar, Amitava Banerjee, Richard J B Dobson, Laurence J Howe, Valerie Kuan,
R Tom Lumbers, and Others. UK phenomics platform for developing and validating elec-
tronic health record phenotypes: CALIBER. Journal of the American Medical Informatics
Association, 2019.
22. Eleni Rapsomaniki, Adam Timmis, Julie George, Mar Pujades-Rodriguez, Anoop D. Shah,
Spiros Denaxas, Ian R. White, Mark J. Caulfield, John E. Deanfield, Liam Smeeth, Bryan
Williams, Aroon Hingorani, and Harry Hemingway. Blood pressure and incidence of twelve
cardiovascular diseases: Lifetime risks, healthy life-years lost, and age-specific associations
in 1·25 million people. The Lancet, 383(9932):1899–1911, 2014. ISSN 1474547X. doi:
10.1016/S0140-6736(14)60685- 1.
23. Anoop Dinesh Shah, Claudia Langenberg, Eleni Rapsomaniki, Spiros Denaxas, Mar
Pujades-Rodriguez, Chris P. Gale, John Deanfield, Liam Smeeth, Adam Timmis, and Harry
Hemingway. Type 2 diabetes and incidence of cardiovascular diseases: A cohort study in
1·9 million people. The Lancet Diabetes and Endocrinology, 3(2):105–113, 2015. ISSN
22138595. doi: 10.1016/S2213-8587(14)70219- 0.
24. Mar Pujades-Rodriguez, Bram Duyx, Sara L Thomas, Dimitris Stogiannis, Anisur Rahman,
Liam Smeeth, and Harry Hemingway. Rheumatoid arthritis and incidence of twelve initial
presentations of cardiovascular disease: a population record-linkage cohort study in Eng-
land. PloS one, 11(3), 2016.
25. Maria Pikoula, Jennifer Kathleen Quint, Francis Nissen, Harry Hemingway, Liam Smeeth,
and Spiros Denaxas. Identifying clinically important COPD sub-types using data-driven
approaches in primary care population based electronic health records. BMC medical infor-
matics and decision making, 19(1):86, 2019.
26. Valerie Kuan, Spiros Denaxas, Arturo Gonzalez-Izquierdo, Kenan Direk, Osman Bhatti,
Shanaz Husain, Shailen Sutaria, Melanie Hingorani, Dorothea Nitsch, Constantinos A
Parisinos, R Thomas Lumbers, Rohini Mathur, Reecha Sofat, JuanP Casas, Ian C K Wong,
Harry Hemingway, and Aroon D Hingorani. A chronological map of 308 physical and men-
tal health conditions from 4 million individuals in the English National Health Service. The
Lancet Digital Health, 1(2):e63–e77, 2019. ISSN 25897500. doi: 10.1016/S2589- 7500(19)
30012-3.
27. MA Hernan and JM Robins. Causal Inference: What If. Boca Raton: Chapman & Hall/CRC.,
2020.
28. Jing Yu, Wen Ouyang, Melvin L K Chua, and Conghua Xie. SARS-CoV-2 transmission in
patients with cancer at a tertiary care hospital in Wuhan, China. JAMA oncology, 2020.
29. Zunyou Wu and Jennifer M McGoogan. Characteristics of and important lessons from the
coronavirus disease 2019 (COVID-19) outbreak in China: summary of a repor t of 72 314
cases from the Chinese Center for Disease Control and Prevention. Jama, 2020.
30. Mahiben Maruthappu, Johnathan Watkins, Aisyah Mohd Noor, Callum Williams, Raghib Ali,
Richard Sullivan, Thomas Zeltner, and Rifat Atun. Economic downturns, universal health
coverage, and cancer mortality in high-income and middle-income countries, 1990–2010: a
longitudinal analysis. The Lancet, 388(10045):684–695, 2016.
31. Candyce H Kroenke, Laura D Kubzansky, Eva S Schernhammer, Michelle D Holmes, and
Ichiro Kawachi. Social networks, social support, and survival after breast cancer diagnosis.
J Clin Oncol, 24(7):1105–1111, 2006.
32. Marko Elovainio, Christian Hakulinen, Laura Pulkki-Råback, Marianna Virtanen, Kim Josefs-
son, Markus Jokela, Jussi Vahtera, and Mika Kivimäki. Contribution of risk factors to excess
mortality in isolated and lonely individuals: an analysis of data from the UK Biobank cohor t
study. The Lancet Public Health, 2(6):e260—-e266, 2017.
33. Institute for Fiscal Studies. Accessed 24 April 2020.
34. Cancer Research UK. Cancer Statistics for the UK. Accessed 20 April 2020.
35. Sur veillance, Epidemiology, and End Results (SEER) Program. Overview of the SEER Pro-
gram. Accessed 19 April 2020.
36. Andrea V Margulis, Joan Fortuny, James A Kaye, Brian Calingaert, Maria Reynolds, Estel
Plana, Lisa J McQuay, Willem Jan Atsma, Billy Franks, Stefan de Vogel, and Others. Vali-
dation of cancer cases using primary care, cancer registry, and hospitalization data in the
United Kingdom. Epidemiology (Cambridge, Mass.), 29(2):308, 2018.
37. Office for National Statistics. Estimates of the population for the UK, England and Wales,
Scotland and Northern Ireland. Accessed 23 April 2020. .
38. United States Census Bureau. National Population by Characteristics: 2010-2019. Ac-
cessed 23 April 2020.
10 Lai et al. | Excess deaths in cancer and multimorbidity