Keywords
intracerebral bleeding - vitamin K antagonists - direct oral anticoagulants - hospitalization
- mortality
Introduction
Intracranial bleedings (ICBs) occur as a consequence of craniocerebral trauma or without
external influence in 20% of strokes, in coagulation disorders, or disorders of blood
vessels (aneurysms, arteriovenous malformation).[1]
Direct oral anticoagulants (DOACs) entered the market in 2008 and have become widely
used since 2010. Their advantages over vitamin K antagonists (VKA) in terms of practicability
and bleeding rates increased their acceptance.[2] A meta-analysis on the risk of ICB among DOACs or VKAs in patients aged ≥75 years
described a significantly reduced risk for DOACs with a hazard ratio (HR) of 0.58
(95% confidence interval [95% CI]: 0.50–0.67).[3] A meta-analysis of 19 trials comparing hematoma volume, hematoma expansion, and
mortality in VKA-ICB versus non-anticoagulation-associated ICB established a more
unfavorable course of all these parameters in VKA-ICB patients.[4] A recent study including 193 ICB patients showed significantly smaller hematoma
volumes under DOACs in comparison to VKAs and a significantly lower 30-day intrahospital
mortality rate of DOAC-ICB, even before the introduction of specific antidotes.[5]
DOACs might have affected the ICB rate in Germany by different aspects. The prescription
rates of DOACs have increased dramatically over the last decade in Germany,[6] which could have led to more ICB cases. The lower bleeding risk of DOACs might have
compensated for such an increase. If DOACs cause smaller intracranial bleeding generally,
overall mortality in ICB cases should have decreased.
To look for these effects in the German population, we compared the nationwide time
trends of general hospitalization and hospitalization for ICB with and without long-term
use of anticoagulants (LUAs) in Germany and finally intrahospital mortality of ICBs
dependent on LUAs.
Patients and Methods
This is a retrospective cohort study including the nationwide hospitalization data
documented in the Diagnosis-Related Groups Statistic (DRG). The nationwide hospitalization
data documented in the DRG represent an annual survey of all hospital cases in Germany
that were accounted for by case rates. All the hospitals in Germany annually transfer
their individual hospitalization data, including one primary diagnosis, up to 89 secondary
diagnoses coded by ICD-10 (International Classification of Diseases, 10th edition),
and up to 100 medical procedures according to a national classification of operations
and procedures to the Institute for the Hospital Remuneration System (InEK). After
a plausibility control, the InEK forwards anonymized data to the Federal Bureau of
Statistics. The principles of the analysis of this hospitalization file have been
published several times previously.[7]
[8]
In brief, we asked the Federal Bureau of Statistics to identify all hospitalizations
in the period from 2006 to 2020 that have a principal diagnosis of ICB (I60: subarachnoid
bleeding and I61: intracerebral bleeding) by calendar year, sex, and 5-year age groups.
Among these cases, we looked for those with LUAs, defined as the additional diagnosis
of “personal history of long-term (current) use of anticoagulants” (ICD code Z92.1).
Finally, we received a defined data set from the Federal Bureau of Statistics including
this information from all fully reimbursed inpatient cases. For this analysis, we
included all the patients aged ≥60 years.
The primary outcome was intrahospital mortality documented in the DRG. According to
the occupational regulations for the North Rhine-Westphalian physicians, retrospective
epidemiological research projects are specifically excluded from the requirement of
an ethical approval. Specific linking of cases and procedures is possible but not
allowed for legal reasons. Thus, institutional review board approval and patients'
informed consent were not necessary for this study.
For all hospitalized patients without I60/I61, we calculated the age-standardized
proportions of patients with anticoagulation for 2006 to 2020 separately for men and
women—hospitalized patients from 2006 were used as reference population for the age
standardization. The same was done for patients with I60/I61—here all patients with
I60/I61 from 2006 were used as the reference population.
For patients with I60/I61, sex-specific age-standardized mortality was calculated
for 2006 to 2020 separately for patients with and without anticoagulation—all the
patients with I60/I61 from 2006 were used as the reference population. Calculations
were done using SAS version 9.4 (SAS Institute, Cary, United States).
Results
Long-Term Use of Anticoagulants
In 2006, a total of 8,026,468 patients were hospitalized. The age-standardized rate
of male patients with LUAs was 7.3% (95%CI: 7.3–7.3) and that of female patients was
5.6% (95% CI: 5.6–5.6; [Table 1], [Fig. 1]). In 2020, hospitalized cases increased to 8,921,978 and the rate of males with
LUAs increased to 22.0% (95% CI: 21.9–22.0) and that of females to 17.7% (95% CI:
17.6–17.7). Among all cases hospitalized for ICB in 2006 (n = 30,792), the rate of males with LUAs was 7.0% (95% CI: 6.5–7.4) and that of females
with LUAs was 5.6% (95% CI: 5.3–6.0). In 2020, cases with ICB sum up to 31,650. The
rates of males and females with LUA increased to 13.7% (95% CI: 13.1–14.3) and 10.8%
(95% CI: 10.3–11.3), respectively.
Table 1
Gender-specific age-standardized proportions of patients with anticoagulation among
hospitalized patients with and without I60/I61, Germany, 2006–2020
|
All hospitalized cases without I60/I61
|
All hospitalized cases with I60/I61
|
Year
|
Cases with anticoagulation[a]
|
All cases
|
Age-standardized proportions of cases with anticoagulation[b] (95% CI)
|
Cases with anticoagulation[a]
|
All cases
|
Age-standardized proportions of cases with anticoagulation[b] (95% CI)
|
Males
|
2006
|
266,061
|
3,696,195
|
7.3 (7.3–7.3)
|
1,041
|
14,524
|
7.0 (6.5–7.4)
|
2007
|
307,783
|
3,829,611
|
8.2 (8.1–8.2)
|
1,255
|
14,767
|
8.2 (7.8–8.7)
|
2008
|
348,199
|
3,948,099
|
9.0 (8.9–9.0)
|
1,571
|
14,944
|
10.2 (9.7–10.7)
|
2009
|
395,130
|
4,069,876
|
9.9 (9.9–9.9)
|
1,693
|
14,721
|
11.1 (10.6–11.6)
|
2010
|
429,206
|
4,172,491
|
10.5 (10.4–10.5)
|
1,583
|
15,512
|
10.9 (10.4–11.4)
|
2011
|
468,658
|
4,285,188
|
11.1 (11.1–11.2)
|
1,790
|
15,071
|
11.4 (10.9–12.0)
|
2012
|
518,113
|
4,405,730
|
12.0 (11.9–12.0)
|
1,932
|
15,341
|
12.2 (11.6–12.7)
|
2013
|
574,544
|
4,485,581
|
13.0 (13.0–13.1)
|
2,124
|
15,430
|
13.2 (12.7–13.8)
|
2014
|
653,868
|
4,598,306
|
14.5 (14.5–14.5)
|
2,417
|
16,246
|
14.5 (13.9–15.1)
|
2015
|
717,707
|
4,702,323
|
15.6 (15.6–15.6)
|
2,172
|
16,582
|
12.6 (12.1–13.2)
|
2016
|
790,771
|
4,777,214
|
17.0 (16.9–17.0)
|
2,079
|
16,391
|
12.1 (11.5–12.6)
|
2017
|
854,078
|
4,858,479
|
18.1 (18.0–18.1)
|
2,113
|
16,517
|
12.1 (11.6–12.6)
|
2018
|
924,302
|
4,887,536
|
19.5 (19.5–19.5)
|
2,067
|
15,918
|
12.3 (11.8–12.9)
|
2019
|
1,007,251
|
4,995,100
|
20.8 (20.7–20.8)
|
2,266
|
16,231
|
13.4 (12.8–13.9)
|
2020
|
941,602
|
4,433,437
|
22.0 (21.9–22.0)
|
2,198
|
15,362
|
13.7 (13.1–14.3)
|
Females
|
2006
|
242,058
|
4,330,273
|
5.6 (5.6–5.6)
|
942
|
16,268
|
5.6 (5.3–6.0)
|
2007
|
279,453
|
4,458,099
|
6.2 (6.2–6.2)
|
1,088
|
16,689
|
6.3 (5.9–6.7)
|
2008
|
317,518
|
4,579,632
|
6.8 (6.8–6.8)
|
1,354
|
16,696
|
7.5 (7.1–7.9)
|
2009
|
357,424
|
4,668,066
|
7.5 (7.4–7.5)
|
1,406
|
16,531
|
8.0 (7.5–8.4)
|
2010
|
387,510
|
4,736,357
|
7.9 (7.9–8.0)
|
1,465
|
16,816
|
7.9 (7.5–8.4)
|
2011
|
429,878
|
4,822,912
|
8.6 (8.6–8.6)
|
1,660
|
16,559
|
9.1 (8.7–9.6)
|
2012
|
474,608
|
4,907,488
|
9.3 (9.3–9.3)
|
1,692
|
16,743
|
9.2 (8.7–9.7)
|
2013
|
526,353
|
4,954,533
|
10.1 (10.1–10.1)
|
1,843
|
16,704
|
9.9 (9.4–10.4)
|
2014
|
602,340
|
5,048,069
|
11.3 (11.3–11.3)
|
2,049
|
17,114
|
10.8 (10.2–11.3)
|
2015
|
658,984
|
5,085,417
|
12.2 (12.2–12.3)
|
1,950
|
17,952
|
9.4 (9.0–9.9)
|
2016
|
728,668
|
5,126,454
|
13.4 (13.4–13.5)
|
1,939
|
17,466
|
9.6 (9.1–10.1)
|
2017
|
788,314
|
5,161,908
|
14.4 (14.3–14.4)
|
1,883
|
17,259
|
9.4 (8.9–9.8)
|
2018
|
849,844
|
5,129,426
|
15.6 (15.5–15.6)
|
1,995
|
16,794
|
10.1 (9.6–10.6)
|
2019
|
917,773
|
5,186,912
|
16.6 (16.5–16.6)
|
1,999
|
16,941
|
10.1 (9.6–10.5)
|
2020
|
846,482
|
4,488,541
|
17.7 (17.6–17.7)
|
2,055
|
16,288
|
10.8 (10.3–11.3)
|
Abbreviation: CI, confidence interval.
a Anticoagulation = ICD-10 code Z92.1 (long-term (current) use of anticoagulation).
b Reference population: hospitalized patients without I60/I61 in 2006.
Fig. 1 Visualization of gender-specific age-standardized proportions of patients with anticoagulation
among hospitalized patients with and without I60/I61, Germany, 2006–2020 I60: subarachnoid
bleeding; I61: intracerebral bleeding; ICD-10 Z92.1 (long-term, current, use of anticoagulation);
CI: confidence interval; weights for age-standardization from hospitalized patients
without I60/I61 in 2006.
Mortality of ICB Cases
In 2006, the age-standardized mortality rates in all ICB cases without LUAs in males
and females were 24.1% (95% CI: 23.2–25.0) and 23.9% (95% CI: 23.1–24.6), respectively.
In 2020, these mortality rates slightly decreased to 22.7% (95% CI: 21.9–23.5) in
males and remained almost unchanged at 23.8% (95% CI: 22.9–24.6) in females ([Table 2], [Fig. 2]). In patients with LUAs, the mortality rates decreased form 30.1 (95% CI: 25.7–34.5)
to 24.3% (95% CI: 22.1–26.5) in males and from 28.4 (95% CI: 24.7–32.1) to 24.2% (95%
CI: 21.6–26.9) in females in the same period.
Table 2
Gender-specific age-standardized case fatality (%) in persons with I60/I61 stratified
by long-term use of anticoagulation, Germany, 2006–2020
|
Without long-term (current) use of anticoagulation[a]
|
With long-term (current) use of anticoagulation[a]
|
Year
|
No. of dead persons with I60/I61
|
No. of persons with I60/I61
|
Age-standardized case fatality (%) in persons with I60/I61 (95% CI)[b]
|
No. of dead persons with I60/I61
|
No. of persons with I60/I61
|
Age-standardized case fatality (%) in persons with I60/I61 (95% CI)[b]
|
Males
|
2006
|
3,088
|
13,483
|
24.1 (23.2–25.0)
|
301
|
1,041
|
30.1 (25.7–34.5)
|
2007
|
2,979
|
13,512
|
23.1 (22.2–23.9)
|
358
|
1,255
|
29.5 (25.5–33.6)
|
2008
|
2,927
|
13,373
|
22.6 (21.7–23.4)
|
395
|
1,571
|
23.2 (20.8–25.6)
|
2009
|
2,669
|
13,028
|
21.3 (20.5–22.1)
|
453
|
1,693
|
24.8 (22.4–27.2)
|
2010
|
2,849
|
13,776
|
21.2 (20.4–22.0)
|
467
|
1,736
|
25.2 (22.7–27.7)
|
2011
|
2,778
|
13,291
|
21.2 (20.4–22.0)
|
506
|
1,790
|
28.1 (25.3–30.8)
|
2012
|
2,792
|
13,409
|
21.0 (20.2–21.8)
|
520
|
1,932
|
26.0 (23.5–28.5)
|
2013
|
2,758
|
13,306
|
20.7 (19.9–21.5)
|
549
|
2,124
|
23.9 (21.7–26.1)
|
2014
|
2,843
|
13,829
|
20.5 (19.7–21.2)
|
643
|
2,417
|
24.5 (22.4–26.5)
|
2015
|
3,169
|
14,410
|
21.6 (20.8–22.4)
|
550
|
2,172
|
23.8 (21.6–26.0)
|
2016
|
3,073
|
14,312
|
21.0 (20.2–21.8)
|
504
|
2,079
|
21.5 (19.4–23.7)
|
2017
|
3,235
|
14,404
|
21.6 (20.9–22.4)
|
569
|
2,113
|
23.9 (21.6–26.2)
|
2018
|
3,047
|
13,851
|
21.2 (20.4–22.0)
|
534
|
2,067
|
23.0 (20.8–25.2)
|
2019
|
3,201
|
13,965
|
21.9 (21.1–22.7)
|
548
|
2,266
|
20.4 (18.5–22.3)
|
2020
|
3,111
|
13,164
|
22.7 (21.9–23.5)
|
610
|
2,198
|
24.3 (22.1–26.5)
|
Females
|
2006
|
4,009
|
15,326
|
23.9 (23.1–24.6)
|
306
|
942
|
28.4 (24.7–32.1)
|
2007
|
3,994
|
15,601
|
23.1 (22.4–23.9)
|
324
|
1,088
|
25.3 (22.2–28.5)
|
2008
|
3,816
|
15,342
|
22.3 (21.6–23.1)
|
424
|
1,354
|
27.0 (23.7–30.3)
|
2009
|
3,669
|
15,125
|
21.5 (20.8–22.3)
|
440
|
1,406
|
28.1 (24.9–31.2)
|
2010
|
3,718
|
15,351
|
21.4 (20.6–22.1)
|
430
|
1,465
|
24.6 (21.6–27.5)
|
2011
|
3,609
|
14,899
|
21.4 (20.7–22.2)
|
492
|
1,660
|
25.5 (22.6–28.4)
|
2012
|
3,684
|
15,051
|
21.3 (20.6–22.2)
|
492
|
1,692
|
24.3 (21.6–27.1)
|
2013
|
3,577
|
14,861
|
21.1 (20.3–21.8)
|
552
|
1,843
|
24.2 (21.6–26.8)
|
2014
|
3,643
|
15,065
|
21.2 (20.5–22.0)
|
598
|
2,049
|
25.0 (22.3–27.8)
|
2015
|
3,997
|
16,002
|
21.8 (21.1–22.5)
|
553
|
1,950
|
24.7 (21.8–27.7)
|
2016
|
3,827
|
15,527
|
21.2 (20.5–22.0)
|
566
|
1,939
|
22.9 (20.3–25.4)
|
2017
|
3,939
|
15,376
|
22.2 (21.4–22.9)
|
515
|
1,883
|
21.0 (18.5–23.5)
|
2018
|
3,895
|
14,799
|
22.8 (22.0–23.6)
|
566
|
1,995
|
22.7 (20.0–25.4)
|
2019
|
3,914
|
14,942
|
22.3 (21.5–23.1)
|
568
|
1,999
|
21.5 (19.1–24.0)
|
2020
|
3,844
|
14,233
|
23.8 (22.9–24.6)
|
645
|
2,055
|
24.2 (21.6–26.9)
|
Abbreviation: CI, confidence interval.
a Anticoagulation = ICD-10 code Z92.1 (long-term [current] use of anticoagulation).
b Reference population: hospitalized patients without I60/I61 in 2006.
Fig. 2 Gender-specific age-standardized case fatality (%) in persons with I60/I61 stratified
by long-term use of anticoagulation, Germany, 2006–2020. I60: subarachnoid bleeding;
I61: intracerebral bleeding; ICD-10 Z92.1 (long-term, current, use of anticoagulation);
CI: confidence interval; weights for age standardization from hospitalized patients
without I60/I61 in 2006.
Discussion
Our analysis shows that the rates of cases hospitalized for ICB having LUAs do not
follow the dramatic increase in LUA prescriptions and were much lower than the rates
in all hospitalized cases, generally. The age-standardized mortality decreased continuously
in those with LUAs but not in those without LUAs. This finding is all the more interesting
as in the ICB cases without LUAs such a significant decrease is not present. Age-standardized
mortality remains unchanged in females and decreases only slightly in males in this
group.
A Global Burden of Disease Study reported the incidence of stroke in 204 countries
and territories from 1990 to 2019.[9] Ischemic stroke constituted 62.4% of all incidents of stroke in 2019, while ICB
constituted 27.9% and subarachnoid hemorrhage (SAH) constituted 9.7%. In France, data
from the National Stroke Action Plan reported a decrease in age-standardized rates
of patients hospitalized for ischemic stroke by 1.5% in those aged ≥65 years from
2008 to 2014. The rate of patients hospitalized for ICB was stable (+2.0%).[10] According to the Danish Stroke Registry, the rates of ischemic stroke and ICB declined
in people aged ≥50 years, but the rate of ICB remained constant at 11.2 to 11.4% in
the period from 2005 to 2018.[11] The rate of anticoagulants in the ICB cases increased from 5.9 to 10.4%. Another
report from Denmark reported that from 2005 to 2018, use of antithrombotic drugs,
especially VKAs, was associated with ICB. Although use of oral anticoagulants (OACs)
in the general population increased substantially during the study period from 3.8
to 11.1%, the incidence rate of ICB decreased.[12] An increase in the prescription rate of DOACs is reported worldwide.[13] Sheth et al analyzed British multisource health and prescription records data to
investigate the association between prescription rate of DOACs and stroke/bleed events
during the period from 2013 to 2016.[14] During the study period, the proportion of DOAC prescriptions increased at an average
rate of 122% per annum. DOAC prescription was associated with 50% reduction in ischemic and hemorrhagic strokes. Among patients with atrial
fibrillation (AF) with prior ICB, DOAC use was associated with lower rates of ICB
and major bleeding compared with warfarin use, whereas the rate of ischemic stroke
was similar in the two groups. A nationwide cohort study from Taiwan showed that compared
with warfarin use, DOAC use was associated with a statistically significant lower
risk of all-cause mortality and ICB.[15]
Our results show two different associations of ICBs and LUAs. First, the increase
in LUAs is much lower in hospitalized ICB cases than in all the other cases. Second,
mortality of ICB cases with LUAs decreases, whereas mortality in ICB cases without
LUA does not.
We cannot really explain the first effect. As ICBs are triggered by LUAs, the rate
of ICB in patients with LUAs should be higher than that in the general population.
One possible explanation might be that the causes of LUA-associated ICBs and non-LUA-associated
ICBs are different. The Oxford Community Stroke Project (OCSP; 1981–1986) and the
Oxford Vascular Study (OXVASC; 2002–2006) reported exactly the opposite.[16] They reported that the incidence of ICB associated with premorbid hypertension dropped
(0.37%; 95% CI: 0.20–0.69; p = 0.002), but the incidence of ICB associated with antithrombotic use increased (7.4%;
95% CI: 1.7–32; p = 0.007) resulting in substantially decreasing overall incidence of ICB (rate ratio:
0.53; 95% CI: 0.29–0.95; p = 0.03). Another explanation could be that preventive treatment of unruptured intracerebral
aneurysm has increased.[17] Data from the National Inpatient Sample (2004–2014) were extracted, including patients
with a primary diagnosis of a subarachnoid bleeding (SAB) or unruptured intracranial
aneurysms treated by clipping or coiling. With each passing year, there is a significant
decrease in relative risk ratio of undergoing treatment for an SAB, concomitant with
a stable annual risk of undergoing treatment for unruptured intracranial aneurysms.
This trend is supported by a study from Korea carried out in the period from 2005
to 2015 using the nationwide database of the Korean National Health Insurance Service
in South Korea. The marked increase in the detection and treatment of unruptured aneurysm
(UA) might have contributed to the decreasing incidence of SAH.[18] Such an increase in preventive UA treatment should affect both groups of ICB cases,
those with and those without LUAs. A third reason could be that the population treated
with LUAs has changed to a healthier population. The change from C, Herzinsuffizienz”
Herzinsuffizienz (“Congestive heart failure”) 1; H, Hypertonie, auch behandelt (“Hypertension”)
1; A, Alter >75 Jahre (“Age”) 1; D, Diabetes mellitus (“Diabetes”) 1; S2, Vorangegangener
Schlaganfall oder transitorische ischämische Attacke” TIA (“Stroke”) 2 (CHADS2 score)
to Congestive heart failure (Herzinsuffizienz) 1 ; Hypertension 1; Age (>75 Jahre)
2; Diabetes mellitus 1; Stroke / TIA 2; Vascular disease (z.B. pAVK, vorangegangener
Herzinfarkt, schwere Verkalkung der Aorta) 1; Age (65–74 Jahre) 1; Sex category (weibliches
Geschlecht) 1 (CHA2DS2-VASc score) increases the rate of patients with AF who should
have OACs from 66 to 94%.[19]
[20] AF patients were treated with LUAs because of minor criteria such as a female with
other vascular diseases. It is unclear which of these three aspects have the largest
impact on our findings. Given the differences between our data and data from other countries, more information
on the type of anticoagulation and primary diagnosis as well as cause of bleeding
in patients without anticoagulation would be necessary.
The second finding might be explained by the differences in stroke volume associated
with DOACs. Patients with DOAC ICB showed lower ICB volume and better clinical outcomes
than patients with warfarin ICB.[21] The onset of warfarin ICB was frequently observed in the morning and evening delaying
diagnosis and treatment, whereas DOAC ICB did not show any specific onset time. In
contrast, there is lack of a decrease in age-standardized mortality in female ICB
cases without LUAs and only a slight decrease in males in the observation period of
15 years, which is more difficult to explain. A population-based study from South
Texas reported a decrease in ICB incidence, but an unchanged case fatality and long-term
mortality in the period from 2000 to 2010.[22] A systematic review from 2019 reported no significant reduction of mortality at
1 month (–0.4% per year; 95% CI: –1.0 to 0.2; p = 0.155) in the last three decades.[23] A population-based cohort study from Denmark analyzed the mortality in ICB patients
from 2004 to 2017 and found an improvement over time in the unadjusted mortality rate
after 30 days.[24] They discussed the changes in the risk profile, the increasing proportion of less
severe strokes, and improvements by early admission of ICB patients to specialized
stroke care units as possible explanations. Such effects should have influenced the
ICB mortality in patients without LUAs in Germany as well.
Strength and Limitation
A major strength of this study is the large data set, which includes virtually all
German hospitals. This allows a unique view at the current clinical practice. Moreover,
to the best of our knowledge, there is currently no other publication addressing this
topic from a population-based viewpoint in Germany.
However, there are factors that limit our results. First, our study design did not
allow control for confounding including indications for treatment, quality of treatment,
polypharmacy, and relevant comorbidities. Second, the analyses are not on a per-patient
basis, and patients may be included multiple times. Moreover, the DRG data do not
allow for the differentiation between first-line treatments and revision procedures.
Third, we cannot assess if and how coding errors may have impacted our analysis. Thus,
undercoding can be assumed, especially for the use of anticoagulants (ICD code Z92.1)
Conclusion
Our results show that the increase in LUAs in ICB cases is much lower than that in
the general hospitalized population. We also found a decrease in intrahospital mortality
rates in ICB cases with LUAs from 2006 to 2020, but not in ICB cases without LUAs.
We cannot explain the specific reasons for these differences, and further research
is needed with better knowledge of the cases.