Article Text

Download PDFPDF

Research paper
Long term (13 years) prognosis after primary intracerebral haemorrhage: a prospective population based study of long term mortality, prognostic factors and causes of death
  1. Björn M Hansen1,2,
  2. Ola G Nilsson3,4,
  3. Harald Anderson5,
  4. Bo Norrving1,2,
  5. Hans Säveland3,4,
  6. Arne Lindgren1,2
  1. 1Department of Clinical Sciences Lund, Neurology, Lund University, Lund, Sweden
  2. 2Department of Neurology, Skåne University Hospital, Lund, Sweden
  3. 3Department of Clinical Sciences Lund, Neurosurgery, Lund University, Lund, Sweden
  4. 4Department of Neurosurgery, Skåne University Hospital, Lund, Sweden
  5. 5Department of Clinical Sciences Lund, Cancer Epidemiology, Lund University, Lund, Sweden
  1. Correspondence to B M Hansen, Department of Neurology, Skåne University Hospital, 221 85 Lund, Sweden; Bjorn.Hansen{at}


Introduction Many studies have focused on short term mortality after primary intracerebral haemorrhage (ICH) whereas long term prognosis and causes of death have been less studied. We therefore examined these issues in a population based cohort of 1 year ICH survivors.

Methods ICH patients in a defined Swedish population (1.14 million inhabitants) were prospectively registered during 1996. Patients surviving 1 year after ICH onset were followed-up regarding survival status and cause of death until December 2009 using data from the National Census Office and the National Cause of Death Register. Patient prognosis was also compared with the general population using official Swedish mortality data. Clinical and radiological prognostic factors were evaluated.

Results Of 323 patients with ICH, 172 (53%) survived after 1 year, 127 (39%) after 5 years and 57 (18%) after 13 years. Mortality of the 172, 1 year survivors (mean age 67.7 years at ICH) persistently exceeded expected mortality; 13 years post ictus survival was only 34% compared with 61% in the general population. Of 115 deaths among the 172, 1 year survivors, 36% were from cerebrovascular disease and 19% from ischaemic heart disease. Independent risk factors for death among 1 year survivors were age (HR 1.08 per year; 95% CI 1.06 to 1.10; p<0.001), diabetes mellitus at baseline (HR 2.10; 95% CI 1.18 to 3.74; p=0.012) and anticoagulant therapy (HR 1.99; 95% CI 1.12 to 3.53; p=0.018) at ICH onset.

Conclusions One year survivors after ICH had a substantial and persisting excess mortality compared with the general population. Major causes of death were stroke and ischaemic heart disease.


Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Approximately 10% of all strokes are caused by primary intracerebral haemorrhage (ICH).1 The short term mortality amounts to about 40% at 1 month and 55% at 1 year after ICH onset.2 Factors predicting mortality at 30 days and 1 year include age, level of consciousness at admission, haematoma volume, presence of intraventricular haemorrhage (IVH), heart disease and infratentorial haematoma.3 ,4

The subsequent prognosis of patients who survive the first critical year after ICH is less investigated, despite considerable interest from both patients and the treating physician. Although previous studies have analysed some aspects of this problem,5–9 there is no prospective study on the long term prognosis (>10 years) after ICH in a population based cohort, reporting both prognostic factors for survival and causes of death.

In the present study, we therefore addressed the following questions in a prospective population based cohort of ICH patients. (1) Do 1 year survivors after ICH have an excess mortality compared with the population in general? (2) Are there variations in excess mortality over time? (3) Which risk factors affect the further survival for late stage survivors after ICH? (4) Which are the main causes of death among 1 year survivors after ICH?


Study population

We performed a long term follow-up of ICH patients registered prospectively during 1996 in a defined population in Southern Sweden. The incidence of ICH and short term mortality has previously been reported.4 ,10 A total of 341 consecutive patients with primary ICH, inpatients as well as those diagnosed at autopsies, were identified; 12 of these patients had two ICH during 1996 and were registered twice.10 In this follow-up, a single entry was created for each patient, with the first ICH in 1996 registered as the index ICH, hence 329 unique ICH patients were followed-up in this study. The population of 1.14 million inhabitants was served by 12 hospitals and four pathology/forensic departments. This long term follow-up study on patients registered in 1996 was approved by the regional ethics review board.


All patients with spontaneous primary ICH (ie, with no evidence of tumour, arteriovenous malformation, arterial aneurysm, cerebral infarction or trauma as aetiology) were registered. The ICH diagnosis was confirmed by CT or at autopsy. Participating hospitals routinely performed CT for all stroke patients within 1–2 days of admission and all included patients had CT performed within this time frame. Patients diagnosed and treated solely as outpatients were not registered. One of the authors (OGN) or a radiologist assessed the neuroradiological characteristics: ICH volume (estimated by the A×B×C/2 method11), ICH location (lobar, deep, cerebellar or brainstem) and substantial ventricular extension (defined as blood in two or more ventricles). Potential risk factors were collected from medical records, as described previously.4


By using a specific personal identification number assigned to all Swedish citizens, individual information regarding survival status, date of death and cause of death were obtained from the National Census Office and the Cause of Death Registry (managed by the National Board of Health and Welfare).12 Survival status and cause of death were followed until 4 December 2009. To determine expected survival in the corresponding general population, age, sex and calendar year specific mortality data were obtained from Statistics Sweden. Cause specific mortality for the main causes of death in the general population (by sex and in 10 year age spans between 1997 and 2009) was acquired from official statistics, available from the National Board of Health and Welfare's online database on cause of death statistics.13

Official death certificates are completed for all Swedish citizens at the time of death and then processed by the Swedish National Board of Health and Welfare to determine the underlying cause of death as well as contributing causes of death. Cause of death data are classified according to the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) and registered in the Cause of Death Registry.12 The authors categorised underlying causes of death in our cohort into the following groups: (1) cerebrovascular disease (corresponding to ICD-10, I60-69; subdivided into death related to index ICH, new ICH, cerebral infarction or cerebrovascular disease without further specification); (2) ischaemic heart disease (corresponding to ICD-10, I20-I25; including patients with cardiac failure with no other apparent cause); (3) cancer (corresponding to ICD-10, C00-D48); (4) infection; (5) trauma; and (6) other causes of death (including for example, renal failure, dementia and gastrointestinal bleeding). Contributing causes of death were used as additional information if categorisation could not be completed with only the underlying cause of death.

Statistical methods

Overall survival was estimated by means of life table methods, and Cox regression analyses were used to study the prognostic impact of clinical and radiological variables with all cause mortality as the endpoint. Univariate analyses of single risk factors were adjusted for age and sex; multivariate analysis included age, sex and significant (p<0.05) univariate factors. The Stata Statistical Software, Release 11 (StataCorp LP, USA) was used for statistical analysis. Cumulative mortality due to cerebrovascular disease, ischaemic heart disease and other causes were determined considering competing risks and illustrated as stacked cumulative mortality curves.

Expected and relative survival,14 and expected total number of deaths, were determined using current age, sex and current calendar specific Swedish mortality rates and a Stata routine developed by Dickman et al.15 The relative survival ratio (RSR) is the quotient between the observed survival of a study cohort divided by its expected survival according to the general population mortality. RSR values below 1 indicate excess mortality in the study group while values above 1 indicate mortality below the reference population average. The annual RSR at a certain follow-up year is defined as the ratio between the observed conditional survival to live 1 year more and the corresponding expected survival rate, and this method has previously been used in both cancer epidemiology and mortality studies of stroke patients.15 ,16 Survival analyses were performed for the whole cohort of 1 year survivors as well as for subgroups determined by sex, age and ICH location.

To calculate the expected number of deaths due to cerebrovascular disease (ICD-10, I60-69), ischaemic heart disease (ICD-10, I20-I25), cancer (ICD-10, C00-D48) and other causes, respectively, the number of follow-up years in the cohort was divided into sex and 10 year age groups, and then multiplied by the average (over calendar years) Swedish official mortality for each of these disease groups.


Study population

A total of 329 patients with ICH were initially enrolled, of which six were lost to follow-up due to missing or incorrect personal identification numbers. Of the remaining 323 patients (mean age 70.4 years), 172 were still alive 1 year after ICH onset (mean age at baseline ICH onset 67.7 years) and they were followed until time of death (n=115, 67%) or until December 2009 (ie, at least 13 years; n=57, 33%). Five year and 10 year mortality rates in the whole original cohort were 61% (n=196) and 77% (n=249), respectively. Additional demographics and distribution of risk factors are shown in table 1.

Table 1

Characteristics of the total primary intracerebral haemorrhage patient cohort and of the 1 year survivors after intracerebral haemorrhage

Long term survival among the 172, 1 year survivors

Among the 172, 1 year survivors, 127 (74%; 95% CI 67% to 80%) survived 5 years after ICH onset and 74 (43%; 95% CI 36% to 50%) survived 10 years after ICH. The corresponding expected survival rates for the age and sex matched general population were 85% and 68%, respectively (figure 1A). The annual RSR among the 1 year survivors was <1 for each specific follow-up year, suggesting that the excess mortality of ICH patients persisted during the study period (figure 1B). The absolute difference between cumulative observed and expected survival 13 years after ICH onset was 27%. Similar differences (ranging between 22% and 29%) were observed for the subgroups determined by age, sex and ICH location.

Figure 1

(A) Observed survival with 95% CI and expected survival for the 172, 1 year survivors after primary intracerebral haemorrhage (ICH). (B) Annual relative survival ratio (RSR) among the172, 1 year survivors after ICH.

Cox regression analysis showed that both diabetes mellitus and anticoagulant therapy at ICH onset predicted (p<0.05) higher long term mortality after adjusting for sex and age at ICH diagnosis. Long term mortality was not significantly affected by antiplatelet therapy, level of consciousness at admission (GCS), ICH location, ICH volume, IVH, hypertension, cerebrovascular disease or heart disease previous to ICH (table 2).

Table 2

Cox regression analysis among 1 year survivors after primary intracerebral haemorrhage, with all cause mortality as the endpoint

Multivariate analysis including diabetes mellitus, anticoagulant therapy, sex and age as covariates showed that age was an independent risk factor for long term mortality for the 1 year survivors (HR 1.08 for each year of higher age; 95% CI 1.06 to 1.10; p<0.001). Diabetes mellitus (HR 2.10; 95% CI 1.18 to 3.74; p=0.012) and anticoagulant therapy (HR 1.99; 95% CI 1.12 to 3.53; p=0.018) also had a negative impact on long term survival (table 2).

Cause of death

The three major causes of death among Swedish inhabitants aged 65–84 years, between 1997 and 2009, were cancer, ischaemic heart disease and cerebrovascular disease. The cumulative numbers of observed deaths among 1 year ICH survivors were: 42 (36%) due to cerebrovascular disease, 22 (19%) due to ischaemic heart disease (of whom three had cardiac failure), 13 (11%) due to cancer and 38 (33%) due to other causes. The corresponding expected numbers of deaths (calculated from Swedish official statistics) were: 6.8 due to cerebrovascular disease, 13.8 due to ischaemic heart disease, 14.1 due to cancer and 18.9 due to other causes. Figure 2 illustrates mortality as stacked cause specific cumulative mortality curves for cerebrovascular and cardiovascular and other causes of death, respectively; together with table 3 it shows that ischaemic heart disease and cerebrovascular disease were the major causes of death at 1–9 years after ICH whereas cancer and other causes dominated at 9–13 years. Contributing causes of death among the eight patients with cause of death determined as related to the index ICH were: dementia, pneumonia and urinary tract infection, due to sequelae of their index ICH.

Table 3

Cause of death 1–13 years after intracerebral haemorrhage and patient years at risk

Figure 2

Stacked cumulative curves for mortality due to cerebrovascular disease, ischaemic heart disease and other causes (including cancer, infection and trauma) for 1 year survivors after primary intracerebral haemorrhage (ICH).


Our study shows that 1 year survivors after ICH have a persisting excess mortality during long term follow-up after surviving the first critical year after ICH. The excess mortality amounts to 27% at 13 years after stroke onset compared with the population in general; the annual RSR values of <1 shows that the excess mortality continues during the whole follow-up period.

One possible cause for the observed excess mortality is that risk factors for ICH occurrence, such as hypertension,17 atherosclerosis and cerebral amyloid angiopathy,18 are important risk factors for other types of chronic vascular disease. In the present study, vascular disease was a common denominator of risk factors affecting long term survival (age, diabetes mellitus and anticoagulant therapy) and a major cause of death (cerebrovascular and ischaemic heart disease). This corresponds well with the perception that ICH is a manifestation of chronic general vascular disease. Several other chronic diseases have also been reported to have an aggregation of comorbidities and multimorbidity, especially with advancing age,19 and this aggregation is likely related to increased mortality. Another possible cause of the observed excess mortality is that the grade of disability after stroke negatively affects survival,20 perhaps due to complications following immobilisation.

The 61% 5 year mortality for our total cohort is in the lower end compared with previous studies (65–75%, estimated from survival tables).6–9 Despite low 5 year mortality rates, the 77% 10 year mortality rate in our study corresponds well with previous studies, spanning from 1985 to 1998, in which 10 year (in one study, 11 year) mortality rates ranged from 76% to 82%.5–9 A Finnish study reported that increased mortality, compared with the general population, could only be observed during the first 6 years after ICH onset.7 This is in contrast with our study where the excess mortality remained during the whole 13 year study period. This difference might be due to chance because of the relatively small numbers of late survivors in both studies, and that the study designs differed: the Finnish study included 28 day survivors whereas we only included 1 year survivors. We showed that cerebrovascular disease (36%) was the major cause of death among our patients, followed by ischaemic heart disease (19%) and cancer (11%). These results correspond with results from two previous studies on 28 day survivors after stroke where 16–23% died from ischaemic heart disease and 32–34% from cerebrovascular disease.5 ,7

Risk factors affecting short term survival, such as level of consciousness at admission, infratentorial ICH, ICH volume, IVH and heart disease,3 ,4 diminished in importance for long term prognosis after ICH in our study. These factors seem to be primary linked to the acute stage of the disease and are less important regarding long term survival. If patients who die in the acute/intermediate phase (0–365 days) after ICH are included in a long term survival analysis (follow-up >10 years), the short term risk factors for survival may conceal the importance of long term risk factors. We think that this is an additional reason for studying long term survival among 1 year survivors after ICH.

In our study, diabetes mellitus at baseline was associated with long term mortality which is in line with results from previous studies where short term and overall survival after ICH was negatively affected by hyperglycaemia at admission and by diabetes mellitus.6 ,21–23 No correlation between diabetes mellitus and ICH occurrence could be found in a meta-analysis of eight studies on risk factors for ICH.24 This supports our finding that the presence of diabetes mellitus is a proper independent risk factor for long term survival after ICH. Anticoagulation therapy is a risk factor for ICH occurrence25 and short term mortality26 but why our results indicate that anticoagulation therapy at baseline affected the prognosis after ICH in the long term (HR 1.99) needs consideration. One explanation is that anticoagulation therapy is permanently withdrawn after ICH onset, and among patients with concomitant atrial fibrillation lack of anticoagulant therapy doubles the overall risk of death.27 Another reason could be that patients with anticoagulant therapy at ICH ictus have a poorer functional outcome28 which might lead to reduced long term survival. The high relative death rates in cardiovascular disease, alongside anticoagulant therapy as a long term risk factor, indicate the importance of future studies to address the effects of reinstating anticoagulant therapy after ICH. Other factors influencing long term prognosis among 28 day survivors have been reported to be older age, male sex and heart failure.7 We could not identify gender or heart failure as long term risk factors in our study which may be related to the fact that our analysis comprised 1 year survivors—that is, a considerably longer time after ICH onset when these factors may have diminished in importance.

Our study has the advantage of investigating a large, prospectively recruited population based cohort of ICH patients in which national personal identification numbers enabled long term follow-up via official sources. One limitation is that no new data regarding risk factors were collected after baseline registration. This mainly affects the analysis of anticoagulant and antiplatelet therapy, as medications may have been altered after inclusion. Additionally, due to the study design, we lacked information on indications for baseline anticoagulant therapy; consequently, the survival analysis could not be adjusted for the conditions that the anticoagulant therapy was intended to treat. Classification of cause of death was based on official death certificates completed by the treating physician and our data are therefore dependent on external judgement. This may have biased the registration of new strokes as causes of death, including whether these were ischaemic or haemorrhagic. However, CT rates among stroke patients in Sweden are high29 ,30 and it can therefore be assumed that the detection rate of ICH among stroke patients is high. Even though our study does not cover patients treated solely as outpatients, the number of stroke patients not admitted to hospital was estimated to be less than 5% in one South Swedish study which overlaps our study in time (1989–1998), and between 7% and 8% in more recent studies (1999–2000 and 2001–2002).29 ,31 ,32 High hospital admitting rates and high CT scan rates among Swedish stroke patients further indicate that our study has a good population coverage despite mainly being hospital based.

Further studies on the long term prognosis after ICH could focus on: (1) risk factor treatment after the acute and subacute phases of ICH; (2) how total burden of comorbidities in patients with ICH affects prognosis; (3) risk of recurrent stroke, ischaemic as well as haemorrhagic; (4) pooling data from previous studies on long term survival to increase the accuracy of prognostication and to address geographical and chronological differences in survival; (5) incorporating data on functional outcomes and hospitalisation patterns after ICH to gain knowledge on how the disease affects patient quality of life; and (6) identifying potential economic benefits of prevention.


One year survivors after primary ICH have a constant excess mortality compared with the general population, and long term mortality seems to be influenced by risk factors that are potentially modifiable. Major causes of death during follow-up were stroke and ischaemic heart disease. Further investigation of how to optimise treatment for late stage survivors after ICH is needed to reduce their excess mortality.



  • Contributors BMH prepared the first draft of the manuscript. BMH and AL made substantial contributions to conception and design; acquisition of the data; analysis and interpretation of the data; drafting the article; revising it critically for important intellectual content; and gave final approval of the submitted manuscript. OGN made substantial contributions to conception and design; acquisition of the data; analysis and interpretation of the data; revising the article critically for important intellectual content; and gave final approval of the submitted manuscript. HS, BN and HA made substantial contributions to conception and design; analysis and interpretation of the data; revising the article critically for important intellectual content; and gave final approval of the submitted manuscript.

  • Funding The study was supported by the Swedish Research Council (K2010-61X-20378-04-3), the Swedish Stroke Association, the Freemasons Lodge of Instruction EOS in Lund, the King Gustaf V and Queen Victoria's foundations, the Rut and Erik Hardebo Donation Fund, Region Skåne and Lund University.

  • Competing interests None.

  • Ethics approval The study was approved by the regional ethics review board, Lund.

  • Provenance and peer review Not commissioned; externally peer reviewed.