Twenty-year mortality of adult patients with primary immune thrombocytopenia: a Danish population-based cohort study

Authors

  • Henrik Frederiksen,

    Corresponding author
    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    2. Department of Haematology, Odense University Hospital, Odense, Denmark
    • Correspondence: Henrik Frederiksen, MD, PhD, Department of Haematology, Odense University Hospital, Sdr Boulevard 29, DK-5000 Odense C, Denmark.

      E-mail: hef@dadlnet.dk

    Search for more papers by this author
  • Merete Lund Maegbaek,

    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    Search for more papers by this author
  • Mette Nørgaard

    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    Search for more papers by this author

Summary

Studies have reported a 1·3- to 2·2-fold higher mortality rate among patients with primary immune thrombocytopenia (ITP) compared to the general population. However, long-term mortality estimates as well as cause-specific mortality data are sparse. In our population-based cohort of adult patients with newly diagnosed ITP and up to 37 years of follow-up, the 5-year, 10-year and 20-year mortality among the ITP patients was 22%, 34% and 49%, respectively. The mortality in the ITP cohort was consistently higher than in the in the general population cohort yielding an adjusted hazard ratio (HR) of 1·5 [95% confidence interval (CI): 1·2–1·8]. The adjusted HRs of mortality due to cardiovascular disease, infection, bleeding and haematological cancer were 1·5 (95% CI: 1·1–1·5), 2·4 (95% CI: 1·0–5·7), 6·2 (95% CI: 2·8–13·5) and 5·7 (95% CI: 2·1–15·7), respectively, whereas mortality due to solid cancer and other causes were similar in ITP patients and the general population. We conclude that mortality rates among ITP patients are higher than in the general population, predominantly as a result of increased cardiovascular disease, infection, bleeding and haematological cancer cause-specific mortalities.

Primary immune thrombocytopenia (ITP) is an autoimmune disease with reduced platelet survival and production. (Schmidt & Rasmussen, 1985; Gernsheimer et al, 1989) Although severe bleeding and infections among patients with ITP are rare (Frederiksen et al, 2012), studies conducted in Holland, UK and Denmark during the past decade have reported that morbidity and mortality among ITP patients is increased compared to the general population. (Portielje et al, 2001; Schoonen et al, 2009; Nørgaard et al, 2011; Frederiksen et al, 2012) Using different comparison groups, these studies reported that adult ITP patients have an increased relative mortality ranging from 1·3- to 2·2-fold that of the general population. (Portielje et al, 2001; Schoonen et al, 2009; Nørgaard et al, 2011) One of these studies comprised both primary and secondary ITP patients, and found that patients who were reclassified as secondary ITP within a median follow-up of 9·4 years had a sixfold higher mortality than the general population while patients with primary ITP had a 1·3-fold increased mortality. (Portielje et al, 2001) Also, among chronic ITP patients, those refractory to splenectomy have been reported to have a substantially higher mortality than patients in remission after splenectomy (McMillan & Durette, 2004).

The excess mortality among ITP patients has been attributed to infection and bleeding with cause-specific mortality estimates ranging from six to eight times that of the general population (Nørgaard et al, 2011).

Long-term mortality data among ITP patients are sparse. In 1999 a population-based cohort was described, comprising incident primary adult ITP patients diagnosed between 1973 and 1995. (Frederiksen & Schmidt, 1999) In the present study we assess long-term overall and cause-specific mortality within this cohort with up to 37-years follow-up as well as comparing it to mortality among the general population.

Methods

This cohort study was conducted in Funen, a Southern County of Denmark, with a population of more than 450 000 inhabitants, which is representative sample including 9% of the entire Danish population. In Denmark, all residents have un-fettered access to free tax-supported health care, including free access to hospital admissions and outpatient visits at hospitals. Since 1968, the Danish Civil Registration System (CRS), has provided all residents of Denmark with a unique, permanent 10-digit civil registration number, enabling unambiguous individual-level linkage among all Danish registries. (Pedersen, 2011) The civil registration number is a prerequisite for obtaining any form of social benefit in Denmark, including health care. (Frank, 2000) The CRS is continuously updated and keeps records of date of birth, sex, date of emigration and vital status for all Danish residents (Pedersen, 2011).

Immune thrombocytopenia cohort

In 1995–96, a population-based cohort of all newly diagnosed primary ITP patients aged ≥15 years, living in the well-defined county of Funen, was identified. (Frederiksen & Schmidt, 1999) The primary ITP patients were diagnosed at any Danish hospital or out-patient clinic, during a 22-year period from 1973 to 1995. (Frederiksen & Schmidt, 1999) The Danish National Registry of Patients (DNRP) contains information on virtually all admissions to public hospitals since 1977; visits to outpatient clinics or emergency rooms at hospitals have been included since 1995 (Lynge et al, 2011). Data recorded in the DNRP include civil registration number, dates of outpatient visits, hospital admissions and discharges, and up to 20 diagnoses coded by physicians according to the World Health Organization International Classification of Diseases, 8th revision (ICD-8) covering 1977–1993 and 10th revision (ICD-10) thereafter (Lynge et al, 2011). In the county of Funen where the primary ITP cohort was identified, registration of patients started earlier than in the DNRP. In the Funen Patient Administrative System (FPAS) of diagnoses, registration commenced 1 April 1973 for in-patient admissions and out-patient clinic contacts were registered from 1983 using the exact same data collection method as the DNRP.

The patients were identified through their ICD-8 or ICD-10 diagnosis codes in the FPAS (from 1973) and DNRP (from 1977) and the primary ITP diagnoses were verified by review of the medical records. (Frederiksen & Schmidt, 1999) From the medical records data including first recorded date with thrombocytopenia (date of diagnosis), platelet count, age, sex, bleeding symptoms, medical treatment and splenectomy were registered.

General population comparison cohort

To compare survival after the ITP diagnosis with that of the general population, a general population comparison cohort was identified matched to the ITP patients at the date of diagnosis. For each patient with primary ITP, ten randomly selected comparison cohort members were identified from the CRS, matched on sex and age in the calendar year at the time of diagnosis. Members of the comparison cohort were assigned the same index date as their matched primary ITP patients. Information on vital status and migration for both the ITP patients and the comparison cohort was retrieved from the CRS.

Cause-specific mortality

Information on causes of death was retrieved from the Danish Register of Causes of Death, which includes data on underlying and immediate causes of death for all deaths among residents of Denmark since 1970. (Helweg-Larsen, 2011) The underlying causes of death are grouped into categories that are based on ICD-8 diagnosis codes from 1970 to 1994 and ICD-10 diagnosis codes thereafter. For our analysis, we included the following six categories of causes of death: cardiovascular diseases other than intracranial haemorrhage, haematological cancers, all other cancers, infection, haemorrhagic episodes and others. The ICD-8 and ICD-10 groupings are presented in Appendix S1. The analyses regarding causes of death were restricted to deaths occurring before 1 January 2011.

Comorbidity and other covariates

We used the diagnoses recorded in the DNRP to assess comorbidity in both the primary ITP and the comparison cohort. Comorbidity was classified according to the 19 diseases included in Charlson Comorbidity Index (CCI), a validated measure to predict 1-year mortality using comorbidity data. (Charlson et al, 1987) We calculated CCI scores on the basis of all hospital diagnoses recorded in the DNRP from 1977, up to the date of primary ITP diagnosis or index date among the comparison cohort. Based on CCI scores we defined only two levels of comorbidity used in the analyses (‘none’, corresponding to no recorded underlying diseases, or ‘any’).

Covariates included in the analyses of overall mortality comparing ITP patients with the general population comparisons were age at diagnosis, sex, year of index date and comorbidity score at diagnosis. We also compared mortality within the primary ITP cohort between patients with different characteristics including the same covariates as well as bleeding symptoms and platelet count at diagnosis.

Statistical analysis

Follow-up started on the date of first primary ITP diagnosis (first recorded date with thrombocytopenia)/index date and ended at emigration, death or 1 January 2012, whichever came first.

We used the Kaplan–Meier Method to assess survival and computed cumulative mortality estimates after 5-, 10- and 20 years, respectively. Cox proportional hazards regression analysis was used to compare rates of mortality among primary ITP patients and members of the comparison cohort. We estimated hazard ratios (HR) and associated 95% confidence intervals (CI), with adjustment for age, sex, year of ITP diagnosis and comorbidity, taking the competing risk of other causes of death into account. Although the comparison cohort members were matched on age and sex, small differences were observed in the age distribution between the two cohorts (Table 1). These covariates were therefore also included in the regression analyses to control for any age- or sex-related residual confounding. For each of the six categories of underlying causes of deaths we estimated the cause-specific mortality among primary ITP patients and the population comparison cohort using the cumulative mortality. We used Cox proportional hazards regression analysis to estimate HR and 95% CI, treating other causes of death as competing risks. For analyses of cause-specific death, follow-up ended on 1 January 2011 for technical reasons.

Table 1. Descriptive characteristics of 221 adult patients with primary ITP and 2210 matched general population comparison cohort members.
 ITP patients N (%)Population comparison cohort N (%)
  1. a

    Year of diagnosis for immune thrombocytopenia (ITP) patients or index date for matched comparisons.

N (%)221 (100)2210 (100)
Age at ITP diagnosis (years)
15–2947 (21·3)468 (21·2)
30–5976 (34·4)774 (35·0)
≥6098 (44·3)968 (43·8)
Sex
Female139 (62·9)1390 (62·9)
Male82 (37·1)820 (37·1)
Yeara
1973–8054 (24·4)540 (24·4)
1981–8748 (21·7)480 (21·7)
1988–95119 (53·9)1190 (53·9)
Comorbidity score
0189 (85·5)1948 (88·1)
≥132 (14·5)262 (11·9)
Death during follow-up
No102 (46·2)1167 (52·8)
Yes119 (53·8)1043 (47·2)

The assumption of proportional hazards was assessed graphically in all models and found to be appropriate. This study was approved by the Danish Data Protection Agency (Record no. 2010-41-5668). The initial ITP cohort study was approved by The Regional Scientific Ethical Committee (Record no. 95/199).

Results

Descriptive data

The ITP cohort comprised 221 primary ITP patients – 139 (63%) were women and 82 (37%) men (Table 1). Median age was 54·9 years. At the time of ITP diagnosis, 175 patients (79%) presented with skin and/or mucosal bleeding and 46 (21%) were asymptomatic. The initial platelet count was <10 × 109/l among 101 (46%), <20 × 109/l among 138 (62%), and <50 × 109/l among 189 (86%). During the first 2 weeks after diagnosis 153 (69%) patients were treated with corticosteroids, 14 (6%) were treated with corticosteroids in combination with other drugs, including intravenous immunoglobulin (IVIg), and 54 (25%) were not treated. Splenectomy was performed in 65 (29%) patients after a median of 4·4 months (inter quartile range (IQR): 2·2–15·7 months). For the survival analyses the ITP patients were followed for a median of 197·3 months (IQR: 85·4–283·2 months). No patients or controls were lost to follow-up.

Mortality

Since diagnosis, 119 (54%) of the primary ITP patients have died and mortality among primary ITP patients was consistently higher than among the comparison cohort members (Fig 1). The overall 5-, 10- and 20-year cumulative mortality rates among the ITP patients were 22%, 34% and 49%, respectively (Table 2). In the population comparison cohort, 5-, 10- and 20-year cumulative mortality rates were 12%, 23% and 42%, respectively (Table 2). This corresponded to an adjusted HR of 1·5 (95% CI: 1·2–1·8). Mortality was highest among males, patients ≥60 years, and patients with comorbidity, as expected (Table 2). Mortality among primary ITP patients younger than 60 years of age was low and not statistically significantly different from comparison cohort members in the same age groups (Table 2). Although estimated with low precision, mortality rates among patients with both ITP and comorbidity were comparable to mortality rates among general population members with comorbidity (Table 2).

Table 2. All-cause 5-, 10- and 20-year cumulative mortality rate among 221 adult patients with primary ITP and 2210 matched general population comparison cohort members. Hazard ratio, taking the competing risk of other causes of death into account, is estimated across the entire study period.
All-cause mortality5-year mortality rate10-year mortality rate20-year mortality rateHazard ratio
ITP cohort (n)Deaths in ITP cohort (n)ITP patients (95% CI)Population comparison cohort (95% CI)ITP patients (95% CI)Population comparison cohort (95% CI)ITP patients (95% CI)Population comparison cohort (95% CI)Crude HR (95% CI)Adjusted HR (95% CI)a
  1. ITP, immune thrombocytopenia; HR, Hazard ratio; 95% CI, 95% confidence interval.

  2. a

    Adjusted for age (15–29, 30–59, ≥60 years), sex, year of ITP diagnosis and comorbidity score.

  3. b

    Year of diagnosis for ITP patients or index date for matched comparisons.

Overall2211190·22 (0·17–0·28)0·12 (0·10–0·13)0·34 (0·28–0·40)0·23 (0·21–0·25)0·49 (0·43–0·56)0·42 (0·40–0·44)1·23 (1·05–1·54)1·48 (1·20–1·82)
Age at ITP diagnosis (years)
15–294740·04 (0·01–0·16)0·00·06 (0·02–0·19)0·00 (0·00–0·02)0·06 (0·02–0·19)0·02 (0·01–0·03)2·23 (0·75–6·62)1·99 (0·64–6·17)
30–5976260·08 (0·04–0·17)0·02 (0·01–0·03)0·11 (0·05–0·20)0·06 (0·04–0·08)0·24 (0·16–0·37)0·17 (0·14–0·20)1·33 (0·88–2·02)1·34 (0·88–2·04)
≥6098890·41 (0·32–0·51)0·25 (0·22–0·28)0·64 (0·55–0·74)0·47 (0·44–0·51)0·91 (0·82–0·96)0·81 (0·78–0·84)1·68 (1·32–2·13)1·52 (1·19–1·94)
Sex
Female139750·19 (0·14–0·27)0·09 (0·07–0·10)0·29 (0·22–0·37)0·18 (0·16–0·20)0·48 (0·39–0·56)0·38 (0·36–0·41)1·62 (1·25–2·09)1·50 (1·15–1·94)
Male82440·26 (0·18–0·37)0·17 (0·14–0·19)0·42 (0·32–0·53)0·32 (0·28–0·35)0·53 (0·42–0·64)0·48 (0·44–0·51)1·52 (1·09–2·11)1·47 (1·04–2·04)
Yearb
1973–8054320·20 (0·12–0·34)0·10 (0·08–0·13)0·28 (0·18–0·42)0·20 (0·17–0·24)0·41 (0·29–0·55)0·40 (0·36–0·44)1·41 (0·96–2·07)1·39 (0·94–2·04)
1981–8748250·19 (0·10–0·33)0·08 (0·06–0·11)0·27 (0·17–0·42)0·16 (0·13–0·20)0·49 (0·36–0·64)0·36 (0·32–0·40)1·57 (1·01–2·44)1·75 (1·12–2·74)
1988–95119620·24 (0·17–0·32)0·14 (0·12–0·16)0·39 (0·31–0·48)0·27 (0·25–0·30)0·54 (0·45–0·64)0·45 (0·42–0·48)1·67 (1·27–2·24)1·47 (1·10–1·96)
Comorbidity score
0189900·19 (0·14–0·25)0·08 (0·07–0·09)0·28 (0·22–0·35)0·17 (0·16–0·19)0·43 (0·36–0·50)0·36 (0·34–0·38)1·76 (1·39–2·23)1·26 (1·02–1·57)
≥132290·41 (0·26–0·60)0·39 (0·33–0·45)0·66 (0·49–0·81)0·65 (0·59–0·70)0·88 (0·74–0·97)0·86 (0·81–0·90)0·96 (0·57–1·63)1·06 (0·72–1·57)
Figure 1.

All-cause mortality among 221 adult patients with primary immune thrombocytopenia and 2210 matched general population comparison cohort members. The graphs represent the survival curves for the two groups with surrounding 95% confidence intervals.

Table 3 compares mortality across different patient characteristics. The 5-, 10-, and 20-year mortality estimates were comparable according to sex, year of ITP diagnosis, comorbidity, symptoms at diagnosis and platelet count at diagnosis (Table 3).

Table 3. All-cause 5-, 10- and 20-year cumulative mortality rates among 221 adult patients with primary ITP. Hazard ratio, taking the competing risk of other causes of death into account, is estimated across the entire study period. The Hazard ratio is estimated with a reference within each category.
 ITP cohort (n)Deaths in ITP cohort (n)5-year mortality in patients with ITP (95% CI)10-year mortality in patients with ITP (95% CI)20-year mortality in patients with ITP (95% CI)Crude HR (95% CI)Adjusted HR (95% CI)a
  1. ITP, immune thrombocytopenia; HR, Hazard ratio; 95% CI, 95% confidence interval.

  2. a

    The HRs are adjusted for age, sex, year of ITP diagnosis, comorbidity score, bleeding at diagnosis and platelet count at diagnosis. Within each strata the same exposure variable is not included in the regression analyses (e.g., age is not controlled for in the age stratum).

Age at ITP diagnosis (years)
15–294740·04 (0·01–0·16)0·06 (0·02–0·19)0·06 (0·02–0·19)1·01·0
30–5976250·08 (0·04–0·17)0·11 (0·05–0·20)0·24 (0·16–0·37)4·65 (1·62–13·33)4·67 (1·61–13·51)
≥6098870·41 (0·32–0·51)0·64 (0·55–0·74)0·91 (0·82–0·96)33·42 (12·08–92·47)32·99 (11·61–93·74)
Sex
Female139750·19 (0·14–0·27)0·29 (0·22–0·37)0·48 (0·39–0·56)1·01·0
Male82440·26 (0·18–0·37)0·42 (0·32–0·53)0·53 (0·42–0·64)1·10 (0·76–1·59)1·18 (0·79–1·75)
Year of ITP diagnosis
1973–8054320·20 (0·12–0·34)0·28 (0·18–0·42)0·41 (0·29–0·55)1·01·0
1981–8748250·19 (0·10–0·33)0·27 (0·17–0·42)0·49 (0·36–0·64)0·93 (0·55–1·57)0·97 (0·57–1·64)
1988–95119620·24 (0·17–0·32)0·39 (0·31–0·48)0·54 (0·45–0·64)1·14 (0·74–1·75)0·75 (0·46–1·21)
Comorbidity score
0189900·19 (0·14–0·25)0·28 (0·22–0·35)0·43 (0·36–0·50)1·01·0
≥132290·41 (0·26–0·60)0·66 (0·49–0·81)0·88 (0·74–0·97)3·33 (2·18–5·11)1·38 (0·84–2·24)
Symptoms at diagnosis
No bleeding symptoms46230·22 (0·12–0·37)0·39 (0·27–0·55)0·51 (0·37–0·67)1·01·0
Bleeding symptoms175960·22 (0·16–0·44)0·32 (0·26–0·40)0·49 (0·41–0·57)0·97 (0·61–1·63)1·04 (0·60–1·81)
Platelet count at diagnosis (× 109/l)
>2080400·18 (0·11–0·28)0·34 (0·25–0·45)0·48 (0·38–0·60)1·01·0
10–2040230·28 (0·16–0·44)0·35 (0·22–0·52)0·48 (0·34–0·64)1·10 (0·66–1·85)1·04 (0·60–1·81)
<10101560·23 (0·16–0·32)0·33 (0·25–0·43)0·51 (0·41–0·61)1·09 (0·72–1·63)0·89 (0·53–1·50)

Information regarding cause of death was available for the 116 (52%) primary ITP patients and 1006 (46%) general population comparison cohort members who died before 1 January 2011. An additional five comparison cohort members died before this date but cause of death was not registered. Mortality rates due to cardiovascular disease, infection, bleeding and haematological cancer among the primary ITP patients were statistically significantly above those of the comparison cohort, whereas mortality due to solid tumours and other diagnoses were comparable between the two cohorts (Table 4). The adjusted HRs of mortality due to cardiovascular disease, infection, bleeding and haematological cancer were 1·5 (95% CI: 1·1–1·5), 2·4 (95% CI: 1·0–5·7), 6·2 (95% CI: 2·8–13·5) and 5·7 (95% CI: 2·1–15·7), respectively (Table 4). Of the eight primary ITP patients that died from infectious causes, the diagnoses were registered as pneumonia in three cases and pleural empyema, pulmonary abscess, miliary tuberculosis, diverticulitis of the colon with perforation and abscess and urinary tract infection in one case each.

Table 4. Cause–specific 5-, 10- and 20-Year cumulative mortality rates as a percentage among 221 adult patients with primary ITP and 2205 matched general population comparison cohort members. Hazard ratios, taking the competing risk of other causes of death into account, is estimated across the entire study period.
Mortality causeDeaths in ITP cohort, n (%)Deaths in general population cohort n (%)5-year mortality rate In percent10-year mortality rate In percent20-year mortality rate In percentHazard ratios
ITP patients (95% CI)Population comparison cohort (95% CI)ITP patients (95% CI)Population comparison cohort (95% CI)ITP patients (95% CI)Population comparison cohort (95% CI)Crude HR (95% CI)Adjusted HR (95% CI)a
  1. ITP, immune thrombocytopenia; HR, Hazard ratio; 95% CI, 95% confidence interval.

  2. a

    Adjusted for age, sex, year of ITP diagnosis, comorbidity score.

Cardiovascular42 (19)391 (17)8·2 (5·0–12·2)5·1 (4·2–6·1)12·3 (8·3–17·0)9·2 (8·1–10·5)17·6 (12·9–23·0)16·7 (15·1–18·3)1·2 (0·9–1·6)1·5 (1·1–2·1)
Infection8 (4)53 (2)1·8 (0·6–4·3)0·3 (0·1–0·6)2·3 (0·9–4·9)1·1 (0·7–1·6)3·8 (1·8–7·1)2·3 (1·7–3·0)1·7 (0·8–3·5)2·4 (1·0–5·7)
Bleeding12 (5)41 (2)5·0 (2·6–8·4)0·6 (0·4–1·0)5·0 (2·6–8·4)1·1 (0·7–1·6)5·4 (3·0–9·0)1·5 (1·1–2·1)3·2 (1·7–6·0)6·2 (2·8–13·5)
Any solid cancer21 (10)230 (10)1·4 (0·4–3·7)2·5 (1·9–3·2)5·5 (3·0–9·1)5·1 (4·2–6·0)9·5 (5·9–14·1)9·4 (8·2–10·7)1·0 (0·6–1·6)1·0 (0·7–1·7)
Haematological cancer7 (3)21 (1)0·5 (0·0–2·3)0·2 (0·1–0·5)1·4 (0·4–3·7)0·4 (0·2–0·8)2·8 (1·2–5·7)1·0 (0·6–1·5)3·7 (1·6–8·6)5·7 (2·1–15·7)
Other26 (12)270 (12)5·0 (2·6–8·4)2·8 (2·2–3·6)7·3 (4·3–11·2)6·1 (5·1–7·1)10·9 (7·1–15·5)10·7 (9·4–12·1)1·1 (0·7–1·6)1·1 (0·7–1·7)
Still alive105 (47)1199 (54)        

Discussion

By means of our population-based data with long and complete follow-up we found that 5-, 10- and 20-year mortality rates among adult ITP patients were consistently higher than those of the general population. During the study period the mortality was 1·5-fold higher among ITP patients than in the general population.

Previous studies of mortality in ITP have been heterogeneous in terms of patient age and disease duration, and many studies lack a control group. Many have focused mainly on the proportion of deaths in the study population (Guthrie et al, 1988; Schattner & Bussel, 1994; Stasi et al, 1995; Neylon et al, 2003; Michel et al, 2011), and only a few studies have assessed time-specific mortality (Sailer et al, 2003) or compared mortality to the general population. (Portielje et al, 2001; Schoonen et al, 2009; Nørgaard et al, 2011) Our overall all-cause mortality rate ratio (MRR) of 1·5 is in line with relative mortality measures in some previous reports (Portielje et al, 2001; Schoonen et al, 2009), whereas Nørgaard et al (2011) reported a higher MRR (2·2) in their cohort including chronic ITP patients only diagnosed between 1996 and 2007. In their study the 5-year mortality rate was 24%, but increased from 19% to 30% in the halves of the study period. (Nørgaard et al, 2011) In our patient population, diagnosed between 1973 and 1995, the 5-year mortality rate was similar, averaging 20% across the three study periods but estimated with low precision due to small numbers in the strata. Sailer et al (2003) studied 130 consecutive ITP patients from a single institution and reported a 5-year and 10-year mortality of just 7% and 14%, respectively. The reason for this discrepancy is unknown.

There are several explanations for the increased mortality in ITP patients and not all seem to be directly associated with diagnosis or treatment. Previous studies have reported that an elevated risk of bleeding, as well as infections and malignancies, contribute to this. (Portielje et al, 2001; Nørgaard et al, 2011) Malignant diseases following an ITP diagnosis may be due to immunosuppressant therapy, including corticosteroids, which increases the risk of malignancies (Sorensen et al, 2004; Marcen, 2009). In fact, one the patients who developed a haematologcal malignancy was an ITP patient who, due to treatment with cyclophosphamide, later developed a therapy-related myelodysplasia. Also, the diagnostic possibilities of haematological malignancies have evolved since the beginning of this study. Some of the ITP patients may have already had an undiagnosed low-grade malignancy, such as a small chronic lymphocytic leukaemia clone at ITP diagnosis.

The mortality risk in relation to splenectomy in ITP patients is difficult to assess because splenectomized ITP patients are younger and have less comorbidity (Jensen et al, 2011). However, newer population-based studies have shown that mortality is equal among splenectomized and unsplenectomized ITP patients (Yong et al, 2010), but the risk of infections requiring hospitalization remains 1·4 times higher for splenectomized ITP patients. (Thomsen et al, 2009).

Although our study is population-based, including a large primary ITP cohort with long follow-up, it also has limitations. The rarity of the disease resulted in mortality estimates in subgroups with limited statistical precision. At the beginning of the study period only patients with in-hospital admissions were registered and therefore included in our study. In-hospital patients may, therefore represent a selected subgroup with a clinical adverse outcome. However, mortality estimates were comparable for patients diagnosed across the different study periods. The causes of death were retrieved from a registry based on information from death certificates. Given that the autopsy rate in Denmark has declined, from 75% in the 1970's to below 20% at present, for death occurring in hospital and is even lower among deaths occurring outside hospital, information regarding cause of death may not be accurate. (Helweg-Larsen, 2011) It seems, however, unlikely that misclassification of causes of death would result in systematic differences between ITP patients and the general population although this cannot be completely ruled out. The impact of this potential bias on our results is therefore not clear.

The proportion of primary ITP patients that were registered with bleeding-related deaths (n = 12) remained stable during the periods of follow-up whereas a slight increase in this proportion was observed for the comparison cohort (Table 4). For an aging population increasingly treated with platelet anti-aggregants and other anticoagulant drugs, the observation is as expected for the comparison cohort. For the primary ITP patients, causality with time is difficult to assess because contraindications to such drugs associated with varying platelet counts are likely to affect bleeding risk.

Within the primary ITP cohort and general population cohort, comorbidity was strongly associated with mortality, as expected (Tables 2 and 3). The difference in mortality rates between ITP patients and the general population were much less pronounced when ITP patients with comorbidity were compared with general population members with comorbidity (Table 2), indicating that among patients with other chronic diseases primary ITP is not the major cause of death. The CCI in our study was based only on hospital diagnoses. Not all comorbid diagnoses are captured in this way because some diagnoses, such as diabetes, can be made and followed solely in general practice and will therefore not be included. This may result in differential misclassification if ITP patients are more likely to have a comorbid diagnosis registered in hospital than the general population comparison cohort members. Whether this potential bias influences our results is unclear. The CCI is a weighted index based on 1-year mortality risks with the included conditions. Although it has been applied in many studies no gold standard for measuring comorbidity exists (Sarfati, 2012). We included baseline CCI scores, which may not account for the effects of comorbidity on long-term mortality.

The CCI includes acquired immunodeficiency syndrome (AIDS) and leukaemia as comorbid diagnoses. Both would result in another diagnosis than primary ITP if such patients were also suffering from thrombocytopenia. Therefore, these two diagnoses can only result in a CCI point score among the comparison cohort members in our study. Given that both AIDS and leukaemia are rare diagnoses this is unlikely to influence our results. However, this potential for more comorbidity among the comparison cohort members would result in our relative mortality estimates being conservative.

We conclude that both long- and short-term mortality rates among primary ITP patients were above those of the general population. Mortality was elevated, predominantly due to increased cardiovascular disease, infection, bleeding and haematological cancer cause-specific mortalities. The cohort was defined in a period were the only treatment options included various immunosuppressive medical or surgical therapies. It will therefore be interesting to follow how newer thrombopoiesis-stimulating agents might affect mortality in this patient group.

Acknowledgement

This study was supported by the Clinical Epidemiological Research Foundation.

Authors' contribution

H Frederiksen conceptualized the idea for the study. All co-authors participated in the study's design. M Maegbaek performed data analyses. H Frederiksen and M Nørgaard directed the statistical analyses. H Frederiksen wrote the first draft of the paper, and all authors participated in writing subsequent drafts.

Conflict of interest disclosure

H Frederiksen has received honoraria for speaker engagements with GlaxoSmithKline. M Maegbaek and M Nørgaard report no potential conflicts of interest. The Department of Clinical Epidemiology, Aarhus University Hospital, receives funding for other studies on ITP from Amgen in the form of research grants to (and administered by) Aarhus University. None of these studies have any direct connection to the present study.

Ancillary