More ‘malignant’ than cancer? Five-year survival following a first admission for heart failure

2.1. Data sources

The Information and Statistics Division of the National Health Service in Scotland collates data on all National Health Service hospital admissions within the country via the Scottish Morbidity Record Scheme 8. As part of this scheme, each person admitted to any Scottish hospital since 1981, has been assigned a unique identifying number. This number is then used for all subsequent hospitalisation. Furthermore, as part of standard procedure, information from patient case records in 1991 were used, at the time of hospital discharge, to code diagnoses according to the Ninth Revision of the World Health Organisation International Classification of Diseases (ICD) 9. The term ‘discharge’ includes both live discharges and deaths. Importantly, continuous auditing of ICD coding of hospital admissions within Scotland, suggest that these data are 90% accurate overall in relation to correctly identifying patient diagnoses 10. These data therefore permit an analysis of each individual's hospitalisation record since 1981 with a high degree or accuracy.

Data concerning individual patients on the Scottish Morbidity Record database are also linked, using probability matching, to information held by the General Register Office for Scotland relating to all deaths occurring within the UK. As migration rates to countries outside of the UK are low, this system is estimated to be at least 98% accurate overall. Therefore, accurate survival data for all individuals admitted to hospital since 1981 are also available.

The Scottish Cancer Registry also contains these data, in addition to more specific information concerning individuals according to the anatomical site and type of their primary tumour 11. The four most common sites of cancer registered in Scottish men during 1991 were, in rank order, lung (30%), large bowel (12%), prostate (11%) and bladder (8%). The equivalent figures for women were breast (24%), large bowel (14%), lung (10%) and ovarian cancer (5%) 11.

Socio-demographic and survival data, including sex- and age-specific life-expectancy, for the whole of Scottish population (approximately 5.1 million) during the period of 1991–1996 are contained in the 1991 and 1996 Annual Reports of the Registrar General for Scotland 12,13.

2.2. Index hospitalisation

Data relating to a first hospitalisation for heart failure and acute myocardial infarction in 1991 were obtained from the linked database containing hospitalisation and survival data. Equivalent data relating to the cancer-related diagnoses of interest were obtained from the Scottish Cancer Registry.

All recorded admissions during the year 1991 were screened in order to identify and select all those patients with a hospitalisation primarily caused (recorded in the first position for the purpose of ICD9 coding) by the following:

We then excluded all patients who had a hospital admission for their ‘index’ condition (recorded in any position for the purpose of ICD9 coding) in the 10 years prior to 1991. For cancer patients a previous admission associated with any malignant neoplasm (ICD9 140–208) also meant exclusion from analysis.

2.3. Baseline data

Each hospital record provides information concerning the individuals’ age, sex, date of admission and usual postcode of residence. The latter were used to derive the Carstairs–Morris Deprivation index 14. This index, based on an official Scottish-wide census performed in 1991, can be used to rank postcodes of residence into five deprivation categories (1-least, 5-most deprived) according to levels of employment, living conditions, car ownership and social class.

2.4. Survival rates

The Linked Database currently permits reliable analysis of survival data until 31 December 1996. All surviving patients were censored at this time-point to provide 5-year follow-up for every patient. If death from any cause occurred, length of survival was calculated from the date of recorded admission to date of recorded death.

2.5. Loss of expected life-years

All deaths occurring in individuals before their ‘expected’ age of death (determined by published life-expectancy tables for the age-matched population cohort in 1991 12,13) were defined as premature. The number of ‘expected’ life-years lost was also calculated by subtracting actual age at death from ‘expected’ age of death. Associated loss of expected life-years was then calculated as a median (IQR) for each diagnosis and as a rate of expected life-years per 1000 population.

2.6. Statistical analysis

Sex-specific life-tables and Kaplan–Meier survival curves for each condition were constructed from survival data using the actuarial life-table method. Comparative, sex- and age-specific survival data for the entire Scottish population in 1991 (a census year) were calculated from mid-year population estimates from 1991 to 1996 12,13. Multiple logistic regression models were then used to calculate the probability of death within 30 days and, for surviving patients, 31 days to 5-years adjusting for age and social deprivation. Age was entered into each logistic model as a continuous variable and social deprivation as the original index score from one to five. The six diagnoses were entered into each logistic model as one categorical variable with heart failure set as the lowest class. For both social deprivation and diagnosis, the lowest class was set at unity. Adjusted odds ratio (OR) and 95% confidence intervals (CI) for myocardial infarction and four types of cancer are therefore relative to that of heart failure. All logistic regression models were subject to the Hosmer–Lemeshow Goodness-of-Fit Test and were accepted as valid if the associated P-value was >0.05. For all analyses SPSS version 9.0 was used.

3.1. Cohort characteristics

In 1991, a total of 16 224 men experienced an initial hospitalisation for heart failure (n-3241), myocardial infarction (n-6932) or cancer of the lung, large bowel, prostate or bladder (n-6051). Similarly, 14 842 women experienced an initial hospitalisation for heart failure (n-3606), myocardial infarction (n-4916), or cancer of the breast, lung, large bowel or ovary (n-6320). Table 1 summarises the demographic characteristics of this cohort according to their index diagnosis. It also shows the associated annual population incidence rate for each type of admission.

Overall, there were more first-time admissions for myocardial infarction than heart failure. This difference was greater in men (more than double) than in women (approx. 40% more). Consistent with data emanating from the UK overall 11, mainland Europe 15 and the US 16, lung cancer was the most and second most common form of cancer in men and women, respectively. With the exception of acute myocardial infarction (mean age 64 years), male patients were predominantly older than 60 years of age; the age distributions being broadly similar for heart failure and the four types of cancer. Among women, however, there was a clear difference in the age distribution of those admitted with either cancer of the breast or ovary (mean age 62–64 years) and the remainder of the cohort (mean age 72–76 years). Although there were approximately 25% more heart failure than breast cancer-related admissions overall, only 20% of these were among women aged less than 65 years compared to 50% for breast cancer.

3.2. Unadjusted survival

For both men and women, lung cancer was associated with the poorest unadjusted survival rate with a median survival time of 3–4 months and only 5% of patients surviving to 5 years. However, heart failure was associated with the second poorest unadjusted survival rate with a median survival time of 16 months and only 25% of men and women surviving to 5 years. Overall, short, medium and longer-term survival rates varied widely according to the index diagnosis (see Fig. 1). For example, whilst myocardial infarction was associated with a similar (and therefore high) initial mortality rate at 30-days (between 20 and 25%), individuals with heart failure fared considerably worse thereafter. Among women, heart failure-related survival was broadly equivalent to that of ovarian cancer, whilst breast cancer was associated with twice the survival rate at 5 years (60%).

3.3. Adjusted survival relative to heart failure

With the major exception of lung cancer (with an approximate twofold and eightfold increased probability of death in the short and longer-term, respectively), heart failure was associated with the poorest longer-term adjusted survival in men. In women both cancer of the breast and large bowel were associated with better short-term survival rates in comparison to heart failure. Subsequent long-term survival was more favourable in the former and equivalent in the latter. Alternatively, cancer of the ovary and lung were associated with poorer adjusted survival rates overall in comparison to heart failure (see Table 2). As expected, age was a powerful prognostic factor in both models-each additional decade of age conferring an approximate four- and fivefold increased probability of death within 5 years for men and women, respectively. Fig. 2 shows the survival rates for the most represented age group for men (65–74 years) and women (70–79 years) relative to the entire population within that age group. A higher deprivation score was also associated with a two- to threefold higher probability of death relative to the lowest deprivation score.

3.4. Loss of expected life-years

Table 3 shows the total number of deaths associated with each diagnosis, the proportion of which occurred prematurely and the median number of ‘expected’ life-years lost during 5-year follow-up. As such, the relative impact of these conditions on loss of expected life-years varied according to the total number of individuals who died and their average age at death. In men, lung cancer was associated with both a large number of premature deaths and a significant number of expected life-years lost. Heart failure was also associated with a significant number of expected life-years lost (on average 9 years per person), being associated with more deaths than the combination of large bowel, prostate and bladder cancer. In women, despite the fact that proportionately more deaths occurred in those who had already exceeded average life expectancy, heart failure was second only to myocardial infarction in terms of the total number of premature deaths. Both lung and breast cancer, however, by virtue of a greater number of expected life-years lost per person had a greater impact on the population as a whole during this period.

4.1. More ‘malignant’ than cancer?

This is the first population-based study directly comparing heart failure vs. cancer-related survival. With the notable exception of lung cancer, a first admission for heart failure is commonly associated with a worse survival rate than an equivalent admission for its common precursor, myocardial infarction, and the most common types of cancer. This applies equally to men and women. The 5-year survival rate in the almost 7000 men and women (representing 0.14% of the population) initially admitted to a Scottish hospital with heart failure in 1991 was a dismal 25%. Even on an adjusted basis, heart failure-related survival was poorer than many of the common types of cancer. Significantly, the number of male deaths associated with heart failure was similar to that of lung cancer (approx. 2500 in each group) and only 400 fewer than that associated with acute myocardial infarction. In women, heart failure, followed closely by myocardial infarction, was associated with the greatest number of deaths overall (approx. 2700 deaths in each group).

4.2. Comparison with other studies

Is it possible to extrapolate these data to other developed countries? Both the 1- and 5-year survival rates in this Scottish cohort are equivalent to those of large-scale studies of hospitalised heart failure patients in both the US 5 and Australia 17. On an age-adjusted basis they are also broadly equivalent to the community-based Framingham cohorts 6,7. Cancer-related incidence and mortality rates in Scotland are consistent with that of the UK overall 11,15,18 and other northern European countries 15 but typically higher than those of the US 16 and southern European countries 15. For those countries like the US with equivalent heart failure-related, but lower cancer-related survival rates, therefore, the relative prognostic importance of heart failure compared to cancer may be greater than suggested by these data.

4.3. The burden of heart failure

Prioritising levels of health-care funding is always difficult. However, the overall impact of any disease state has to be measured not only in terms of absolute mortality rates, but the average age of those affected, its effect on quality of life and the type and cost of health-care resources it typically engenders. Despite relatively fewer deaths and better survival rates overall, we have again shown that breast cancer-related deaths often occur in relatively young women. The level of publicity and health-care funding breast cancer currently receives as an important health issue for women is no doubt justified on this basis 18. However, despite a greater number of women affected and dismal survival rates (in all age groups), heart failure is rarely identified as an important health issue for women. Additional data from the Scottish Cancer Registry 11, suggest that during the same 5-year period heart failure was associated with a greater number of premature deaths and associated loss of expected life-years than cancer of the ovary, cervix and uterus combined. Coupled with the data that link heart failure with poor survival rates overall 6,7,19, an increasing number of expensive hospital admissions 5, and poorer self-reported quality of life than any other common medical disorder 20, the current data reinforce the view that heart failure is a major health issue for both men and women.

4.4. Benefits of screening and palliative care programs

In considering the modern-day burden of heart failure, a further comparison with cancer is probably appropriate. In most developed countries, there are specific cancer registries, routine screening programs (e.g. mammography, cervical smears and faecal occult blood tests) and medical care that is closely regulated to ensure optimal treatment and the rapid introduction of new therapies 11,13,16,18. Moreover, the widespread implementation of formalised palliative care services also ensures that comprehensive care is delivered to patients with end-stage cancer 21. Despite recent concerns about the effectiveness of breast screening programs 22, data from both Europe 13,15 and the US 16 suggest that this co-ordinated response to cancer has resulted in significant reductions in the age-adjusted cancer death rate. Another analogy is perhaps, hypertension. The severe and deadly form, appropriately named ‘malignant hypertension’, has all but disappeared as a result of the introduction of hypertension screening programs and the widespread use of effective anti-hypertensive therapy 23. As a direct consequence, the prognosis of hypertension has improved and the incidence of hypertensive heart disease has declined dramatically 23.

4.5. Heart failure-specific initiatives

Despite the high prevalence of individuals within the community who have left ventricular systolic dysfunction associated with either untreated symptoms of heart failure, or at risk for developing overt heart failure in the future 24, there is a general lack of screening programs for the early detection and treatment of such individuals 25. Moreover, pharmacological agents with proven clinical benefit in the treatment of heart failure (e.g. inhibitors of angiotensin converting enzyme) are commonly under-utilised 26. More importantly, perhaps, is the general lack of specific heart failure registries and co-ordinated health-care services once heart failure is diagnosed. As such, there is only limited evidence to suggest that heart failure-related survival has improved in recent years 27. Both the Framingham 7 and Rochester 19 studies failed to document improved survival rates among individuals with heart failure prior to the widespread introduction of inhibitors of angiotensin converting enzyme. Although we, and others, have recently reported a decline in case fatality rates associated with this type of cohort in recent years 28,29, any improvement in this regard is modest when compared to those associated with most types of cancer. Within this context, these data support the need for health care initiatives that firstly recognise heart failure as a distinct and significant component of cardiovascular-related morbidity and mortality, and secondly, afford it the same type of commitment and health care resources usually reserved for high profile disease states. Nurse-led, comprehensive management programmes, for example, have been shown to minimise readmissions and prolong survival in heart failure patients 30 but remain under-funded. Such a response needs to be acted upon as a matter of urgency, especially if, as many suspect, heart failure reaches ‘epidemic’ proportions within the next 20 years 1–3,31.

4.6. Limitations

As with any study of this type there are a number of limitations that require comment. Firstly, we had to rely on discharge coding to identify the study cohort. Although internal validation studies of the Scottish Morbidity Record Scheme have proved this data to be quite accurate 10 and hospitalisation for heart failure is usually associated with more definitive investigation and treatment 32, the diagnostic accuracy and overall quality of data may vary on an institutional basis. Moreover, we do not know what proportion of individuals with a diagnosis of heart failure had left ventricular systolic dysfunction. Heart failure patients with normal systolic function probably have better survival rates than those with depressed function 33. We have therefore described the survival rate for a mixture of such patients, as has the Framingham 7, Rochester 19 and other studies of this type. We have also only studied hospitalised heart failure patients who potentially represent the severe end of the spectrum of this syndrome. However, community surveys demonstrate that most heart failure patients are admitted to hospital within 2 years of identification 32. Finally, we also have no data concerning causality of subsequent death. This is always problematic as heart failure commonly omitted from the death certificate as a contributory cause of death 34.