OUP user menu

Is there unmet need for implantable cardioverter defibrillators? Findings from a post-mortem series of sudden cardiac death

Deborah Chase, Paul J. Roderick, Hayley Burnley, Patrick J. Gallagher, Paul R. Roberts, John M. Morgan
DOI: http://dx.doi.org/10.1093/europace/eun114 741-746 First published online: 7 May 2008


Aims To establish whether sudden cardiac death (SCD) victims could have been identified prior to their event and considered for an implantable cardioverter defibrillators (ICDs).

Methods and results Consecutive post-mortem cases of adult SCDs presumed to be caused by a ventricular arrhythmia over 12 months (2002–03) from a defined catchment population, Southampton, UK (n = 443 824 adults aged ≥16 years). Pathological data were extracted from the post-mortem reports. Hospital and general practice (GP) notes provided data on previous symptoms, investigations, and cardiac disease history. Two electrophysiologists judged the appropriateness of each case for an ICD against National Guidance. Two hundred and fifteen cases met the inclusion criteria and lived within the catchment area. Agreement between experts on appropriateness for an ICD in those aged <80 years was good (kappa score of 0.64). Only one case (<1%) was considered appropriate for an ICD without requirement for further investigation. Forty-nine per cent of cases were considered to have required further cardiac investigations to determine appropriateness; these were mainly heart failure patients who had suffered a myocardial infarction (MI). Forty per cent of cases had no previous clinical evidence of confirmed or suspected heart disease. However, pathological data showed that 51% of cases had suffered a previous MI.

Conclusion Two-fifths of SCD victims had no recorded health service contact that would indicate increased risk of SCD within their lifetime. A large number of patients suffered previous cardiac events or symptoms suggestive of increased SCD risk but were not referred for further investigations. There is a need for better care pathways for patients post-MI to identify those requiring an ICD. The impact on the ICD rate of undertaking these extra investigations is uncertain.

  • Sudden cardiac death
  • Implantable cardioverter defibrillator
  • Case series
  • Post-mortem
  • Appropriateness


Implantable cardioverter defibrillators (ICDs) have been shown to be a cost-effective intervention in the management of patients at high risk of sudden cardiac death (SCD).1 Ventricular arrhythmias are the main triggers for SCD. These arrhythmias may develop during the course of an acute coronary event and can occur spontaneously in chronic heart disease.

In 2000, the National Institute for Clinical Excellence (NICE) recommended the use of ICDs for:

  1. secondary prevention, i.e. for patients who present, in the absence of a treatable cause, having survived a cardiac arrest or sustained ventricular tachycardia (VT) (and are significantly compromised);

  2. primary prevention, i.e. patients with a history of previous myocardial infarction (MI), with non-sustained ventricular arrhythmia (NSVT) and inducible VT on electrophysiological study.1 These recommendations were updated in 2005 to include patients with a previous MI and left ventricular dysfunction (without associated ventricular arrhythmia).

The 2000 NICE guidelines recommended a rate of 50 ICDs implanted per million population (pmp). At this time, the UK ICD rate was less than half this rate at ∼20 pmp.2 This rate has increased and the recommended UK rate of 50 pmp was reached in 2005.2 The UK ICD implantation rates are lower than rates in many European countries, and markedly lower than in the USA.3 Within England and Wales, there is a considerable variation in ICD rates between areas, even after adjusting for proxy measures of need.4 These discrepancies in rates suggest unmet need for ICDs. In fact, discrepancies are even greater now because indications for ICD were expanded in 2005.

The aim of this study was to establish whether victims of SCD, thought to be primarily due to ventricular arrhythmia, could have been identified prior to their death based on health service contacts and potentially been considered for an ICD.


The research design was a consecutive case series of SCD within a defined population. All records of post-mortem examinations undertaken at Southampton University Hospitals (SUHT) over a 12-month period were screened against the inclusion criteria from October 2002 to October 2003.


All SCDs were categorized using a modification of the method originally described by Davies in 1998 (Table 1).5 In deaths that were due to acute coronary disease, there was either evidence of an acute ischaemic event, such as a fresh coronary thrombus, or an acute MI with or without ventricular rupture (SCD category 1 within Table 1). Other deaths were classified according to the presence or the absence of: (i) significant coronary artery stenosis (defined as a luminal diameter of ≤1 mm or >75% stenosis of the lumen); (ii) healed MI; and (iii) ventricular hypertrophy with or without post-mortem evidence of heart failure. Patients dying of pulmonary embolism and vascular disorders such as dissecting or ruptured atheromatous aneurysms were excluded. Time of death from onset of symptoms was not used as a criterion for defining sudden death because of the large number of unwitnessed cases.

View this table:
Table 1

Cause of death in 366 sudden cardiac deaths

Category of SCDPathological findingsNo. of cases (%)
1. Certain evidence of an acute ischaemic eventAcute myocardial infarction, with or without rupture and/or fresh coronary thrombosis134 (37%)
2. Certain evidence of ischaemic heart diseaseClear evidence of healed myocardial infarction and at least one significant coronary artery stenosis but no fresh thrombosis or acute myocardial infarction116 (32%)
3. Evidence of ischaemic heart diseaseAt least one significant coronary artery stenosis but no fresh thrombosis, acute, or healed myocardial infarction. No other cause of death demonstrated63 (17%)
4. Evidence of heart muscle diseaseNo evidence of significant coronary heart disease. Left or right ventricular hypertrophy or dilatation with or without evidence of congestive cardiac failure. In some cases, specific macroscopic changes indicated a particular form of cardiomyopathy46 (13%)
5. No structural evidence of heart diseaseNo significant abnormalities on macroscopic or histological examination7 (2%)

Cases with evidence of an acute MI or acute coronary thrombosis (SCD category 1) were excluded from this study because ICD implantation is unlikely to influence the development of acute coronary disease, and most of these patients would not have benefited from ICD implantation during their lifetime.


The catchment area for SUHT was the former Southampton and South West Hampshire Health Authority area (population of ∼500 000 persons). This is a largely urban area in the South of England with an all-cause directly standardized mortality rate of 737 per 100 000 males and 501 per 100 000 females (in comparison with 807 and 553 per 100 000, respectively, in England).6 Cases who resided outside of this area were excluded from this study.

Data collection

Previous health service history and pathological data were collected from post-mortem reports, police forms, and from hospital and general practice (GP) notes. Health service history data were collected to determine whether patients could have been identified for further investigation or actual referral for ICD prior to their death from health service contacts.

All records of post-mortem examinations undertaken within the study timeframe were reviewed for study inclusion. These were identified from the SUHT post-mortem database. Pathologists were aware of the nature and purpose of the investigation. They were not required to alter their method of performing examinations but were encouraged to describe the findings in the heart and coronary arteries in more detail than usual.

Appropriateness rating

A simplified version of the Rand method was used.7 Experts were asked to independently rate the appropriateness of each case for an ICD, using information from each patient's health service history, and then to reach consensus through meeting and discussing any disagreements.

Criteria on the appropriateness for ICDs were developed by which each case could be judged. These were based on national guidance (the 2000 NICE recommendations1). Those cases where one could be most certain that patients should have been considered for an ICD were those meeting the NICE criteria for an ICD.

Those cases where one could be most certain that patients would not have been identified prior to death for an ICD were those having no recorded contact with health service providers in association with cardiac problems and/or symptoms suggestive of ventricular arrhythmias. These were cases with definite ‘de novo’ disease; with no opportunity for secondary prevention.

For cases not considered to be at either end of this spectrum, consensus from experts was used to determine the level of certainty that each case could or could not be identified.

Consensus method

Two SUHT Consultant Cardiologists (with special interest in cardiac electrophysiology), A and B, were provided independently with a standardized ‘proforma’ on each case in this study. The proforma provided information from hospital and GP notes on each patient's CHD history, i.e. symptoms, test results, and management.

The Consultants were familiar with the study case definition and case finding methods but were not provided with any pathological data on these cases. They were asked to determine if cases should have been considered for an ICD or further testing to determine appropriateness for ICD during their lifetime on the basis of their health service history. They were asked to assign each case to one of the five categories: (i) appropriate for ICD; (ii) inappropriate for ICD (contraindications); (iii) should have had further diagnostic tests (such as Holter monitoring, echocardiography, and electrophysiological study); (iv) no justification for further testing, inappropriate for ICD; and (v) no evidence of heart disease before death. They then met to discuss cases where they disagreed. A and B undertook this process for all cases. Analyses were undertaken on all cases and just those cases who were under 80 years at their time of death.

Statistical analysis for judging appropriateness of cases

The number of SCD cases that would have been appropriate for an ICD was presented. Characteristics of patients including pathology and previous CHD history were detailed. Agreement between raters (A and B) in judging the appropriateness of cases was also measured using the kappa statistic for all cases and just those cases who were under 80 years at the time of death.


Descriptive and pathological data

A total of 1038 adult post-mortems were undertaken in the study period at SUHT (94% Coroner's cases and 6% consented in hospital deaths). Six hundred of these cases were sudden deaths, and of these 366 (61%) were SCDs. The pattern of post-mortem diagnoses of the SCDs is summarized in Table 1.

Of the 366 SCDs, 134 (37%) had evidence of an acute MI or acute coronary thrombosis. These cases were excluded from further study. The remaining 232 SCDs (all cases in SCD categories 2–5 in Table 1) met the criteria for this study.

One hundred and seventy-nine of these patients (77%) had evidence of coronary narrowing and 116 had an associated healed myocardial infarct (50%). In 46 patients, there was undoubted cardiac disease such as left ventricular hypertrophy or dilatation and/or evidence of cardiac failure, but no significant coronary artery narrowing was demonstrated. Seven patients had died suddenly but had no macroscopic or microscopic evidence of cardiac disease (sudden arrhythmic death syndrome).

Two hundred and fifteen (93%) of the 232 cases meeting the study criteria resided within the study catchment area and were therefore included in this study. Descriptive data on the characteristics of the cases and data relating to the event itself are shown in Table 2. Most of the 215 cases were males and the elderly. The youngest case was 40 years of age at the time of death.

View this table:
Table 2

Descriptive and pathological data on the case series, number of cases, and percentage (n = 215, unless otherwise specified)

Descriptive data
Male gender138 (64%)
Age in years, median (min, max)75 years (range 40–94)
Ethnic originWhite 215 (100%)
Place of deathHome 155 (72%), hospital 17 (8%), public place 17 (8%), care home 11 (5%), other person's house 3 (1%), work 5 (2%), and no details 7 (3%)
Death witnessed (n = 194)72 (37%)
Symptoms reported prior to death? (n = 139)67 (48%)
Pathological data
Previous MI110 (51%)
Left main stem disease (LMS)7 (3%)
No. of vessels diseased (RCA, LAD, and circumflex excluding LMS)One vessel = 65 (30%), two vessels = 56 (26%), three vessels = 33 (15%), and no vessels diseased = 61 (28%)
Evidence of CABG15 (7%)
Aortic valve disease22 (10%)
  • RCA, right coronary artery; LAD, left anterior descending; CABG, coronary artery bypass grafting; PTCA, percutaneous transluminal coronary angioplasty.

Most of the deaths occurred at home (72%) and were not witnessed (63%). For 35% of cases, there was no information on symptoms prior to death. Symptom descriptions varied considerably; breathlessness was reported as a symptom for 34% of the 67 cases with symptoms reported prior to death, chest pain for 24%, blackout 5%, and more ambiguous symptoms, such as neck or shoulder pain (categorized under ‘other’ in the table), dizziness, ‘felt unwell’, etc., reported for other cases.

Previous history

Of the 215 cases, hospital notes could be identified for 138 (64%) cases and GP notes for 201 cases (93%). Hospital and/or GP notes could be identified for 209 (97% of cases). Table 3 shows the proportion of cases with a history of CHD as recorded in their hospital and/or GP notes.

View this table:
Table 3

Heart disease history in hospital and/or GP notes (n = 209)

HistoryTotal number of patients (%)
Previous cardiac arrest4 (2%)
Diagnosed with ventricular arrhythmia7 (3%)
Suffered previous clinical MI37 (18%)
Diagnosed with heart failure27 (13%)
Severe left ventricular dysfunction on echocardiogram9 (4%)
Evidence of heart failure on angiogram or chest X-ray10 (5%)
Diagnosed with other type of arrhythmia50 (24%)
Radiofrequency ablation0
CABG14 (7%)
PTCA2 (1%)
Other cardiac surgery6 (3%)
Pacemaker4 (2%)
Cardiomyopathy5 (2%)
SCD familial conditiona1 (<1%)
Angina56 (27%)
Diabetes20 (10%)
  • aExcluding hypertrophic cardiomyopathy, this is included under cardiomyopathy.

Only 2% (four cases) had a previous cardiac arrest and a further 3% (seven cases) diagnosed ventricular arrhythmia recorded in their notes. Of the seven patients with diagnosed ventricular arrhythmias, four suffered sustained VT (three of these in the acute phase post-MI) and three patients suffered non-sustained VT. As would be expected, high risk conditions for SCD, i.e. cardiomyopathy (2% of cases in this study) and hereditary conditions (one case), were rare.

Eighteen per cent of cases had a previous MI recorded in their notes but 51% of cases had clear pathological evidence of a previous MI at post-mortem suggesting either missing data or silent MI. Twenty-seven patients (13%) had a clinical diagnosis of heart failure, but only 44% (12) of these patients had had an echocardiogram.

Diagnosis of other types of arrhythmia, mainly atrial arrhythmias, was more common (24%). Almost half the patients had treated hypertension and about a third of cases had had angina.

Expert's appropriateness ratings

The kappa score was 0.50 for the appropriateness of ICD, suggesting moderate agreement between the two raters. Both raters agreed that only one case was appropriate for an ICD. Their reasons for deeming this case to be appropriate were (i) ‘known to be at high risk of SCD’ and (ii) ‘left ventricular function <35%’. This patient had cardiomyopathy during their lifetime but had no previous history of a cardiac arrest or ventricular arrhythmia. Of note, this patient was referred for electrophysiological examination within their lifetime but declined investigations. Both raters agreed that three of the four patients who suffered a previous cardiac arrest were not appropriate for an ICD, having suffered a cardiac arrest within 48 h post-MI. They also agreed that all seven of the patients who suffered a diagnosed ventricular arrhythmia should have been considered for an echocardiogram and/or electrophysiological study.

Reasons for assigning cases to the ‘should have been considered for further testing’ varied. The main reason was exploration of whether patients met the NICE recommendation for primary prevention of SCD. There was some disagreement between A and B on who was inappropriate for an ICD. Expert A used older age, i.e. over 80 years at the time of death, as a contraindication to ICD implantation and therefore classified these cases as inappropriate for an ICD. Expert B commented on the older age of patients but did not use this information to determine appropriateness.

Exclusion of cases over 80 years of age at the time of death

If cases aged 80 years or more at the time of death were excluded from the cohort, this left 146 cases with hospital and/or GP notes for ICD appropriateness assessment. The kappa score for agreement between experts A and B for this sub-group of the cohort was stronger than for the whole cohort, with k = 0.64.

Of these 146 cases, there were 31 cases in which there was disagreement between raters. The disagreement was not regarding appropriateness for an ICD but whether further testing was justified. These cases were discussed at a joint meeting between the raters. Final decisions on all cases aged <80 years at the time of death are shown in Table 4.

View this table:
Table 4

Final appropriateness ratings as agreed by the two experts in cases under age 80

CategoryNo. of cases (%)
Appropriate for an ICD1 (<1%)
Should have had further tests72 (49%)
No evidence to justify further testing11 (8%)
De novo: No pre-mortem evidence of heart disease or symptoms suggestive of heart disease58 (40%)
Further information required4 (3%)

A and B agreed that only one case aged <80 years at the time of death would have been appropriate for an ICD. Forty per cent of cases could be classed as de novo. Most cases, 49%, required further investigations to determine appropriateness of an ICD. The main reason for assigning the cases (∼80%) to this category was that the patient had suffered an MI with/without heart failure or NSVT but was not referred for further investigation of potential life-threatening ventricular arrhythmias.

For four cases, the experts requested further information on heart rhythm recordings to determine whether these patients had been misdiagnosed with a supra-ventricular arrhythmia rather than a ventricular arrhythmia during their lifetime.


The aim of this study was to establish whether SCD cases thought to be due to a ventricular arrhythmia could have been identified prior to death and potentially considered for an ICD. This study has suggested that very few of the 215 cases (<1% as agreed by experts) could have been definitely identified having met the NICE criteria for an ICD. Approximately 40% of SCD cases had no prior health service history of cardiac abnormalities or symptoms suggestive of ventricular arrhythmias.

Results from further investigations would have been required to determine the appropriateness of a large number of cases. The experts agreed that 49% of cases under 80 years at the time of death required further investigations (heart rhythm monitoring and subsequent electrophysiological study for cases with positive test results) to determine ICD appropriateness. The main reason for assigning the cases (∼80%) to this category was that the patient had suffered a clinical MI with/without heart failure or NSVT but was not referred for further investigation of potential life-threatening ventricular arrhythmias.

This is the first study to assess whether SCD cases could have been identified and potentially benefited from an ICD during their lifetime. Published studies have assessed the effectiveness and cost-effectiveness of ICDs and audited their use in current practice.4,8,9 Other studies have assessed the scope for preventing SCDs by identifying risk factors for SCD and found that the presence and severity of underlying heart disease is one of the most predictive risk factors for the future occurrence of SCD.10

In terms of age and gender and the proportion of witnessed deaths, our findings are similar to previous studies of SCD11 In the earliest contemporary study of 168 SCD patients, intraluminal or occlusive thrombi were detected histologically in 73%.12 A detailed histological study of coronary artery disease in 231 out of hospital sudden deaths in Nottingham in the early 1990s13 demonstrated coronary artery thrombosis in 47% of cases. A smaller but broadly similar study in the USA found thrombi in 52% of cases.14 In the UK, it is no longer possible to make detailed histological studies of coronary artery pathology in unselected post-mortems. Largely, on the basis of macroscopic evaluation, we found coronary thrombosis in 39% of cases, a broadly similar figure to that reported recently in a national survey of sudden deaths in England and Wales.11 It is possible that the underlying pattern of coronary artery disease causing sudden death has changed over the last 15 years since Davies' study in London.12 However, it is equally possible that without detailed histology, we underestimated the number of acute ischaemic deaths. This would result in the misclassification of some SCDs as being primarily caused by a ventricular arrhythmia.

There is evidence that the incidence of sudden death is declining in the USA15 and in Finland.16 However, there has been no significant reduction over a long period in Northern Ireland where the current incidence is 122 per 100 000 for men and 41 per 100 000 for women.17 There are no other more recent studies from outside the UK on the pathology of SCD in all age groups.

We used a broad definition for SCD that did not require the death to be witnessed. Other definitions include just those cases known to have experienced symptoms 1 h before death. Only 37% of cases in this study were witnessed. If only those cases known to have experienced symptoms within 1 h prior to death were included, this would have constituted just 33 cases within our series. There was a striking disparity between the number of cases with pathological evidence of healed MI and those with a clinical history of MI (51 vs. 18%). This could be due to missing data within the notes and/or patients experiencing a silent MI.

Independently, the experts were in agreement about those cases that were appropriate for an ICD or ‘de novo’. However, there was some disagreement regarding cases within the ‘should have had further tests’ category. In particular, one expert deemed cases over 80 years at the time of death as ‘inappropriate for an ICD’. Age per se is not a barrier to clinical treatment in the National Health Service. However, this expert took account of the need to balance the potential to benefit from therapy with co-morbidities in judging appropriateness for further tests. To enable consensus, cases over 80 years at the time of death were excluded from the final data analyses.

The strengths of this study are that cases were consecutive and were identified from within a defined catchment area and an explicit and standardized approach of pathological assessment, data extraction, and appropriateness rating was applied.

There were several potential limitations. The information on previous medical history was derived from medical notes; we had GP/hospital notes for 97% cases but data maybe inaccurate and or missing. Furthermore, one trained but non-clinical researcher extracted data from the notes using a standardized and comprehensive checklist. It is possible that information could have been missed. Sudden cardiac death cases that did not have a post-mortem were missed from this study. Only 22% of deaths in England are followed by post-mortem. This rate is lower still in Scotland and Northern Ireland (12 and 9%, respectively). Of concern is whether the cases in the SCD case series are representative of all SCDs primarily due to ventricular arrhythmias. An extrapolation of routine data on ischaemic heart disease mortality on deaths in the community and out of hospital cardiac arrest data collected within the same catchment area and timeframe of this study suggested that a small number of cases (∼30 SCD cases) may have been missed. This suggests that the SCDs in this case series represented the majority of cases occurring in the community in 2002–03. Lastly, appropriateness was assessed by two experts working within the same specialist centre. Ratings from a group of experts from different centres would have provided a better indication of national consensus on appropriateness. However, it was not feasible to convene a group of national experts within the timeframe of the study.

The generalizability of our results should be considered. Although the demographics of this study population are similar to England, albeit with a better health experience (i.e. a lower all-cause SMR), it is served by a tertiary cardiac centre, which could result in a higher investigation and referral rate. Therefore, our findings may underestimate unmet need.

This study has shown that although only one case within the case series should definitely have had an ICD, there may be significant unmet need, as very few patients had been investigated to determine their appropriateness for an ICD. The majority of patients rated as ‘should have had further investigations’ by raters were included within this category because there was no information within their notes on their left ventricular function. Now that the NICE guidelines (2000 and revised in 2005) are well established, further work should focus on developing care pathways taking account of the current availability and uptake of echocardiography and other investigations, particularly in patients who have survived an MI.

Furthermore, pathology data from this study highlight the significant number of patients who are unaware of suffering from a previous MI and the large number with no recorded health service contact in relation to cardiac abnormalities or possible arrhythmia symptoms within their lifetime. This finding highlights the importance of population-based primary prevention measures to prevent SCD.

Finally, we propose that pathologists should be encouraged to use the SCD grading scale (shown in Table 1) to determine the level of certainty associated with an SCD. The implementation of this scale would better enable analysis of variations in SCD rates and associated factors.

Conflict of interest: none declared.


This work was funded through an NHS South East Research and Development Directorate Health Services Research Training Fellowship awarded to D.C.


View Abstract