CHANCES : Follow-up of coronary heart disease, stroke and diabetes

Prepared by: Mark ODoherty and Maria Hughes, UKCRC Centre of Excellence for Public Health, Queens University Belfast, United Kingdom

1. Introduction

The majority of cohorts have/had access to a death register that fully covers the country or region where the study took place or have obtained death certificates for study participants who died during follow-up. In these cases ICD codes for causes of death are available. However, SHARE does not have this information as it relied on an ‘end of life interview’ which was conducted with a household member or a neighbour or another person with social contacts with the study participant.

The purpose of this section, therefore, is to give an overview of the issues that need considered in respect of the follow-up of coronary, stroke and diabetes events i.e. beyond what has already been considered for mortality follow-up (see Mortality).

The cohorts were asked to report on a range of issues (e.g. the outcome event definitions used) to allow an overview of the types and quality of the procedures and measurements employed by each cohort. Their responses to the questionnaire, and also information collected from published manuscripts, personal communication and cohort websites can be found in attached Excel tables for coronary endpoints and stroke endpoints.

For coronary and stroke events, information from all cohorts except EPIC Elderly Bilthoven has been obtained. The information below refers to the cohorts on which confirmed feedback has been obtained.

2. Coronary events

The majority of cohorts have follow-up available for coronary events. We have proposed two variables for coronary events. The first variable acts as a broad definition of fatal or nonfatal acute coronary events, and satisfies the criteria used by each cohort. It also includes unclassifiable coronary deaths which cannot be confirmed but no other cause is known. The second variable is similar to the first but does not include unclassifiable coronary deaths. Therefore, it may be more specific in capturing coronary events.

Cohorts have used various different methods for follow-up of coronary events within their cohorts, and also for the validation of such events. The documenting of coronary events was achieved through linkage to different registries, such as hospital discharge registers or mortality registries (such as those established for MONICA); through follow-up questionnaires or interviews (either from the participant themselves or sometimes from a close relative/friend/caregiver, if the participant was deceased or unable to complete the questionnaire themselves); or through contact with a physician. Once a coronary event had been identified, validation was mainly achieved (in eight cohorts) through the retrieval of medical records, hospital discharge diagnoses, death certificates and/or autopsy reports (necropsy only reported by a few cohorts); or through contact with and a supplementary questionnaire to the treating physician or general practitioner (ESTHER). Validation of events was achieved through the examination of clinical symptoms and signs, findings in ECG reports, and through analysing values of cardiac biomarkers and enzymes, usually by physicians or experienced epidemiologists. In the event of a fatal coronary event, death certificates and/or autopsy reports were also scrutinized to ensure coronary disease was the underlying and most plausible cause of death. SENECA and SHARE are the only cohorts which did not validate self-reported nonfatal coronary events, and fatal events were ascertained by either a questionnaire sent out to a close relative if the municipality could not provide a death certificate for the participant (SENECA), or by an ‘end of life interview’ which was conducted with a household member or a neighbour or another person with social contacts with the deceased participant (SHARE).

Cohorts have used various definitions for the coronary events reported within their cohorts, such as: those recommended by WHO MONICA project (EPIC Elderly Greece, EPIC Elderly Umeå, MORGAM (modified WHO MONICA), and Tromsø (modified WHO MONICA)); the 1971 WHO criteria (NHS and Zutphen (with updates)); and the 2003 AHA criteria (EPIC Elderly Denmark). ESTHER verified self-reported nonfatal coronary events through questionnaires sent to physicians and coded these as definite MIs and possible MIs. However, these definitions are based solely on whether the physician responded, and definite MIs relied on the physician’s answer to the question ‘did the participant have a myocardial infarction in the last X years (Yes/No and Date).’ RES has outlined the procedures they used to define coronary events reported within their cohort, but have not yet stated if a particular standardised procedure was used, though they reference the use of medical records, ECG reports and laboratory results. Overall, the diagnoses of coronary events are quite heterogeneous amongst the participating cohorts, and care will need to be taken when deciding upon which cohorts to include in future analyses. For example, the quality of the data available on coronary events within SENECA or SHARE is not as robust as those in Tromsø or MORGAM.

NIH_AARP is a large cohort with follow up for nonfatal coronary events which could be useful if a self reported response to questionnaire ‘if the participant had ever been diagnosed with a heart attack, angina, or coronary disease, heart-rhythm disturbance and coronary artery bypass or angioplasty surgery’ is acceptable; the primary endpoint of this cohort is cancer. Additionally, HAPIEE has attempted ongoing confirmation and validation of nonfatal coronary events (e.g. in the Czech Republic and Poland - all reported incident events were verified and validated through discharge reports and hospital medical records; in Russia and Lithuania - all incident events were verified and validated through MI, stroke and mortality registers established during the MONICA study in 1980s and 1990s, with data sources including hospital medical records and death certificates). However, the availability of different data varies considerably between countries. Additionally, Zutphen has information on fatal coronary events only for use within CHANCES, and these fatal events were defined using the 1971 edition of the WHO ICD codes and were verified by using hospital records, death certificates and information from the general practitioner.

Of the participating cohorts who provided information, the percentage of those participants lost to follow-up was quite low e.g. <10% overall loss to follow-up, with <1% loss to follow-up for deaths in NHS and almost 100% follow-up in EPIC Elderly Denmark. In most cases, missingness was related to a change of contact details or emigration from study area/country, a denial of access to appropriate registers and only in a minority of cases was it specifically due to non-response/refusal to follow-up questionnaires.

The follow-up of nonfatal coronary events other than MI, is more difficult within the participating cohorts, and follow-up of such events as angina pectoris or cardiac revascularization (e.g. Percutaneous Transluminar Coronary Artery (PTCA) and Coronary Artery Bypass Graft (CABG)) depends on the procedures utilized to define these coronary events. EPIC Elderly Greece, EPIC Elderly Denmark, MORGAM, and Tromsø have identified incident nonfatal angina pectoris using their own definition procedures, whilst ESTHER and NHS have self-reported angina pectoris only. EPIC Elderly Umeå, SENECA and SHARE did not record incident angina pectoris within their cohorts. RES also has pharmacy records available which serve 99% of the cohort from January 1991. Therefore, a pragmatic definition of two or more nitrate prescriptions to indicate angina may be possible. A similar pattern was evident in regards to cardiac revascularizations, with EPIC Elderly Greece, MORGAM, RES, and Tromsø having identified incident cases as part of their follow-up, and NHS having self-reported cases only.

The proposed harmonized variables for coronary event endpoints can be found in page Coronary end-point variables.

3. Stroke events

The majority of cohorts have follow-up available for stroke events. Three variables for stroke events have been proposed. The first variable acts as a broad definition of fatal or nonfatal acute stroke events, and incorporates all acute strokes from the cohorts which fulfill the typical symptoms of a definite stroke (see page STR1 - Stroke event type 1) and also unclassifiable strokes where the data concerning origin or duration of symptoms is insufficient to define definite stroke. Strokes validated by the systematic review of diagnostic data or by routine clinical or death certificate diagnosis are also included in this variable definition. The other two stroke variables are more specific: likely cerebral infarction and likely haemorrhagic stroke. Ideally, the diagnosis of cerebral infarctions or haemorrhagic strokes is based on specific diagnostic procedures, such as computer tomography, magnetic resonance imaging, or on necropsy findings. However, when these are unavailable or their availability is not known, the clinical or death certificate diagnoses can also be used within these variables, but the ICD codes for type unspecified are not permissible

Cohorts have used various different methods in the follow-up of stroke events and also in the validation of such events. The documenting of stroke events was achieved through linkage to different registries, such as hospital discharge registers or mortality registries; through follow-up questionnaires or interviews (either from the participant themselves or sometimes from a close relative/friend/caregiver, if the participant was deceased or unable to complete the questionnaire themselves); or through contacting with a physician. Once a stroke event had been identified, validation of events was mainly achieved through the retrieval of medical records, hospital discharge diagnoses, death certificates and/or autopsy reports; or through contact/supplementary questionnaire with the treating physician or general practitioner. The cohorts have used further detailed information from specific diagnostic tests, such as, computerized tomography, magnetic resonance imaging, lumbar puncture, angiography, surgery, or autopsy to classify the stroke subtypes (e.g. ischaemic stroke (cerebral infarction) or haemorrhagic stroke), and its hoped that CHANCES derived variables will capture this level of detail. SENECA and SHARE are the only cohorts which did not validate self-reported nonfatal stroke events, and fatal events were reported by either a non-responder questionnaire sent out to a close relative if the municipality could not provide a death certificate (SENECA), or by an ‘end of life interview’ which was conducted with a household member or a neighbour or another person with social contacts with the deceased participant (SHARE). SHARE also does not have subtypes of stroke available, and SENECA is only capable of coding fatal stroke events using the ICD codes from death certificates, when available (similar in Zutphen). ESTHER verified self-reported nonfatal stroke events through questionnaires sent to physicians. However, to date, this cohort has used stroke events without sub classifications in its own analyses, but it may be possible to derive the following stroke subtypes for 83.8% of the cohort: TIA, Prolonged Reversible Ischemic Neurologic Deficit (PRIND), complete infarction, and unknown.

Overall, the diagnosis of the stroke events and the severity of nonfatal stroke events are quite heterogeneous amongst the participating cohorts. Therefore, great care will need to be taken when deciding upon which cohorts to include in future CHANCES analyses. For example, the quality of the data available on stroke events within SENECA or SHARE is not as robust as those in Tromsø or MORGAM.

The few exceptions to this would include the NIH_AARP cohort, which does not technically have follow-up available for nonfatal stroke events as this cohort was concerned primarily with cancer. However, the NIH_AARP follow-up questionnaire contained questions asking if the participant had ever had a stroke or a TIA. However, this was self-reported and not subject to validation. Additionally, HAPIEE has confirmation of nonfatal stroke events, through hospital discharge diagnoses for all reported incident strokes, and has collected all available data to validate these events. However, the availability of different data varies considerably between the HAPIEE countries, making it hard to validate events consistently. Therefore, HAPIEE is currently not in a position to provide validated nonfatal stroke events. Additionally, Zutphen has information on fatal stroke events only for use within CHANCES, and these fatal events were defined using the 1971 edition of the WHO ICD codes and were verified by using hospital records, death certificates and information from the general practitioner.

The proposed harmonized variables for stroke event endpoints can be found in page Stroke end-point variables.

4. Type II diabetes mellitus

Type II diabetes mellitus is a chronic condition which can accelerate the risk of developing cardiovascular disease and a range of other chronic diseases. CHANCES cohorts tend to have low numbers of fatal diabetes cases and the majority of cases are nonfatal. Defining incident diabetes in the CHANCES cohorts is a challenge because it is not the primary endpoint for many of the cohorts, and biomarker measurements are not always available to provide unequivocal diagnoses of diabetes, therefore a robust method of diagnosing diabetes at baseline is often required.

9 cohorts have followed up diabetes events, including ESTHER, EPIC Elderly Umeå, RES, EPIC Elderly Greece, Tromsø, Zutphen, NHS, EPIC Elderly Denmark and SENECA. HAPIEE is collecting incident self-reported diabetes cases but without validation at present. SHARE collected self-reported diabetes but has not validated events and cannot distinguish between type I from type II diabetes. ESTHER, EPIC Elderly Umeå and RES have systematically searched for and validated diabetes cases in their cohorts. EPIC Elderly Greece and Tromsø have collected some incident diabetes cases during follow up for other endpoints and robustly validated these with medical records. Zutphen has reliably diagnosed prevalent diabetes at baseline using medical reports and non-fasting glucose measures, and incident diabetes cases by the same method using the WHO definition. NHS, SENECA and EPIC Elderly Denmark have used self-reported prevalent diabetes at baseline and self-report for incident diabetes during follow-up, augmented by self-reported current diabetes treatment. NHS has collected additional self-reported information to support diabetes definitions.

Self-reported incident diabetes is available in NHS, SENECA, EPIC Elderly Denmark, EPIC Elderly Greece and Tromsø. Those self-reporting incident diabetes in NHS in the initial follow up questionnaire provided additional details on diagnosis, laboratory results (e.g. fasting glucose measures) and treatment for diabetes with a 90% response rate. A random sample of 62 cases had their medical records checked and had 98% agreement with self-reports. SENECA and EPIC Elderly Denmark identified incident diabetes cases during follow up through self-report and current diabetes treatment but are not validated further (UPDATE: EPIC Elderly Denmark may be able to provide validated incident events through record linkage). SENECA robustly identified prevalent diabetes at baseline (self-report subsequently validated), which strengthens their self-reported incident data. This contrasts with EPIC Elderly Denmark which used self-report at baseline and follow up. Zutphen has not studied incident diabetes in their cohort and thus are unsure of the validity of incident diabetes cases in their cohort. They have reliably diagnosed prevalent diabetes at baseline using medical reports, current treatment and non-fasting glucose measures, and incident diabetes is defined only by self-report and current treatment and non-fasting glucose (WHO). Fasting glucose and OGTT were measured in 1990, possibly identifying incident diabetes cases at five year follow-up but other relevant measures are not available. EPIC Elderly Greece and Tromsø also identified incident diabetes cases during follow up through self-report and current treatment, but have not widely validated these cases. EPIC Elderly Greece has a total of 1050 self-reported diabetes cases. However, 450 of these self-reported cases have been validated during follow up through hospital records when diabetes co-occurred with CVD, with a mean follow up 6 years. Tromsø T4 cohort has identified 538 incident diabetes cases validated using medical records, with a mean follow up 9.5 years, and some of these involved the measurement of non-fasting glucose measures above 11.1 mmol/l, while other cases co-occurred with CVD. These validated cases could be included with the validated cases from ESTHER, EPIC Elderly Umeå and RES.

In general the year of diagnosis is available from all cohorts: ESTHER can provide month and year, EPIC Elderly Greece, EPIC Elderly Umeå and RES can provide date of onset in most cases (EPIC dates through self-report). Complications associated with diabetes have been collected in EPIC Elderly Umeå and ESTHER, in particular peripheral neuropathy and diabetic nephropathy can be compared in these cohorts. These can be potentially supplemented with cases with complications from EPIC Elderly Greece.

ESTHER, RES and EPIC Elderly Umeå have documented prevalent diabetes at baseline and have used various ways of identifying incident diabetes in their cohorts. ESTHER base incident diabetes in their physician validation survey (validation from medical records) at 8-years follow up. This diagnosis can be supported by fasting glucose (81.3%) and HbA1c measurements (98.3%) which allow participants to be classed as having diabetes, impaired glucose tolerance or impaired fasting glucose based on new ADA criteria. This work is ongoing, but the number of incident diabetes cases is approximately 388 with a mean 4.5 years follow up. RES has identified incident diabetes cases at 4, 6, 9, 15 and 20 years of follow up through linkage to general practitioner and pharmacy registers. Fasting glucose measures and OGTT based on ADA criteria were taken at each recontact (although baseline glucose was nonfasting) to support incident diabetes definition. The number of incident diabetes cases is approximately 754 with a mean of 8 years follow up. EPIC Elderly Umeå has followed up incident diabetes from 1992 (BL) to 2007 by checking the cohort through linkage to the diabetes register (DiabNorth). Inclusion on the registry requires fasting glucose and OGTT tests according to ADA criteria (which may be available to CHANCES) with medication use and medical record confirmation of diabetes. Biomarker measurements can also define diabetes, impaired glucose tolerance and impaired fasting glucose based on ADA criteria and these have identified 793 incident diabetes cases in EPIC Umeå and 283 cases in EPIC Elderly Umeå.

Two variables for diabetes have been proposed. One variable is related to documented Type II diabetes mellitus (T2DM) during follow-up, based on either fasting glucose, OGTT, or HbA1c results; a documented diagnosis by a medical doctor; or a clinical or death certificate diagnosis indicating T2DM. Documented current treatment of T2DM (e.g. insulin or anti-diabetic drugs) is also incorporated within this variable. The second variable includes documented or self-reported diabetes. The reason for deriving two variables is that documented T2DM can be analysed separately from documented or self-reported T2DM. Use of these variables requires the availability of variables DIAB_HIST1 or DIAB_HIST3 to exclude participants with diabetes (diagnosed or self reported) at baseline. Note that some cohorts (e.g. EPIC Elderly Greece) may have different follow up periods within the cohort.

CHD endpoints.xls (application/vnd.ms-excel)

stroke endpoints.xls (application/vnd.ms-excel)

CHD endpoints.xls (application/vnd.ms-excel)

stroke endpoints.xls (application/vnd.ms-excel)

CHD endpoints.xls (application/vnd.ms-excel)

stroke endpoints.xls (application/vnd.ms-excel)

CHD endpoints.xls (application/vnd.ms-excel)

CHD endpoints.xls (application/octet-stream)

Diabetes endpoints.xlsx (application/vnd.openxmlformats-officedocument.spreadsheetml.sheet)

1. Introduction

2. Coronary events

3. Stroke events

4. Type II diabetes mellitus

Attachments: