The hypothesis is that heavy alcohol consumption carries an increased risk of HS. The respective null hypothesis would be that heavy alcohol consumption does not carry an increased risk of HS. b) Timeframe, disease incidence, latency period, urgency of study, and cost would be important in sselection of study design. Making no assumptions for cost or resource limitation then the most appropriate study would be a prospective cohort. HS has low incidence and high mortality . Hence data for exposure would be better obtained here than reliance on retrospective data as in case control studies. Elements of the risk factors in terms of their natural history could also be evaluated.
Although the incidence of the HS is low if a large enough cohort could be engaged and followed up then enough cases could occur for an evaluation of incidence rates. This would be a large undertaking and government support would be welcome. The study could also be used to look at other diseases and risk factors if planned appropriately.
Also in this cohort study the ability to use the cases in a further nested case-control study could be useful.
If low cost and time frame were critical issues then a case control study may be practical.
c) The cohort would be selected from the general population and would be patients registered with a large number of general medical practices in the United Kingdom (UK) who presented for consultation during a specific time period (3 months of the year). A large number is required as the incidence of HS is low. In the UK there are approximately 110,000 occurences of first stroke annually . Of these an estimated 13% will be HS. This in a UK population approaching 62 million . Several geographical areas would be included to try have general representation for features such as social class, sex and age. This cohort would then be followed for a time period (5 years).
participants are excluded if they:
were not between 18 and 75 years of age (this is because HS stroke is unlikely in children and alcohol consumption is influenced by prohibition laws under this age. Also HS is associated with a higher incidence after 75 years old )
have already had a first ever HS.
This cohort would then be followed up over the time period and cases would be identified by computer GP records and hospital episodic statistics. If GP records are unclear then patient hospital records are examined.
d) The exposure would be classified as consumption of beverages containing alcohol and would be measured by questionnaire at consultation. Participants would be asked to give information on the amount of alcoholic beverages consumed per day. The questionnaire would request alcohol consumption amounts in categories of wine (number of glasses), spirits (the number of standard measures) and beer (the number of cans/ bottles). The number of units would then be calculated by trained project workers (could be trained medical practice staff) who would then categorize the result as level 1 (men less than or equal to 3 units a day, women less or equal to 2 units a day), level 2 (men over 3 and under 8 units a day, women over 2 and under 6 units a day) and level 3 (men more than 8 units a day, women more than 6 units a day). For purposes of interpretation heavy alcohol consumption in individuals would be regarded as levels 2 and 3. This is based on the UK recommended limits .
The outcome that is HS would need to have a strict definition. HS, also known as intracranial haemorrhage or intra cerebral haemorrhage (ICH) is defined as haemorrhage from one of the arteries that supply the brain. Frequently in studies and literature sub-arachnoid haemorrhage (SAH) is included under the umbrella of HS. The International Classification of Disease tenth revision has a code that includes SAH in the definition (I60-I62). For purposes of this study we would include SAH. Diagnostic conditions of HS for this study would then be: Symptoms of commencement of severe headache with variations in the level of consciousness that are related with haemorrhage within the subarachnoid space or brain parenchyma and identified on Magnetic resonance imaging (MRI) or computerised tomography (CT ) scans or on autopsy.
This outcome would be further defined and measured with the following criteria;
Cases should fit the definition above and CT and MRI scans should have been done.
Cases must have been diagnosed by a physician/Surgeon at hospital.
Cases would also undergo medical record analysis by a trained study physician to confirm diagnostic agreement and of first ever occurence.
If HS has resulted in mortality then cause of death must be recorded as such on death and hospital records.
Cases would be excluded if HS was result of brain trauma, neoplasm, surgical mishap, established aneurysm, cerebral amyloid angiopathy.
Repeated postal questionnaires to the participants and examination of medical/ hospital records would be done annually to measure exposure and outcome. Here consent would need to be established early on.
e) Non-respondent/ volunteer bias could be an issue in that non-participants may be different to those who consent in their relationship between outcome and exposure. Here project staff could be trained to better inform potential participants of its importance in healthcare.
Another selection bias issue is the samples coming from General Practice attendees. This cohort may not represent the general population in that they may have general medical problems that precipitate more frequent attendance at the practice. This bias could be reduced by widening the source of participants to places other than just medical practices eg community centres and ensuring no duplicity.
Bias could arise if the selected medical practices were unrepresentative of the country demography. If practices chosen to participate were all in inner city areas and or more socially deprived regions then this would exclude representation from rural and less deprived regions. Also factors such as ethnicity are important considerations so as to make the sample population representative. This bias could be minimised by ensuring a wide selection of practices from differing demographic regions.
Unequal efforts by investigators to collect follow-up data could bias results. This could be minimised by standardising protocols in data collection and also blinding of investigators with regards to previous exposure status of the individuals.
Information biases could also be a factor in the study.
If participants could not understand and answer the questionnaire correctly and may give incorrect answers for exposure status. Issues such as language difficulties could also affect this. This could be minimised by piloting the questionnaire to check validity. Also the staff involved in recording units of alcohol from the information given by the participants could be assessed and then high sensitivity/ specificity and uniformity reached.
Also as the exposure measurement is only being done annually then participants who have stopped consuming alcohol after an evaluation may still be recorded as per the last evaluation. This bias could be reduced by measuring exposure status more frequently.
Loss to follow-up bias could be a very significant issue that affects the study. This would be especially a concern if it affected the exposed and unexposed differently. Here efforts to persuade participants to remain in the study may help. The use of medical practices to encourage participants who have not returned the follow-up questionnaires could be helpful in reducing this bias.
Case ascertainment bias may appear if some cases of HS are diagnosed as something else eg haemorrhage secondary to an infarction, and also if true cases of HS are given a different diagnoses. Here tight definition and examination of records and scans may improve the sensitivity and specificity of correct recording.
f) Age could be a confounding factor if it is felt that the level of alcohol consumption varies among age groups. It could be that younger age groups have a higher alcoholic consumption. Also older age groups are more prone to haemorrhagic stroke .
Smoking would be a confounding issue in that individuals that consume or alcohol may be more likely to smoke more as well. Smoking may also lead to increase risk of HS .
Hypertension could also be confounding for the study. For people that consume greater quantities of alcohol may be more prone to having high blood pressure. Hypertension is also associated with increased risk for HS .
Low Serum cholesterol has been associated with increased risk oh HS in women. Serum cholesterol could also be higher in people with heavy alcohol consumption due to lifestyle .
Anticoagulation medication could confound. Anticoagulant therapy may be associated with lower levels of alcohol consumption and also is a risk factor for HS.
Gender may confound in that males may drink more alcohol. Males may also be more associated with ICH and females more with SAH .
Race could confound if alcohol consumption varies in certain ethnic groups. Black race is more associated with HS .
Cocaine and amphetamine drug use may be a problem in that drug users may consume more alcohol. They would also be more prone to HS .
To minimise effects of confounding of age, smoking, hypertension and serum cholesterol, restriction, matching and statistical techniques such as standardisation and stratified analysis could be applied.
To minimise effect of confounding of race, gender and drug use restriction of study population and matching would be options.
In general to control for confounding the confounders need to be recognised and measured, then matching, restriction, standardisation, stratified analysis and multiple regression techniques are usable methods.
g) Analysis on the results could be carried out by looking at and comparing incidence rates among the alcohol categories levels 1 to 3 and with level 1 (low consumption) and levels 2+3 (heavy consumption). The incidence rates could be time based (denominator person years) especially if the exposure varies considerably in individuals over the duration of the study.
From this risk ratio and attributable risk values would be calculated. These would be presented with 95% confidence intervals. Hypothesis testing would be at p value of 0.05.
With regards to age this could be categorised into 10 year age groups and relative risks could be amended.
These would be presented before and after modification (standardisation) for separate confounding factors.
The sex- specific and gender specific rates would also be presented.
h) The study strengths are:
Collecting data prospectively gives more complete information on exposures, periods and cases. Data collection would be substantive and reasonably strong. Calculation of incidence rates and relative risk/ attributable risk is applicable. Comparison is possible with studies of a similar design.
Its tracks participants for 5 years and is population based.
Temporal relation and dose â€"response between exposure and outcome can be established. Hence can have input into understanding of a causal association and burden of disease.
It could generate considerable data to examine other risk factors and diseases.
Population is sizeable and sufficiently indiscriminate to generalise findings.
Funds and consent allowing, the study could continue past the planned end date.
Option of performing a nested case control study from within the cohort could be very useful.
The weaknesses of this study are:
Non-participation at start, loss to follow-up and unrecorded diagnoses of HS in participants at hospitals other than local to medical practices would be a issue and may limit generalisability.
Failure to record possible non-hospital presentations of HS.
The incorporation of both SAH and ICH into the definition of HS may give findings that are not specific to each one.
It fails to take into account binge drinking and unless exposure recording is frequent will not give accurate data for variable consumptions.
Confidentiality and consent could be concerning factors when attempting numerous communications.
Restricted management of unrelated extraneous factors.
This is a major undertaking. The time frame and number of participants needed would be large in order to gather enough cases. Resources needed for multiple recordings over a large period would be sizeable. Problems associated with keeping trained staff for a considerable period. Therefore it would be time consuming and expensive.
