1MATH6033 Epidemiological Methods –Referral Assignment 2022
This assignment will constitute 100% of your overall mark for this course. Both questions
should be attempted, and each part will contribute towards 50% of the marks. The marks
available for each question are given as an approximate guide to the amount of work that is
required from you.
Page limitYour solution to Part 1 should not exceed four A4 sides, your solution to Part 2
should not exceed three A4 sides (including tables). Please start Part 2 on a new page.
These limits assume that a sensible font size has been used e.g. Times New Roman size 12,
or Arial 11, and that standard margins are used (i.e. 2.54cm around the text). Font size and
margins can be reduced if necessary for tables, but not otherwise. Marks will be deducted
for work exceeding the stated page limits. Please note that good solutions could well be
written up in fewer than the maximum allowable number of pages. Do not present extra
work as appendices. Any appendix submitted will not be read and marks will be deducted
for exceeding the page limits.
Academic integrity You are allowed to talk to your peers about the coursework, but you
need to write up the solutions independently, i.e. while you are on you own. In particular,
copying the solutions from a classmate or anywhere else would constitute a breach of
academic integrity.
Deadline Submit your completed assignment on Turnitin on Blackboard by 4pm on August
19th 2022.Work submitted up to 5 days after the deadline will have the final mark reduced by
the factors in the following table, as per the university regulations:
University Working Days late Mark
1 (final agreed mark) * 0.9
2 (final agreed mark) * 0.8
3 (final agreed mark) * 0.7
4 (final agreed mark) * 0.6
5 (final agreed mark) * 0.5
More than 5 Zero
If you have any queries concerning this assignment, you should e-mail Sarah Crozier
(src@mrc.soton.ac.uk).
Part 1Life course determinants of cardiovascular mortality and morbidity
Background Life course epidemiology is the study of the effects of exposures acting during
gestation, childhood, adolescence and adulthood on risk of morbidity (disease) and mortality
(death) in later adulthood. The study of the developmental origins of health and disease
involves researching how poor growth of the fetus in the womb leads to permanent changes
in the body’s function that cause increased rates of many different types of disease,
including cardiovascular disease, in subsequent adult life.
The Hertfordshire Cohort Study was conducted to investigate the relationship between fetal
growth (quantified by birth weight) and mortality in later adulthood. The following tables
present the standardised mortality ratios (SMRs) for ‘all cardiovascular disease’ (ICD9 codes
390-459) and ‘all other causes’ (ICD9 codes <390 or >459) by birth weight for 15,500 men and
women in the 1911-1930 Hertfordshire birth cohort. The SMRs were calculated for mortality
between 1951 and the end of 1992 using the person-years approach.
2Birth
weight (lb)
All cardiovascular disease All other causes
Observed deaths Expected deaths Observed deaths Expected deaths
5.5 65 68 67 74
-6.5 155 194 162 213
-7.5 353 441 383 485
-8.5 377 477 401 521
-9.5 144 236 190 257
>9.5 78 113 97 123
Your Task (Part 1a)Enter the relevant data into Stata (hint: use the Data Editor – you open
this by clicking the button on the top left hand side of the screen that looks like a
spreadsheet with a pencil over it).
1. Investigate whether for (a) all cardiovascular disease, and (b) all other causes of death,
there is evidence that the SMRs are homogenous across the birth weight groups,
follow a linear trend, or are all different. Briefly present the deviance statistics, residual
degrees of freedom, and deviance difference tests for the models that you fit.
Summarise your findings in a few sentences providing a clear interpretation of the
meaning of the models (14 marks)
2. Calculate SMRs and associated 95% confidence intervalsin each birthweight group for
all cardiovascular disease and for all other causes of death(4 marks)
Further background The Hertfordshire study was unable to account for other life course
factors which may influence risk of cardiovascular morbidity and mortality. Adjustment for
such factors might have altered the conclusions of the study if they acted as confounders of
the relationship between birth weight and cardiovascular disease. Accordingly, a further
study was carried out to explore the relationship between fetal growth (quantified by birth
weight) and high blood pressure; this study included information about adult risk factors.
Elevated blood pressure is one form of cardiovascular morbidity; over time, elevated blood
pressure leads to arterial and organ damage and increased risk of stroke, kidney problems
and heart failure. Identification and management of factors that influence blood pressure is
an important public health issue.
The participants for this study were 712 men and women born in Sheffield between 1923-
1930 and still living in the area in the mid-1990s. Data on the birth weight and gender of these
participants (410 men, 302 women) were abstracted from hospital records kept in the 1920s.
Participants were invited to a clinic to undergo measurement of blood pressure according to
a standard protocol and to obtain information about other risk factors for cardiovascular
disease including smoking and body mass index.
Data The Stata file sheffield.dta contains the variables listed below
idnum Study ID number
sex Sex (1=Male, 2=Female)
bwt Birthweight (oz)
smokstat Smoking status (0=Never, 1=Ex, 2=Current)
sbp Systolic blood pressure (mmHg)
bmi BMI (kg/m2)
Your Task (Part 1b)Answer the following questions to explore the risk factors for blood
pressure in this cohort of elderly men and women.
3. Examine the distributions of the outcome variable (systolic blood pressure) and the
primary exposure variable (birthweight) to determine whether any transformations of
3these data are required for use in subsequent linear regression analyses. Present
appropriate summary statistics of location and spread to describe the distributions in
these subjects. (5 marks)
4. Smoking, gender and body mass index are all potential risk factors for blood pressure.
Explore the univariate associations of each of these risk factors with birthweight. (3
marks)
5. Is there a difference in systolic blood pressure between men and women? Comment
on what impact you expect gender to have on the birth weight-blood pressure
relationship. (4 marks)
6. Use linear regression to describe the magnitude and direction of relationship between
systolic blood pressure and birth weight in this cohort. Present the results of
appropriate statistical analyses to determine if gender is a confounder, or if there is an
interaction with gender. (4 marks)
7. Describe the relationship between systolic blood pressure and smoking. Determine if
gender is a confounder, or if there is an interaction with gender. (6 marks)
8. Select a final multiple linear regression model that you think best summarises the
relationship between systolic blood pressure and birth weight in these subjects, taking
account of any other potential risk factor(s) you now think important. Briefly explain
your reasons for including/excluding gender, body mass index, and smoking status in
your chosen model. A non-statistical but numerate colleague needs to understand the
results; interpret the model in a way that he/she would understand. (10 marks)
(50 marks in total for Part 1)
Part 2Tea drinking habits and oesophageal cancer in a high risk area in northern Iran:
population based case-control study
Background A friend of yours in Southampton enjoys drinking tea, but has heard that tea
drinking may be associated with cancer. He has been searching on the internet and has found
a paper published in the BMJ reporting on a study looking at the association between tea
drinking and oesophageal cancer. Your friend knows that you are studying statistics and have
done a course in epidemiological methods and thinks that you might be able to help him
understand the paper. He wants you to explain the paper but also to interpret the findings and
help him decide whether he should change his tea-drinking habits.
A copy of the paper is attached.
Your task (Part 2)Read the paper and answer the questions below. Remember your
solution to Part 2 should not exceed three A4 typed sides.
Remember that this paper has been published in a major journal and will have undergone
extensive review. It may have some weaknesses, and alternative methods may well have
been possible or better, but it is unlikely to be seriously flawed. Also remember that your
non-statistical/epidemiological friend needs to have everything clearly explained in lay
language so avoid using technical terms unless you explain them; the clearer the
explanation, the greater the number of marks that will be awarded for each part.
1. There are two parts to the study in this paper. Explain the two study designs and how
they each contribute to answering the study question. What are the advantages and
4disadvantages of using the case-control study compared to others that are possible?
(9 marks)
2. How were the cases and controls selected in the case-control study? Describe what
matching means and why it is used. Were the methods of selecting cases and
controls appropriate? (8 marks)
3. Explain what an odds ratio is and how you interpret it using an example of one of the
crude odds ratios from Table 2. (5 marks)
4. Your friend does not understand what a confounding variable is. Explain, with the aid
of tobacco intake as an example, what a confounding variable is and how the
confounding variables have been accounted for in the analysis. (7 marks)
5. Summarise the message of the first two sections of Table 2 (those relating to tea
temperature and interval between tea being poured and drunk). (6 marks)
6. Describe the evidence for whether you think the study conclusions are correct.
Discuss whether the association between drinking hot drinks and oesophageal cancer
is likely to be causal. Explain whether you think the findings are relevant to your friend
and what (if any) changes he should consider to his tea-drinking habits. (15 marks)