Example Of Exam Question With Answers - Copy.docx

A group of researchers conducted a study to examine the association between vegetable and fruit consumption and lung cancer in Greece. They identified a group of 329 patients (298 men) newly diagnosed with lung cancer who were admitted to a major teaching hospital during a 16 month period. They also selected a group of 570 patients (262 men) admitted to the same hospital for minor conditions believed to be unrelated to nutrition. All patients were asked about their usual intake of 20 different types of vegetables and fruits during the year preceding their hospitalisation. Respondents reporting usual intakes of at least 400 grams of vegetables and fruits per day were categorised as having a high intake of vegetables and fruits. Other respondents were categorised as having a low intake. A total of 104 patients with lung cancer and 210 patients with other conditions had a high intake of vegetables and fruits. a)

What type of study is this?

(10 marks)

Hospital based case-control study. (10 marks, take off 2 marks if doesn’t mention the casecontrol study is ‘hospital based’) b)

Calculate an appropriate measure of effect to assess whether there was an association between vegetable and fruit intake and lung cancer. How do you interpret your result? (20 marks)

Give 5 marks for giving correct table. Give 5 marks choosing the correct measure of effect. Give 5 marks for the calculation of the odds ratio. Odds ratio = odds of exposure in cases/odds of exposure in the controls Vegetable and fruit intake Lung cancer Yes No High 104 210 Low 225 360 329 570 Using the 'high intake group' as the exposed group: = (104/225)/(210/360) = ad/bc = 0.79 The students could also have calculated the odds ratio using the 'low intake group' as the exposed group: = (225/104)/(360/210) = 1.26 Give 5 marks for the interpretation of the result: -

The results suggest that a high intake of vegetables and fruits is associated with a reduction in the risk of lung cancer in the group examined

OR The results suggest that a low intake of vegetables and fruits is associated with an increased risk of lung cancer in the group examined c)

List the potential sources of bias that may occur in such studies, and discuss how each might apply to this study. (30 marks)

Give 15 marks for describing information bias: 5 marks for correctly classifying reporting bias and observer bias as information bias. 5 marks for explaining how reporting bias would affect the results in the study. 5 marks for explaining how observer bias would affect the results in the study. Reporting bias  Possible recall bias if lung cancer cases report their dietary intake differently than the controls. Compared with patients with minor conditions believed to be unrelated to diet, lung cancer cases might be more likely to have wondered about their diet and about how it might have influenced their condition (although the association between diet and lung cancer is not known so well in the public - compared with the smoking/lung cancer association). However, we do not know what the controls were hospitalised for. Observer bias  Possible observer bias if the interviewer is collecting data on dietary intake differently for the cases and controls (interviewer bias). It will be hard for the interviewers to be blind to the case/control status. If the interviewers know the hypothesis, they might be more inquisitive as to fruit and vegetable intake in the cases or in the controls. A standardised questionnaire should be used. Give 15 marks for describing selection bias: 7.5 marks for describing bias in the selection of cases 7.5 marks for describing bias in the selection of controls. Bias in the selection of the cases:  Cases were selected from only one hospital (major teaching hospital) and may not be representative of all cases of lung cancer in Greece. NOTE TO MARKER: Some people would not classify this as a selection bias but as an issue of representativeness. We teach them the latter. Do not mark down if gives the former however.  We do not have information on how the cases were selected (all cases seen during the study period?response rate?).  However, we cannot assume that cases who have a low intake of vegetables and fruits are more likely to be selected for the study than cases with a high intake. Bias in the selection of the controls:  Controls may not be representative of the source population (vegetable and fruit intake has been linked with various diseases including cardiovascular diseases - thus, people who are sick/hospitalised may be more likely to have a low intake of vegetables and fruits).  Controls may not be representative of the population that gave rise to the cases. This could occur because the ‘catchment population’ may be different for different diseases and the cases may not be comparable to the controls.  No information given on response rate in the eligible controls. This is a key potential source of bias. Note: Ok to give marks reserved for next question if give other reasons for findings eg temporality (2.5 marks) and reverse causation (2.5 marks) here d)

What other non-causal reasons might there be for these results?

(20 marks)

Give 10 marks for describing issues around confounding.  Give 5 marks if explain what confounding is (2.5 if only states term)  Give 5 marks if suggest potential confounders for this study - they should at least mention age, sex, socioeconomic status and smoking => smokers are known to have a lower vegetable and fruit intake and it is a risk factor for cancer Give 10 marks for describing issues around chance. (2.5 if only states chance, more marks if goes onto mention need for sample size estimates and CI)

e)

If there is a true association, do you think that it is likely to be causal? Explain your answer. (20 marks)

The Bradford Hill Criteria (give 5 marks for discussing temporality, then 2.5 marks each for discussing other criteria up to max of 15). Note: if only gives the other criteria and no explanation give one mark for each criteria given. 1. Temporality: in a case-control study data on exposure are collected retrospectively so the timing of exposure may not be good. There is also the potential for reverse causation 2. Strength: the association is not very strong (OR=1.26) 3. Consistency: information from several studies would be needed - particularly using a better design (cohort studies) 4. Biological gradient: we cannot assess this here (this could be assessed if fruit and vegetable intake was quantified). 5. Specificity: low fruit and vegetable intake not specific to lung cancer (related to several diseases) but this does not meant that the association is not causal 6. Plausibility: some biological mechanisms have been suggested 7. Coherence 8. Experiment: has been done with supplements or education to increase fruit and vegetable intake. 9. Analogy