HP3 Quiz 3 Practice

21

Multiple Regression Analysis

With regards multiple regression analysis, which of the following statements is/are correct?


  It allows us to determine the association between multiple independent variables and outcome

  Independent variables may be continuous or categorical

  Logistic regression is used when the outcome variable is binary (i.e. present/absent)*

  Linear regression# is used when the outcome variable is continuous

  All of the above


ANCOVA is used when the outcome variable has multiple categories but you don’t need to know this for exam. You are not expected to interpret results of linear regression analysis for the exam.

22

Table 2 - Multiple Logistic Regression analysis for CAD:



With regards Table 2, which of the following statements is/are correct?


  Compared with male sex, female sex is protective against CAD.

  Every one year increase in age is associated with a statistically significant increase in CAD risk.

  After controlling for the effect of sex and age, patients with CAD are 1.2 times more likely to have hypertension than controls (those without CAD).

  There is 95% confidence that the true adjusted odds ratio for hypertension is in the interval 1.1 to 1.3

  All of the above


A is correct, because the OR is <1 i.e. protective, and its 95% CI doesn’t include 1 B is correct because the OR is related to a one unit (e.g. one year increase in age) or one category change in the independent variable C is a correct interpretation of an adjusted OR derived from logistic regression - you did this in the assignment D is correct as an interpretation of the 95% CI around an OR - we revised this in the first revision lecture and it was also in assignment.

23

With regards Table 2, what is the p value of the adjusted odds ratio for female sex likely to be and why?


  P value likley to be 1 exactly, as the OR is adjusted for other covariates

  0.05 < p < 1.0 because the 95% CI around the adjusted OR does not include the no-effect value of zero.

  P likely < 0.05 because the 95% CI around the adjusted OR does not include the no-effect value of zero.

  P likely < 0.05 because the 95% CI around the adjusted OR does not include the no-effect value of one.

  None of the above


95% CI and p value are related. If the 95% CI around an odds ratio does not contain the no-effect value of one, then the p value will also be statistically significant (i.e. <0.05)

24

Clinical Trial Profile and Loss to Follow-up

Questions 24 to 26 inclusive are related to the following information and Figure 7.

The trial profile in Figure 7 comes form a randomised controlled trial of compression stockings vs no stockings for prevention of symptomless DVT in long-haul flights. (Scurr et al. The Lancet 2001;357:1485)

Figure 7 - Trial profile:



With regards the Trial Profile (Figure 7), which of the following statements is/are correct?


  The number of patients lost to follow-up was almost the same in both groups.

  The % loss to follow-up in each group is within the expected range for a randomised controlled trial.

  Loss to follow-up will not pose a threat to the validity of the findings of this study.

  a and b

  b and c


A is correct because the number of losses was not differential between treatment arms (16/116 control and 15/115 stocking group) B is correct, because % loss in control group is 16/116=14%, % loss in stocking group is 15/115=13% and both %s are in expected range of 10-15% C is incorrect because the ‘fate’ of those lost to follow-up may have differed in the two groups.

25

There were 12 DVTs in the control group and no DVTs in the stocking group. In a worst case scenario analysis performed in order to quantify the potential impact of loss to follow-up on the findings of the study, what would be the incidence rate of DVT in each group?


  DVT incidence rate 12/116 in no-stocking and 0/115 in stocking group.

  DVT incidence rate 12/116 in no-stocking and 15/115 in stocking group.

  DVT incidence rate 28/116 in no-stocking and 0/115 in stocking group.

  DVT incidence rate 28/116 in no-stocking and 15/115 in stocking group.

  None of the above


In worst case scenario analysis assume all those lost to follow-up in *intervention* arm experienced an event, while assuming none of those lost to follow-up in control group had event. Note: c is *best* case scenario in favour of showing a benefit with intervention. Here assume all those lost in control group had event and none lost in intervention group had event.

26

Regarding an intent-to-treat (ITT) analysis, which of the following statements is/are correct?


  It is done to reduce selection bias due to loss to follow-up / noncompletion

  In this case, the analysis would include 116 patients in the control group and 115 patients in the stocking group

  In this case, the analysis would include 100 patients in the control group and 100 patients in the stocking group

  a and b

  a and c


A is correct B is correct because in ITT analysis patients are retained in the groups to which they were randomised regardless of whether they complete the study; c would be a ‘per protocol’ analysis of only those that completed the study.

27

Survival Analysis

The information below and Figure 8 relate to questions 27 to 30.

A randomised controlled trial was undertaken to compare the efficacy and safety of open versus endovascular repair of abdominal aortic aneurysm. Kaplan Meier (K-M) survival analysis on an intention-to-treat basis was undertaken to compare outcomes in the two groups. For the outcome ‘survival free of severe events’ K-M survival curves for the two groups are presented in Figure 8.

Figure 8:



At the commencement of the study, how many patients were randomly assigned to each of the open and endovascular repair groups?


  178 assigned to open and 173 assigned to endovascular repair.

  150 assigned to open and 155 assigned to endovascular repair.

  146 assigned to open and 151 assigned to endovascular repair.

  121 assigned to open and 125 assigned to endovascular repair.

  89 assigned to open and 91 assigned to endovascular repair.


This is the ‘number at risk’ at time=0

28

After 12 months of follow-up, what was the cumulative probability of survival in the open repair group?


  Approximately 90%

  Approximately 85%

  Approximately 100%

  Approximately 95%

  None of the above


Draw vertical line from 12 month mark up to ‘open’ curve, then trace line horizontally until you meet the vertical axis.

29

At each time point, what constitutes the ‘number at risk’ in each group?


  Those still alive

  Those still alive and free of severe events

  Those still alive, free of severe events and still in study

  Those still alive, free of severe events or lost to follow-up

  None of the above


At each time point, the ‘number at risk’ are those still alive, free of event, excluding those who have withdrawn or been lost to follow-up.

30

Which of the following statements is/are correct?


  The survival curves start to diverge early in follow-up, around the 2-3 month mark

  Based on p=0.39, there is no statistically significant difference in the survival of patients, free of severe events, over the two years of follow-up

  This study is susceptible to sampling selection bias as the patients in this study may not be representative of all patients undergoing AAA repair

  This study is susceptible to bias through loss to follow-up

  All of the above