Reference no: EM132389433

Design of Experiments & Randomised Clinical Trials [DES] Assignment -

Please answer your questions concisely.

Question 1 (Parts (a) through (d)) -

Data from various studies indicate that men who exercise are less likely to sustain a bone fracture compared to men who don't exercise. For a number of reasons, it is difficult to use bone fracture as an endpoint and we need to find an appropriate surrogate measure. In his paper "Skeletal Effects of Exercise in Men", Karlsson mentions the use of surrogate endpoints for fractures, including bone mass.

[Karlsson, M.K. Skeletal Effects of Exercise in Men. Calcif Tissue Int, 2001; 69:196-199..]

(a) Identify three important considerations that an investigator must make when deciding upon an appropriate surrogate outcome (in any study, not just a study on bone fractures).

(b) Discuss the extent to which these considerations have been addressed in the Karlsson paper in relation to the surrogate endpoint of bone mass.

(c) Explain briefly how a statistician could attempt to validate bone mass as a surrogate for use in future trials if a prospective randomised study with fracture as an endpoint and bone mass as a prognostic factor existed.

(d) Discuss issues relating to the extrapolation of treatment benefit seen in the surrogate to that for the clinical endpoint (in any study, not just a study on bone fractures).

Question 2 (Parts (a) through (e) with sub-parts) -

This question refers to the following articles.

Chang AB, Torzillo PJ, Boyce NC et al. Zinc and vitamin A supplementation in indigenous Australian children hospitalised with lower respiratory tract infection: a randomised controlled trial. Med J Aust, 2006; 184(3): 107-112

and

Diamond TH et al. Clinical outcomes after acute osteoporotic vertebral fractures: a 2-year non-randomised trial comparing percutaneous vertebroplasty with conservative therapy. Med J Aust, 2006; 184(3): 113-117.

(a) Consider Table 3 from Diamond et al's paper on vertebral fractures.

(i) Perform a Bonferroni adjustment on the five (5) P values shown under "Physical functioning" (that is, the bottom panel of the table) for the absolute changes at 0 hours, 24 hours, 6 weeks, 6-12 months and 12 months. (Do not concern yourself with the test for the percentage changes). Use a family-wise error rate of 0.05.

(ii) Perform a Simes adjustment on the same P values specified in part (i) above.

(iii) Comment on the results you obtained in parts (i) and (ii).

(b) Consider Diamond et al's Table 2 (baseline clinical demographics etc) and Chang et al's Table 2 (characteristics of infants with ALRI at trial entry).

Given the fundamental difference in study designs, comment on the appropriateness of the information presented.

(c) In Chang et al's paper, in the statistical analysis section, the authors write: "Post hoc power calculations of our sample size with 107 per group was 95% (5% significance) to detect an increase readmission rate as found in the study;". Comment.

(d) Back to Diamond's paper on vertebral fractures. At the end of the statistical analysis section the authors state: "An intention to treat analysis was used, with the last value [observation] carried forward for those individuals who died or who were lost to follow up". Comment.

(e) Consider again Table 2 of Diamond et al. Using Stata's immediate commands, appropriate R procedures, or otherwise (maybe even manually!), consistently reproduce the P values shown in the table. By "consistently" I mean, say, for at least 3 tests - take them at random - you will calculate the same P value (to 2 decimal places) as did the authors. You need not bother will all 14 tests. You may well try tests with and without continuity correction and/or with and without equal variance specification. Comment.

Question 3 (Parts (a) through (h) with sub-parts) -

Seretide is a combination asthma inhaler which contains two different therapeutics; an inhaled corticosteroid (ICS) and a long-acting beta2 agonist (LABA). These two classes of medication act together to achieve both an anti-inflammatory and a bronchodilator action in the management of asthma. The alternative to combination therapy is to use two separate inhalers concurrently; one containing an ICS and one containing a LABA. Evidence from previous studies suggests that the use of combination therapy may lead to better health outcomes and may also offer other advantages, such as convenience for the patient and increased adherence to treatment regimens.

A randomised controlled trial will be conducted to assess whether the administration of a single device containing the combined products achieves better health outcomes and therefore better control of asthma, when compared with concurrent administration of the two individual products. A parallel groups design will be used. Asthma patients will be recruited from general practices in South Australia over a 2-year period. The randomisation will be performed using blocks of varying size (2, 4 and 6) to ensure that approximately equal numbers of patients are assigned to each treatment group.

The randomisation will be stratified by practice and severity of asthma (severe and non-severe). Patients will not be blinded to treatment; however, outcomes will be measured by a blinded assessor. The primary outcome is the morning peak expiratory flow (PEF) at 6 months (a measure of lung function).

(a) Suggest three baseline covariates that you would consider adjusting for in the analysis of the primary outcome and give a brief explanation as to why you think each covariate might be important in this context.

(b) Briefly discuss the advantages and disadvantages of adjusting for baseline covariates in the analysis of the primary outcome in this study.

(c) Briefly specify the reasonable alternatives to adjusting for baseline covariates in the analysis (in any study).

(d) Assuming the primary outcome is normally distributed, calculate the sample size required per group to detect a difference in mean scores of 0.2 standard deviations between the treatment groups, using alpha=0.05 and 90% power. (At this stage, do not allow for non-compliance, crossovers, or loss to follow up).

(e) If we now wish to allow for non-compliance with treatment, assuming 10% drop-out and 5% drop-in, calculate the required sample size per group.

(f) Suppose the trial investigator wants to analyse the primary outcome after the trial has been running for 1 year, 18 months and 2 years (as well as at the end of the trial), using alpha=0.05 each time, and will stop the trial early if a significant result is obtained on any one of these occasions. State the overall alpha that would apply to the primary outcome in this case. How could you control the overall alpha to be 0.05?

(g) Three months into the recruitment phase, it is discovered that different practices in the study have been using different definitions of severity of asthma to determine which stratum each patient belongs to for the purpose of randomisation.

(i) Discuss the possible implications of the use of different definitions of severity of asthma.

(ii) Suggest an approach to avoid this problem for the remainder of the trial.

(h) What do you believe will be the main sources of bias remaining in the trial after randomisation?

Question 4 (Parts (a) through (c) with sub-parts) -

The Second Australian National Blood Pressure Study (ANBP2) was a prospective, randomised, open-labelled, blinded to endpoint (PROBE) trial of Angiotensin Converting Enzyme (ACE) inhibitors versus diuretic in the treatment of mild to moderate hypertension in the elderly - defined as 65 to 84 years of age at last birthday. It was carried out in over 1500 general practices in Australia in the mid-late 1990s. The primary (combined) endpoint was the incidence of all defined cardiovascular events and deaths from any cause. There were several secondary outcomes, including incidence of first cardiovascular events and various measures of blood pressure control. The main entry criterion (there were, of course, many criteria) was a systolic BP ≥160mmHg, or a diastolic BP of ≥ 90mmHg if systolic BP was >140mmHg, prior to randomisation. It was intended that all subjects would be followed for 5 years.

(a) One of the sample size calculations was based on data from a preceding UK MRC trial wherein a yearly incidence of first cardiovascular endpoints (of the types also to be used in ANBP2) among those on the comparator class of drugs (diuretics) was quoted as 21 per 1000 subjects per year. The protocol for ANBP2 called for a two-tailed test with α = 0.05 and β = 0.10.

(i) If it were desired to detect a 25% reduction in the cumulative incidence of first events in the ACE-I group (relative to the diuretic group) at the end of the study, what would be the required sample size per group? (At this stage, do not allow for non-compliance, crossovers, or loss to follow up).

(ii) Further, now assume that 15% of subjects will have missing data on the first cardiovascular event endpoint. What is the required number of subjects needed to be randomised into each group?

(b) Over 53,000 subjects were screened for entry into ANBP2. The screening process detected many people who were unaware they had high blood pressure, but, of course, a large proportion of the subjects eligible for recruitment into ANBP2 were already diagnosed as having hypertension and were already being treated. Are there different ethical implications with respect to recruitment of the already treated group compared with newly discovered hypertensives? Give reasons for your answer.

(c) The ANBP2 protocol specified that the primary endpoint analysis would be adjusted for a limited number of pre-specified potential confounders, should these variables actually confound the results of this study. The ANBP2 statisticians, blind to (i) the assigned treatment (ii) the p-values with and without adjustment, and (iii) the direction of any change in the treatment effect between adjusted and unadjusted analyses, found that age and sex were the only potential confounders in the defined list that actually changed the treatment effect by an appreciable amount. Accordingly, the final result was adjusted only for age and sex. The p-value associated with the adjusted analysis for the primary outcome was exactly 0.05 (to at least 4 decimal places). The p-value without adjustment for age and sex was 0.08. Comment.

Question 5 Parts (a) through (e) -

Researchers assessed the effectiveness of a range of weight management programmes for weight loss. A randomised controlled trial study design, incorporating eight treatment arms, was used. Each intervention-Weight Watchers, Slimming World, Rosemary Conley, group-based dietetics led programme, general practice one to one counselling, pharmacy led one to one counselling, and a choice of any of the six programmes-lasted for 12 weeks. The control treatment consisted of 12 vouchers enabling free entrance to a local leisure (fitness) centre. Participants were 740 obese or overweight men and women identified from general practice records. The primary outcome was weight loss at the end of the programme (12 weeks). Secondary outcomes included weight loss at one year. Baseline characteristics were available for all participants, whereas follow-up data were available for 658 (88.9%) participants at the end of the programme and 522 (70.5%) at one year. Analyses were performed according to intention to treat, using "baseline observation carried forward" to account for missing data. All treatment programmes achieved significant weight loss from baseline to programme end. When compared with the control treatment at 12 weeks, the only programmes that resulted in significantly more weight loss were Weight Watchers (mean difference 2.53 kg, 95% confidence interval 1.30 to 3.76; P<0.001) and Rosemary Conley (2.18, 0.96 to 3.41; P=0.004). All programmes except general practice and pharmacy provision resulted in significant weight loss at one year. At one year, only the Weight Watchers programme resulted in significantly greater weight loss than the control treatment (2.5 kg, 0.8 to 4.2; P=0.022). It was concluded that commercially provided weight management services are more effective than primary care-based services.

a. If missing data were imputed using baseline carried forward (BCF) method, then what is the assumption that the researchers made about the weight distribution of the participants who did not have their weight measured at 12 weeks and one year. Based on your judgement about the assumption do you think the BCF is a suitable method of imputation?

b. Under the BCF assumption do you think the trial provided biased/unbiased answers?

c. If the researchers had not imputed missing data, but just used complete cases, which assumption of the randomisation would be violated?

d. Comment on what happens to power if the researchers would have just used complete cases?

e. When one analyses the data using intention to treat do you think that the researchers are answering the question that they are interested in?