UK-CAB 7 – Meeting Report

Introduction to statistics 1 – Gilead

21 November 2003


Members attending
Introduction Aims of day
Training: Introduction to statistics

Dr Caroline Sabin, Royal Free Hospital

PowerPoint slides from the meeting

Colour slide collection PDF File [628k]
Black and white slide collection PDF File [448k]

3. Gilead

Geraldine Reilly

FTC PowerPoint Slides [5.6MB]
FTC Slides as PDF File [2.7MB]
Tenofovir Powerpoint Slides [2.2MB]

i) FTC update
ii) Recent tenofovir interactions:

  • ddI
  • atazanavir
  • methadone, oral contraceptives, other HIV drugs
  • triple nucleoside combinations (TDF/ABC/3TC and TDF/ddI/3TC)

iii) Update on side effects

4. Internal: Programme for next CAB

Members attending

Joyce Attaro Positively Women London
Steve Atkinson HIV iBase London
William Babumba African HIV Policy Network London
Gus Cairns Positive Nation London
Polly Clayden HIV iBase London
Simon Collins HIV iBase London
Ben Cromarty North Yorkshire AIDS Action Yorkshire
Marc Ennals HIV iBase London
Robert Fiedhouse UK Coalition London
Paul Foster HIV iBase London
Jenni Fredriksson Avert West Sussex
Cathal Gallagher Whipps Cross Hosp/Living Well London
Jim Jewers The Globe Center London
Ben HillJones Avert West Sussex
Blessing Sibanda West Yorkshire African Group Yorkshire
Mohamade Jowata Brent PCT London
Edith Kaggwa UK Coalition London
Martin Leigh UK Networx HIV Network Middlesex
Peter Lovell THT West Devon
Badru Male Brent and Harrow PCT London
Tendai Ndanda Rain Trust London
Rachel Nkama Uganda AIDS Foundation London
Jo Robinson THT London
Prasa Velisetty Mildmay London
Carmen Tarrades Intl Community of Women London
Report by Simon Collins and Steve Atkinson, HIV iBase

Link to top of page

1. Introduction to the day

i-Base welcomed everyone to the seventh UKCAB met on Friday 21 November.

Instead of the usual two training sessions in the morning, this meeting included a double session from Dr Caroline Sabin on an introduction to statistics.

So many terms are used when discussing trials and research that we wanted to include a very comprehensive training that would help people find it easier to understand trial results.

Unless we understand the basic differences behind different studies it is impossible to decide whether we should believe the results as they are reported, and report those results to the wider community.

By understanding the terms used, and the way statistics are presented, we are in a more powerful and better informed position to make these judgments.

The afternoon meeting was with Gilead on recent approval of FTC and drug interactions between tenofovir and other HIV drugs.

A small internal meeting at the end of the day discussed the programme for the next meeting.

Powerpoint slides from the meeting are now posted to the iBase website.
Link to top of page

2. Training: Introduction to Statistics

Dr Caroline Sabin, Royal Free Hospital, Reader in Medical Statistics and Epidemiology

Caroline has been a statistician involved in HIV research for many years – first starting with the Haemophilia Cohort at the Royal Free Hospital and has been involved with many of the largest and most important UK international cohort studies.

The training covered two parts of a six part course in statistics, and the next module from the course will be given at the next CAB meeting in February 2004.

Please refer to the excellent slides for this training when reading these notes.

Colour slide collection PDF File [628k]

Black and white slide collection PDF File [448k]

Part one: different types of study

The first paragraph of any study summary, or any good report of a study, will always include a few words describing the format of the study:

It will usually be either

‘experimental’ or ‘observational’

  • ie does the study look at an intervention like a new drug or a new test (for example, deciding to study people in a preplanned way to switch patients from d4T to abacavir to improve lipodystrophy) OR does it just observe what occurs in a group of patients where there is no planned intervention but, for example, where a group a patients on treatment would be followed over time and the results analysed by whether they switch or stayed on the same treatment.

‘crosssectional’ or ‘longitudinal’

  • ie does the study just take a ‘snapshot’ of results at one point in time, for example, to see how many people in a clinic had lipodystrophy in April 2003 to see the rough prevalence of this side effect OR does it follow people over time to see how many of them develop lipodystrophy over the first 2 years on HAART.

‘prospective’ or ‘retrospective’

  • does the study start, and then collect data going forward in time, ie decide to see how many treatment naïve patients will develop lipodystrophy when using tenofovir compared to d4T in a comparison between the two drugs OR does it collect data from medical records and look backwards to try to find a pattern, ie take patients at a clinic and decide who does or doesn’t have lipodystrophy, and collect information about which drugs and other factors they had in the past that are linked with each group form their patient notes.

Different study designs have advantages and disadvantages, Unless the study is designed carefully to answer specific questions, the results can be very limited. Many studies leave more questions unanswered than they answer. For example, you can see from the examples above, that you would get very different information about lipodystrophy from each of these approaches.

The main types of study, ranked in order of the quality of evidence they provide, are listed below, along with whether they are experimental or observational, crosssectional or longitudinal, and prospective or retrospective.

Another set of terms describing different studies relates to the ‘quality’ or the results in terms of how ‘definite’ or ‘reliable’ they will be seen. So the examples above describe different studies needed to answer different questions, the following terms are ranked generally in order of the quality of the evidence they provide.

Randomised controlled trial (RCT)

This is seen as the ‘gold standard’ for collecting evidence. Patients are randomized to two groups, one of which will receive regimen A and the other regimen B. The process of randomization means that the two groups will have similar characteristics (ie same age, CD4 count etc) at the start of the trial. These trials are usually also ‘blinded’ or ‘doubleblinded’ where patients, or patients, doctors and researchers analysing the results do not know who is getting which treatment.

By definition these are ‘experimental’ and usually ‘longitudinal’. Although such studies are the ‘gold standard’ they still do not provide answers to every question. For example in placebo studies you end up taking more pills, so benefits of reduced pills and effect on adherence counts for some regimens are not captured in the results.

As well as ‘parallel’ trials – where two or more different groups are followed over time, you can also have ‘crossover’ studies, where the same people try both treatments. The benefits of crossover studies are that they require fewer people, and can allow for individual differences, but they are only useful to study an intervention that provides a shortterm benefit. You also need to make sure that other factors are accounted for – ie a treatment for arthritis would be better during the summer than in winter.

Cohort study

Cohort studies are ‘observational’ and ‘longitudinal’ a group of patients is followed in a way that is more inclusive of the wider population. In the US and Europe, cohort studies are like any other study whereby patients attend a special clinic for their ‘cohort visit’. In the UK, however, many of these cohort studies have evolved from the databases that clinics have set up to monitor the use and outcomes of therapy. Thus, no active patient involvement is required. These studies have been invaluable for our understanding of HIV and effective treatment. These cohorts include all types of people, including those who would not normally join, or who are excluded from joining clinical trials (ie people with coinfections, who are very ill, or who have less organised lifestyles).

Disadvantages of cohort studies is making allowances for ‘confounding factors’ other factors that could influence the results – and that people who change clinic can be lost from the study.

Casecontrol study

These are ‘observational’ and ‘longitudinal’ (and traditionally retrospective). A group of patients with a disease (cases) are compared to a group of patients without the disease (controls). The aim is to see whether exposure to any factor has occurred more or less frequently in the past in the cases than in the controls. Cases and controls may often be matched on basic demographic information (eg. sex and age) to make the two groups as similar as possible.

Advantages are that they are relatively cheap, quick and easy to carry out, with no losstofollowup and are suitable for rare events. Disadvantages include the potential for ‘recall bias’ when the timing or accuracy of events cannot be reliably established therefore more difficult to assess causality.

Crosssectional study

These are carried out at a single point in time with no followup. It usually involves a questionnaire or survey to assess the prevalence of a condition, to describe the current situation or to assess attitudes and beliefs. Advantages include relatively cheap and quick. Disadvantages are that is not possible to estimate incidence of disease and you don’t know what happened before or after to each of the patients.

Case series/case note review

These are considered a fairly low form of evidence but can provide useful preliminary data ie. to define the natural history of disease or to describe current practices. There is no comparative element therefore it is not possible to show a link between exposure and disease. By definition, these studies are usually retrospective therefore, potential for problems with historical data.

Literature review/systematic review

This can be useful when there is already a lot of published data on a topic. A literature review is any review of the literature, not necessarily done in a systematic manner, and with no attempt to identify all potential studies. A systematic review is more comprehensive and useful as attention is paid to ensure that all available studies are included. Aim of both is to describe findings and variations between studies, rather than to try and establish a common estimate of the effect size


This is when results from several studies are pooled give a combined estimate of the results. It can be useful when individual published studies are generally too small to reach reliable conclusions. It can be done using published data (extension of a systematic review) or by obtaining raw data from the study investigators. Effectiveness is limited by ‘publication bias’ – where only trials with significant results are published or presented. They are more comprehensive if nonpublished studies (eg. conference abstracts, studies in progress) are also included.

Part Two: Describing and displaying data

Just as there are new terms to be able to understand trials, we need a few specialised terms to understand the data that studies produce.

Before analysing data, the first approach is to look at the data that are collected and then decide how to present them. Tables of figures are difficult to read, but graphs and charts are usually much better tools to see whether there are any clear results.

Decide what type of data we have:

i) Qualitative (categorical) ie like race/ethnicity [sometimes with 2 choices, sometimes more and sometimes the categories can be ordered in some way]


Quantitative (numerical) ie age, height etc [sometimes this can be only ‘whole numbers’ ie number of sexual partners, and sometimes continuous like height]

ii) Check for possible errors or outliers (unusual values) – these can highlight unexpected results that may just be errors in the way the information was collected, or real but unusual results that may skew the final figures. The show easily on a when plotted in a graph.

spotting errors powerpoint slide
Click on the images for a full sized view

ii) If appropriate, describe the shape of the distribution.

shape of data powerpoint slide
Click on the image for a full sized view

The shape of distribution of data is actually very important, but is a new concept if you have not studied statistics. When looking at results in a group of people, you expect most people to be roughly in the middle and fewer people to tailoff at the upper and lower levels – so most people are average height with equal, but fewer, numbers of people being either very tall or very short, and if plotted on a graph you get a ‘bellshape’ distribution.

It is important to know the shape of the distribution because when analysing numerical data, the choice of the most appropriate method will depend on the shape of the distribution. Also, the choice of summary statistics used to display the data will depend on whether the data are have a ‘normal’ distribution.

Describing quantitative data

The two main measures are used to describe data are a measure of ‘average’ and a measure of ‘variability’. Two groups for example could have the same average results but a very different range of results.

Every study includes a measure of ‘average’ (if for CD4 count or viral load at baseline and at the end of the study)

This can be:

i) (arithmetic) mean (all the results added up and then divided by the number of results)


ii) median (the result in the middle, when all the results are lined up in order)

The mean performs well for symmetrical data and the median for when data is skewed.

mean curve median curve

Assessment of variation is slightly more complicated but it is even more important. You will still see many poorly reported studies that report average values but do not report how variable the data are.

The four main measures of variability (or variation) are:

Variance [this is the sum of (each result minus the mean result) 2/ divided by the number of results minus1 – and is always expressed as a positive value]

variance and deviation powerpoint slide
Click on the image for a full sized view

Standard deviation

Because variance is a large number, unrelated to any measure, the ‘square root’ of the variance – called the ‘standard deviation’ or SD is the measure that is really used.

Standard deviation is only used when data are ‘normally’ distributed – and 95% of values will be within 2xstandard deviation each side of the mean value. So if mean CD4 count is 500 and SD is 200, then 95% people in that group will have CD4 counts within the range of 100900.

more on SD powerpoint slide
Click on the images for a full sized view


Interquartile range

The ‘range’ and ‘interquartile range’ (IQR) are much easier to understand. These values are used when the distribution of data is ‘skewed’ or not normal.

The range is simply the difference between the two most extreme values, between the highest and lowest figures.

The ‘interquartile range’ (IQR) is the section where the central 50% of data lie. It is worked out, like the median, by placing the results in order and then identifying the values below which 25% and 75% of the values lie.

range powerpoint slide
Click on the image for a full sized view

Which values are used therefore depends on whether data distribution is ‘normal’:

So, usually use:

EITHER Mean + variance/SD OR Median + range/IQR

Even though studies sometimes quote the mean and range, or the median and SD, statistically this is nonsense…

Finally, because the calculations that can be run on skewed data are limited, sometimes all the data can be transformed (as with using logs for viral load) to make the data distribution similar to a normal bell shaped distribution.

‘Confidence intervals’: dealing with uncertainty!

The last technical term used in describing data covered in the lecture was the term ‘confidence interval’ which is often abbreviated to ‘CI’. This is a mathematical calculation for the possible range of values that could still be consistent from the sample data. The 95% CI gives you a range of values in which you are 95% certain that the true value lies – it is a range of values that is consistent with the data in your sample.

If the confidence interval is wide, then the estimate is not very good. If it is narrow then the results are more reliable. So if results from a study say that the chance of a drug working is 70% and the 95%CI is 6575%, then you’d could think this is okay. But if the 95%CI is 35%95%, then the estimate from the study is much less reliable. A good way to judge whether the results of a study are reliable is to look at both the upper and lower limits of the CI. If the true value took either of these limits, would your conclusions from the study change?

Sometimes, a wide confidence interval just means that the study was very small and that the results need to be confirmed with more patients. The slide below shows that with exactly the same results, width of the CI is closely related to sample size and the CI becomes narrower as sample size increases. However, there is also an upper cutoff where increasing the study even more, only has a minimum effect on the CI.

interpretation of CI powerpoint slide
Click on the images for a full sized view

The final point about ‘confidence intervals’ shown in the second slide above, is that when confidence intervals overlap, we cannot draw any conclusions about whether or not there is a real difference between the groups.

Even if it appears as if there is a difference between the two groups, we need to assess whether this difference could have occurred by chance or is likely to be a real finding…

This links to the meaning of ‘Pvalues’…. which will be continued in the next CAB meeting on Friday 27 February 2004.

If you have already made it this far, you probably think you are learning more than you ever wanted to about statistics, but there is a real practical importance to these terms.
Link to top of page

3. Meeting with Gilead

The afternoon meeting with Gilead began with a minutes silence. This was in memory of people who were unable to benefit from treatment, or who are not able to access current treatment today because of where they live. This was followed by members of the group introducing themselves.

Geraldine Reilly trained as a nurse and specialised in HIV care for 14 years. She only recently started working for the pharmaceutical industry when she joined Gilead a couple of years ago.

Comprehensive slides from Gilead on the data from this session are posted to the iBase website for the November CAB. Gilead manufacture tenofovir and FTC and the hepatitis B drug adefovir.

FTC PowerPoint Slides[5.6MB]

FTC Slides as PDF File[2.7MB]

i) FTC (emtricitabine, Emtriva)

The first part of the talk focused on new information about FTC.

This is a onepill, oncedaily nucleoside that is very similar to 3TC in its chemical structure, but it has a longer ‘halflife’. This means that it takes longer for drug levels to drop in plasma and inside cells, and theoretically will have less risk for developing resistance if you are late or miss a dose for any reason.

FTC and 3TC have the same resistance pattern. If you have the 3TCassociated M184V mutation, then you will have crossresistance to both FTC and 3TC.

In registrational studies FTC with (ddI and efavirenz) was compared to d4T (in twicedaily combinations) and was more effective with fewer side effects (Study301). The oncedaily combination of FTC/ddI/efavirenz was studied as firstline therapy and in a PIswitch study (ANRS091 and 099). Study 303 switched patients from 3TC to FTC and few differences were seen when compared to 3TC.

A broad conclusion from these studies is that FTC is an effective and tolerable nucleoside. A headtohead study against 3TC in oncedaily combinations has not been performed.

Hyperpigmentation, (slide shows small darker frecklelike skin changes) primarily on the palms of hands and soles of feet, was reported in 3% patients but this is higher at 8% in Black patients. This occurred after around 3 months on treatment, is predominantly grade1, and did not lead to discontinuation of FTC.

The development programme for FTC includes coformulation with tenofovir in a single pill.

Further research is planned with hepatitis B.

ii) Tenofovir drug interactions

Several other antiHIV drugs interact with tenofovir, the mechanisms for some of which are not completely understood.

Tenofovir and ddI

This interaction should now be widely known. The practical conclusion is that when used in the same combination, the dose of ddI should be reduced (from 400mg down to 250mg; or from 250mg down to 200mg). Both drugs should then be taken together, with or without food (though many people find tenofovir easy to take with food).

Tenofovir and Kaletra (lopinavir/r)

Kaletra increases tenofovir levels (total drug exposure – ‘Area under the curve’) by around 30%, but this isn’t considered significant, and hasn’t led to any increased reports of tenofovir toxicity.

Tenofovir and atazanavir

Tenofovir reduces levels of atazanavir (AUC by 25% and trough levels by 40%) and this is significant. The recommendation is to boost 300mg atazanavir with 100mg ritonavir, when tenofovir is in the same combination. Atazanavir concentrations are then well above the levels achieved with the 400mg unboosted dose. Although tenofovir levels are also increased, this has not led to increased tenofovir side effects.

Methadone, oral contraceptives

No interactions have been found between tenofovir and methadone, oral contraceptives. other nukes, PIs or NNRTIs.

Tenofovir in triple nucleoside combinations

Two separate triplenucleoside combinations – i) tenofovir/abacavir/3TC and ii) tenofovir/ddI/3TC performed very poorly in different studies. These combinations should not be used. The mechanism for the poor performance is not understood. Various possible explanations for single factors are not supported by other studies. Even the overlapping resistance profile, although the current explanation most favoured, is not bourne out by the resistance profile of patients who failed in these studies.

iii) Update on side effects and tenofovir

The meeting also briefly covered monitoring of kidney function, recent recommendations for dose adjustment in people with renal impairment, and again briefly, some data on bone disease. The slide set for the talk has more detail about all these areas.

Renal impairment

  • Dosing interval adjustment is recommended in all patients with creatinine clearance < 50 mL/min. Recommendation based on creatinine clearance are ‘every 48h if 3049 mL/min; twice a week if 1029 mL/min and every seven days or approx 12 hours after dialysis for patients on hemodialysis.
  • The majority of reported cases of renal impairment in patients taking tenofovir occurred in patients with underlying systemic or renal disease, or in patients taking nephrotoxic agents
  • Patients at risk for, or with a history of, renal dysfunction and patients receiving concomitant nephrotoxic agents should be carefully monitored
  • Tenofovir should be avoided with concurrent or recent use of a nephrotoxic drugs.
    Examples of drugs that can cause kidney damage and dysfunction may include the following:
  • aminoglycoside antibiotics – amikacin (Amikin), gentamicin, paromomycin (Humatin), streptomycin, tobramycin
  • other antibiotics – Septra (Bactrim, cotrimoxazole, trimethoprimsulfamethoxazole)
  • antifungals – amphotericin B (Fungizone) and related formulations of this drug
  • antivirals – acyclovir (Zovirax), adefovir (Hepsera), cidofovir (Vistide), foscarnet (Foscavir), indinavir (Crixivan), Valtrex (valacyclovir)
  • antiparasite drugs – intravenous pentamidine
  • NSAIDS (nonsteroidal antiinflammatory agents) – acetaminophen (Tylenol), ibuprofen (Advil, Motrin), indomethacin (Indocid), naproxen (Naprosyn)

Link to top of page

4. Internal meeting: Programme for 2004

The next meeting will either be the Friday 27 February. We also have a programme of dates in advance for 2004. All dates are for Fridays.

February 27 To include feedback from Retrovirus conference (Feb 04). In addition to the second part of Caroline Sabin’s Statistics course, training will include how to access to treatment and tests when included in guidelines but not provided in your clinic – ie general advocacy; and the afternoon session will also include a section for feedback from other groups (ie the new Treatment Action Campaign support group in the UK).

May 21 To include feedback from BHIVA conference ? (Training perhaps on immunology? and paediatric care?)

August 13 To include feedback from World AIDS Conference, (Thailand July 04)

November 19 To include feedback from Autumn Meetings (Lipodystrophy, Glasgow)
Link to top of page

5. AOB

The meeting finished at 4.30pm.

Report: Simon Collins and Steve Atkinson, HIV iBase
November 2003
Link to top of page

Published: November 21, 2003
Last edited: December 19, 2010