How can ART observational cohorts and randomized clinical trials inform policy and practice

1
How can ART observational cohorts and randomized
clinical trials inform policy and practice?

• Praphan Phanuphak, M.D., Ph.D.
• Thai Red Cross AIDS Research Centre, Bangkok
  IAS 2005

2
Strengths of randomized controlled trials (RCT)

• Random allocation minimizes selection bias
  confounding. All confounders are evenly
  distributed between the groups being compared.
• Thus, RCT is the gold standard since the results
  are most reliable especially if the results can
  be duplicated by another RCT.

3
Rating scheme for clinical practice
recommendations

• Category Definition
• Strength of recommendation
• A Strong should always be offered
• B Moderate should usually be offered
• C Optional
• D Should usually not be offered
• E Should never be offered
• Quality of evidence for recommendation
• I At least one randomized trial with
  clinical results
• II Clinical trials with laboratory results
• III Expert opinion
• Dybul M et al, Guidelines for using ARVMMWR
  2002

4
Strengths of RCT (II)

• Results can be obtained relatively short.
• HIV diseases are relatively urgent. It is a
  dead or live situation such as AZT monotherapy
  could double the life span of AIDS patients in
  1990 or an infected vs. uninfected situation such
  as the use of AZT in preventing mother-to-child
  transmission (ACTG 076)

5
HIV observational cohorts

• Multicenter AIDS Cohort Study (MACS)
• French Hospital Database on HIV (FHDH)
• Swiss Cohort
• EuroSIDA
• Australian HIV Observational Database (AHOD)
• TREAT Asia HIV Observational Database (TAHOD),
  etc, etc.

6
How the HIV observational cohorts being used?

• Predict disease progression
• Monitor trends in HIV disease outcomes
• - Too sad a story, we are treating physician,
  not horoscopist!
• Monitor trends in HIV treatment
• -Do we try to be in the norms?
• Assess association between HIV treatment and
  disease outcomes
• -Dont we all know?

7
How the HIV observational cohorts being used (2)?

• Compare different populations
• -Can Asians be compared to Caucasians?
• Monitor long-term efficacy long-term
  side-effects
• -Can also be learned by clinical experience
  post-marketing surveillance
• See whether knowledge from clinical trials can be
  applied in clinical practice
• -It depends on how convincing the results are and
  its feasibility

8
How the HIV observational cohorts being used (3)?

• Assess effectiveness of particular ARV
  combinations some agreement such as
• -ZDV monotherapy could have a short-term
  benefit
• -Graham et al, Lancet 1991338265
• -Moore et al, NEJM 19913241412
• -Vella et al, JAMA 19922671232
• -Lundgren et al, JAMA 19942711088
• - Double nucleosides was better than
  monotherapy
• -Graham et al, Ann Intern Med
  19961241031
• -Phillips et al, JAIDS 199717239
• Do we really need these information?
• And sometimes it can be misleading!!

9
Results from some observational cohorts can be
misleading

• Phillips et al, AIDS 1999 132075-82.
• Compared progression to AIDS death between RCT
  (Delta, ACTG175, CPCRA 007, ACTG320) and 3
  cohorts (EuroSIDA, FHDH, Swiss) all agree except
  FHDH which showed higher risk of progression on
  triple therapy as compared to double nucleosides
  (RR 1.20, 95CI 1.01-1.44) while ACTG320 results
  showed a RR of 0.50, 95CI 0.33-0.76. The reason
  was due to those who received triple therapy in
  FHDH were more advanced patients.

10
Weakness of observational cohorts

• Selection bias
• Information (observation, classification,
  measurement) bias
• Confounding bias known unknown
• By chance? what p value? adequate power?
• Causal association?
• At the end, critical clinicians will always ask
  about the experimental evidence, i.e., randomized
  control trials

11
Statisticians magic

• Observational cohorts open the door for
  experienced statisticians to use various magic
  statistical methods to prove or disprove their
  hypotheses such as Weibull proportional hazards
  model, parametric survival model, log-logistic,
  lognormal, generalized gamma and gompertz
  distributions, etc.
• Do any practitioners or policy makers really
  understand these techniques?
  Thus, should one believe or not believe?

12
Translation of RCT results into practice policy

• Practice of experts or general internist Who is
  more difficult to convince?
• Practice in resource-limited settings not very
  many options!!
• Policy guidelines IAS/USA, DHHS, WHO
• Government policy how committed?
• Policy of funding agencies always thinking
  about the best buy for their money, example
  Mwanza trial

13
For practicing physician

• Results from reliable (investigators,
  institutions, journals, carefully planned
  methodology analysis, p value lt0.00..) and
  sizable (Ngt100) RCT, particularly double-blinded
  and confirmed by at least another group of
  investigators
• Recommended in treatment guidelines by
  professional or reliable organizations, not
  necessarily government
• Observational cohorts are nice to learn but will
  believe only if can be confirmed by RCT, e.g.,
  D/C cotrimoxazole prophylaxis after CD4 rise.

14
For policy makers (government funding agencies)

• For a good government policy on treatment
  guidelines should be based on the best reliable
  data, i.e., RCT results because public money will
  be used
• For a bad government wait for long term efficacy
  safety data, i.e., data from long term
  observational cohorts so that they dont have to
  spend money early

15
For policy guidelines makers

• IAS/USA, DHHS other developed country
  guidelines based on the most solid evidence (AI,
  AII, BII)
• Developing country guidelines the minimum
  standards, adjusted on affordability and
  willingness to pay (by the government)
• WHO guidelines minimum standards and applicable
  to all resource-limited countries, no distinction
  between low vs. medium income developing
  countries

16
Examples of translating RCT results into practice
policy

• HAART is better than double NUCs
• (ZDV/3TC vs. ZDV/3TC/IDV Hammer et al, NEJM
  1997 337725,
• 
  Gulick et al, NEJM 1997 337734)
• Half-dose d4T is as good as full-dose d4T (ddI)
• (Ruxrungtham et al, AIDS 2000 14 1375)
• NNRTI is as good as or better than PI
• (EFV vs. IDV Staszewski et al, NEJM 1999
  3411665)
• (EFV vs. ATV Squires et al, JAIDS 2004 361011)
• LPV/r is better than 1st generation PI
• (LPV vs. NLV Walmsley et al, NEJM 2002
  3462039)
• ATV does not cause dyslipidemia as compared to
  other PIs
• (ATV vs. NLV Murphy et al, AIDS 2003 172603,
  Wood et al, JAIDS 2004 36684)
• PMTCT regimens (ACTG 076, HIVNET 012)

17
Moving clinical trials to the South

• Plenty of examples of good clinical trials from
  the South HIV-NAT (Bangkok), Chennai (India),
  South Africa, Argentina, Brazil
• Knowledge, skill, patient population, need and
  willingness are there.
• Infrastructure local funding may not be optimal
  but can be strengthened, not starting from
  scratch.
• But not many trials from the South. Why how can
  we improve it?

18
Rationales of doing clinical trials in the South

• Enormous number of naïve patients HIV-infected
  pregnant women for PMTCT pediatric trials
• Recruitment is rapid, adherence, tolerance
  retention are much better than the trials in the
  North.
• Personnel laboratory costs are lower.
• A way to sensitize academicians, government NGO
  in the country in the region for research ARV
  use
• Opportunity for research on top of the ongoing
  services provided by PEPFAR, GFATM
• Equal gain for the collaborating institutions
  from the North
• What we dont need is a token or a Saint from
  the North.

19
What type of clinical trials that are appropriate
for the South?

• When to start ARV in children (CD4 lt15 vs.
  15-25) and in TB co-infected patients (2-8 wks
  vs. gt8 wks in CD4 lt200)
• Drug interaction studies (RFM NVP 400 vs. 600
  mg, EFV 600 vs. 800 mg)
• Reduced dose regimens to save cost and reduce
  side-effects, e.g., ZDV 200 mg vs. 300 mg bid,
  d4T 20 vs. 30 vs. 40 mg bid in BWgt60 kg, IDV 400/
  600/ 800 mg boosted with RTV 100 mg bid, ATV
  200/r OD

20
What type of clinical trials that are appropriate
for the South (2)?

• Equivalence study of generic drugs in countries
  where only few can afford brand name drugs
• Regimens that one knows that it is safe but
  contra-indicated such as EFV in 2nd-3rd
  trimesters or NVP in pregnancy
• PMTCT regimens such as ZDV/3TC/TDF vs. ZDV/3TC/PI
• STI to reduce cost side-effects

21
Structured treatment interruptions

• Treatment interruption is common in real life.
• STI is proposed to guide safe treatment
  interruption
• Dybul et al were strong advocates for 1 week on,
  1 week off STI
• Both of 2 small cohorts (8-10) showed no rebound
  viremia in any patients for as long as 72 weeks
• PNAS 2001 9815161 (4/10 on NNRTI, 6 PI)
• JID 2004 1891974 (ddI/3TC/EFV OD)

22
1 week on, 1 week off STI was proved ineffective
by RCT

• Ananworanich et al, AIDS 2003 17F33 (Staccato
  study, planned N600 randomized in continued
  treatment arm, CD4-guided arm and
  1-week-on-1-week- off arm). Scheduled preliminary
  safety analysis showed 53 (19/36) virologic
  failure (i.e., 2 consecutive VLgt500 copies/ml) in
  wk-on-off arm, as compared to 2/37 in continuous
  arm (plt0.001) and 0/39 in CD4-guided arm. Thus,
  wk-on-off arm was terminated. Most (11/19) who
  failed were on d4T/ddI/SQV/r (1600/100 OD).

23
Obstacles of doing trials in the South

• Less pharmaceutical interest because not being
  the market or uncertainty about the quality of
  the trials
• Certain limitations from funding agencies such as
  grant cannot be used for ARV, needles, generic
  drugs
• Too much paper work site preparation
• Limitations on boundaries, territories
  networking for international collaboration

24
Conclusions

• RCT is most reliable to establish treatment
  efficacy.
• Its researchers innovation, not
  epidemiologists imagination.
• It can be accomplished rapidly and can be adopted
  rapidly by policy maker practitioner.
• Moving clinical trials to the South is
  compelling.
• Observational cohorts are fun data collection and
  analysis to prove or disprove things.
• If disprove, then another RCT, i.e.,
  RCT is the ultimate goal. Then, why to
  bother about observational cohorts?