The%20SAINT%20Study

1
The SAINT Study

• Jonathan Levin

2

• 1. Results of HIVNET 012
• 2. Design of SAINT
• 3. Results of Petra
• 4. Results of SAINT
• 5. Conclusions from SAINT
• 6. A possible new study

3
1. Results of HIVNET 012

• Lancet 354 No 9181 1999 pp 795-802 Kampala,
  Uganda
• Treatments
• Test (n310)
• 200mg oral nvp to mothers at onset of labour
• 2mg/kg to babies within 72 hours of birth
• Reference (n308)
• 600mg oral azt to mother at onset of labour,
  300mg every 3h until delivery
• 4mg/kg to babies bid for 7 days

4

• Results - HIV transmission rates
• 6-8 weeks
• nvp 11.9
• azt 21.3
• absolute decrease in risk 9.4 (p0.003)
• relative decrease 44
• 14-16 weeks 12 months
• nvp 13.1 nvp 15.7
• azt 25.1 AZT 24.1
• abs decrease 12 8.4
• rel decrease 47 35
• The absolute decrease is maintained (biologically
  plausible)
• The reference is of unproven efficacy - but since
  nvp is superior, its effectiveness is
  demonstrated

5
2. Design of SAINT

• Multicentre South African Study - 11 sites
• Treatments
• Test (n655)
• nvp 200mg oral in labour to mother, followed
  by second dose 24-48 hrs post delivery 6mg oral
  to infant at 24-48 hrs
• Referene (n662)
• Petra B
• Mother ZDV 600mg then 300mg every 3hrs 3TC
  150mg bid plus ZDV 300mg bid for 7 days 3TC
  150mg bid for 7 days infant ZDV 12mg bid for 7
  days 3TC 6mg bid for 7 days

6

• Rationale
• Evaluate safety and effectiveness of arv regimen
  in late presenters
• Primary Objective
• evaluate efficacy of nvp versus AZT 3TC in
  reducing MTCT of HIV

7

• Sample size calculation
• Assume
• (a) With no treatment transmission at 8 weeks is
  25
• (b) Petra B will reduce this to 16
• Then sample size of 655 per group will give 90
  power to detect as significant a reduction by nvp
  to 11.5
• I.e. Assumption was that study would show nvp
  superior to Petra B

8
3. Results of Petra

• HIV transmission at 6 weeks
• Petra A 5.7 (n354)
• Petra B 8.9 (n354)
• Petra C 14.2 (n353)
• Placebo 15.3 (n339)
• So SAINT protocol overestimated transmission in
  placebo group
• Estimated benefit of Petra B over placebo is 6.4
• (Relative decrease is 42)

9

• HIV transmission at 18 months
• Petra A 14.9
• Petra B 18.1
• Petra C 20.0
• Placebo 22.2
• Estimated benefit of Petra B over placebo is 4.1
• (Relative decrease is 18)

10
4. Results of SAINT

• Overall HIV transmission at 8 weeks
• NVP 12.3 (n643)
• Petra B 9.3 (n646)
• Estimated benefit of Petra B over nvp is 3
• no follow up to 18 months

11

• Maternal NVP resistance
• At 4 weeks after delivery 68 (74/109) of mothers
  had resistant virus
• Follow up specimens were collected 6-18 months.
  11.8 of isolates were still resistant
• Resistance in SAINT more serious than in HIVNET
  012
• Perhaps due to 2 doses of nvp to the mother

12
5. Conclusions from SAINT

• Researchers
• Considerable overlap of confidence intervals
  and reduced transmission rate (compared to
  hypothesized placebo rate of 25) confirm that
  both treatments are effective
• N.B. Altman and Bland
• Absence of evidence is not evidence of absence

13

• Correct Statistical conclusion
• There is insufficient evidence to conclude that
  Petra B is superior to nvp
• SAINT is inconclusive
• it does not show that nvp is effective
• it does not show that nvp is not effective

14

• HIVNET 012 and SAINT do not give consistent
  results
• Possibilities
• Results of SAINT unduly pessimistic (? Effect of
  using late presenters)
• Results of HIVNET 012 unduly optimistic
• True results are not consistent over the two
  sites (RSA versus Uganda)
• (? Subtype C versus subtype A)
• We do not have enough information to make a
  conclusion on above

15
6. Possible Alternative Study

• Was SAINT an equivalence study?
• No because it was not designed as an equivalence
  study.
• For an equivalence study we must pre-specify
  equivalence limits
• e.g. since in Petra, Petra B was better than
  placebo by 6.4 could choose limits of 3.2

16

• If we had done this, confidence interval for
  difference between Petra B and nvp must lie
  between (-3.2 3.2).
• The actual confidence interval for the difference
  is
• (-0.4 6.4)
• So even if it were designed as equivalence study
  it would not have shown equivalence

17

• Possible study - non inferiority of nvp versus
  Petra A or nvp versus Thai AZT regime (Botswana)
• Specify non inferiority margin of 5
• (So this says that nvp would lead to at most 5
  more transmissions than Petra A/AZT Thai)