Male circumcision and risk of HIV infection:

Current epidemiological data

Helen Weiss

London School of Hygiene & Tropical Medicine, UK

HIV seroprevalence in adults, end 2000HIV seroprevalence in adults, end 2000

Systematic review, 1999

• Inclusion criteria:– Studies in Africa– Female to male transmission of HIV-1– Published papers only (up to April 1999)

– 28 studies identified

• Summary risk ratio (RR) obtained using random-effects meta analysis

Adjusted relative risk

.1 .2 .3 .4 .5 1 2 3 4 5

Combined

Urassa-4

Seed

Tyndall

Simonsen

Sassan-Morokro

Mbugua

Diallo

Cameron

Bwayo

Urassa-3

Urassa-2

Serwadda

Quigley

Kelly

Barongo-all

Other studies

High risk studies

Population-based studies

RR<1reduced risk of HIVamong circumcised men

RR=1 (no effect)

Updated analysis - Sep 2002

• Aim: To update the meta-analysis and include data from non-African countries with high HIV prevalence

• Inclusion criteria:– Published studies of F-M transmission in

developing countries– Abstracts from XIV AIDS conference included

Studies included• 11 additional studies identified

– Published literature (9)– Abstracts from XIV International AIDS

conference (2)– 5 cohort studies– 2 non-African studies

• Total of 38 studies, of which 22 adjusted for confounding

Study characteristics• 17 population-based

– 12 cross-sectional, 3 cohort, 2 case-control– 6 Mwanza, 4 Rakai, 3 Kenyan

• 18 high risk groups– STD clinic attendees, truck drivers, TB patients, discordant couples– 11 cross-sectional, 5 cohort, 3 case-control– 7 Nairobi studies

• 3 others - Volunteers, factory workers

Population-based studies - crude RRs

* Additional study - not included in published meta-analysis

cor.1 .2 .3 .4 .5 1 2 3 4 5

Combined

*Kumwende *Kumwende

*Auvert *Agot

*4city-Nd *4city-Kis Wawer-2 Wawer-1 Urassa-3 Urassa-2

Serwadda Sataran Quigley

Pison Kelly

Carael Barongo-3 Barongo-2 Barongo-1

Population based studies - adjusted RRs

* Additional study - not included in published meta-analysis

or2.1 .2 .3 .4 .5 1 2 3 4 5

Combined

*Kumwende

*Kumwende

*4city-Nd

*4city-Kis

Urassa-3

Urassa-2

Serwadda

Quigley

Kelly

Barongo-2

Population-based studiesAnalysis 2000

analysis2002analysis

Crude NRR95% CI

120.930.71-1.21

190.760.59-0.99

Adjusted NRR95% CI

60.560.44-0.70

100.570.47-0.70

RR

High risk groups - crude RRs

* Additional study - not included in published meta-analysis

cor.1 .2 .3 .4.5 1 2 3 4 5

Combined

*Quinn

*Mehendale

*McDonald

*Lavreys

Tyndall

Simonsen

Sassan-Morokro

Nasio-2

Nasio-1

Lankoande

Hira

Greenblatt

Gilks

Diallo

Cameron

Bwayo

High risk groups - adjusted RRs

* Additional study - not included in published meta-analysis

or2.1 .2 .3 .4 .5 1 2 3 4 5

Combined

*Mehendale

*McDonald

*Lavreys

Tyndall

Simonsen

Sassan-Morokro

Mbugua

Diallo

Cameron

Bwayo

High risk group studiesAnalysis 2000

analysis2002analysis

Crude NRR95% CI

120.270.22-0.33

160.300.23-0.40

Adjusted NRR95% CI

70.290.20-0.41

100.310.23-0.42

RR

Analysis by type of studyAnalysis Cohort

studiesX-sectionalstudies

Crude NRR95% CI

90.460.27-0.78

240.510.39-0.66

Adjusted NRR95% CI

50.390.20-0.76

140.420.33-0.53

RR

Is the effect real?• Strong, consistent effect

– very unlikely to be to due to random error

• Significant, strong effect in cohort studies (less susceptible to bias)

• Effect strengthens on adjustment for confounders– effect unlikely to be due to residual

confounding

Limitations• Not a fully systematic review

– Strength of effect may be over-estimated as studies not finding an effect are more difficult to identify

– But - included studies found in recent Cochrane systematic review

• Observational studies only– Possibility of selection biases and residual confounding

• Significant heterogeneity between studies– Effect may differ in different populations

Effect of age at circumcision• Many African tribes circumcise around

puberty. • Biologically plausible that MC has similar

effect irrespective of age at circumcision• Only 2 studies have examined HIV risk in

relation to age at circumcision– Kelly et al; AIDS 1999; 13:399-405– Quigley et al: AIDS 1997; 11:237-248

• Conflicting and inconclusive results

Does MC affect risk of HIV transmission?

• Difficult to assess epidemiologically– Women may have more than one partner– More scope for misclassification

• Biologically less plausible than effect of acquisition of HIV

M-F transmission of HIV• Uganda - cohort study of discordant couples

Quinn et al; NEJM 2000; 342:921-9– Some evidence of reduced transmission among

circumcised males– RR=0.41, 95% CI 0.1-1.1

• Brazil - cross sectional couples studyCastilho et al; XIV AIDS conf. abstr. C10907– No effect of circumcision on HIV prevalence in female

partners of 377 HIV positive men

Lack of circumcision

HIV

STIs

Male circumcision & other STIs

Infection Protectiveeffect?

GUD (syphilis, chancroid)

Herpes simplex virus

Gonorrhoea

Non-gonococcal urethritis X

MC & cervical cancer• Most common cancer in many developing countries

• HPV infection - major cause

• Geographically clusters with penile cancer – Both cancers associated with HPV infection– Lower risk of HPV infection among circumcised men– Lower risk of penile ca. among circumcised men

MC & cervical cancer

• Multi-country analysis of 1913 couples Castellsague et al: NEJM 2002:346:1105-12

– Brazil, Colombia, Thailand, Philippines, Spain• Adjusted OR = 0.72, 95% CI 0.49-1.04

– In monogamous women: • Adjusted OR = 0.75, 95% CI 0.49-1.14

– Penile HPV infection in male partner:– Adjusted OR = 0.37 (95% CI 0.2-0.9)

Current research needsBiological mechanism

Attitudes & feasibility of introducing MC among non-circumcising communities

Effect of age at circumcision

Effect of hygiene practices

? Classification of circumcision through physical examination rather than self-report

Data on safety of current MC practices

? Effect of MC among MSM

? Male-female transmission

? Effect of MC on other viral infections of public health importance (e.g. HPV, HSV)

Conclusions• Observational evidence for a protective effect

of MC on risk of HIV infection is strong and consistent

• BUT cannot exclude selection biases and residual confounding in observational studies

• RCTs will address many of these limitations

Probably not ready to actively promote MC as an HIV prevention measure

What should we do now?• Disseminate current evidence• Continue studies of acceptability & feasibility of MC in

non-circumcising populations with high incidence of HIV• Assess safety of current circumcising procedures• Develop affordable services for safe voluntary MC• Develop educational materials that:

– emphasise that MC may reduce but not eliminate risk of HIV infection

– Separate out issues of male and female circumcision

Summary of 2002 analysis• All studies (n=38)

– crude RR=0.52; 95% CI: 0.42 to 0.64– adjusted RR=0.44; 95% CI: 0.37 to 0.53

• Population-based studies - adjusted (n=10)– RR=0.57; 95% CI: 0.47 to 0.70

• High risk groups - adjusted (n=10)– RR=0.31; 95% CI: 0.23 to 0.42