25 aÑos de batalla contra el sida:...
TRANSCRIPT
25
AÑOS DE BATALLA CONTRA EL SIDA: 1983-2008
Alvaro Muñoz, PhDProfesor
de Epidemiologia
[y de Fabian Mendez]Cali,
2008.450
AMuñoz
y UniValle4
1983: Conferencia: The use of an autoregressive model for the analysis of longitudinal data in epistemiological studies.
4
1984: Goldberg MB, Guerrero R, Muñoz A. Mortalidad neonatal temprana en Cali. Un estudio de casos y controles. Colombia Medica
4
1985: Curso: Análisis de Estudios de Casos y Controles. Cali. Co-instructor con Norman Breslow.
4
1990: Co-institucion con Hopkins en beca de la Fogarty en SIDA4
1991: Curso: Epidemiologia del SIDA. PalmiraVinieron
de todas
partesdizque
a darsen
la buena
vidapero
resultaron
metidosen la epidemiologia
del sida
[por
SOspina]4
1992: Trujillo JM, Concha
M, Muñoz
A, Bergonzoli
G, Mora C, Borrero
I, Gibbs CJ, Jr., Arango
C. Seroprevalence
and cofactors of HTLV-I infection in Tumaco, Colombia. AIDS Res Hum Retroviruses
4
1993: Alzate A, Crespo MDP, Carrasquilla G, Corral R, Sánchez N, Muñoz A. PPD and HIV infection in Cali, Colombia. J Acquir Immune Defic Syndr
AMuñoz
y UniValle
(cont.) 4
2000: Méndez F, Carrasquilla G, Muñoz A. Risk factors associated with malaria infection in an urban setting. Trans R Soc Trop
Med Hyg
4
2002: Mendez F, Muñoz A, Carrasquilla G, Jurado D, Arevalo-
Herrera M, Cortese JF, Plowe CV. Determinants of treatment response to pyrimethamine-sulfadoxine
and subsequent transmission potential in non-severe falciparum
malaria. Am J Epidemiol
4
2004: Muñoz
N, Mendez F, Posso
H, Molano
M, van den Brule AJC, Ronderos
M, Meijer C, Muñoz
A, for the INC HPV Study Group. HPV infection. J Infect Dis
4
2005: Mendez F, Muñoz N, Posso H, Molano M, Moreno V, van den Brule AJC, Ronderos M, Maijer C, Muñoz A, for the INC HPV study group. Coinfection
with HPV types. J Infect Dis 4
2006: Mendez F, Muñoz
A, Plowe
CV. Use of area under the curve to characterize transmission potential after antimalarial
treatment. Am J of Tropical Med and Hygiene
4
2007: Mendez F, Herrera S, Murrain B, Gutierrez A, Moreno L, Manzano M, Muñoz A, Plowe C. Selection of Antifolate
resistant P. Falciparum
by Sulfadoxine-Pyrimethamine
treatment and infectivity to Anopheles Mosquitoes. Am J of Tropical Med & Hygiene
Epidemiologia
del VIH/SIDAInfeccion con HIVBiomarcadores•
Virologicos: Cantidad
del virus en la sangre
(copias/ml) •
Immunologicos: Grado
de immunodepresion
(conteo
de celulas
CD4 por
mm3)Intervenciones: terapia antiretroviral, profilaxis contra infecciones oportunisticasEventos clinicos: SIDAMuerte: tiempos de supervivencia
Distinguishing Efficacy, Individual and Population Effectiveness of Therapies
Muñoz
A, Gange
S, Jacobson LP. AIDS, 2000; 14:754-6.
6
Epidemiologic ConceptsMuñoz, Gange, Jacobson AIDS’00
Clinical TrialsObservational/Cohort Studies
Efficacy
Effectiveness
Individual Population[Efficiency]
Phillips et al., AIDS ’99Sabin et al., AIDS ’99
Benson et al., NEJM ’00Concato
et al., NEJM ’00
Hoover et al. Stat Med’94Detels
et al., JAMA ’98Porter et al., Lancet ’03
Schneider et al., AIDS’05Cox et al., Stat in Med’07Brenner et al., Lancet ’02
Cole et al., AJE ’03
Hammer et al., NEJM ’97Gulick
et al., NEJM ’97
Stukel
et al.,
JAMA’07
7
Concato
et al Conclusions4“The popular belief that only randomized,
controlled trials produce trustworthy results and that all observational studies are misleading does a disservice to patient care, clinical investigation, and the education of health care professionals.”
4“La creencia
popular de que
solamente ensayos
controlados
y con aleatorizacion
producen
resultados
fidedignos
y que
todos los estudios
observacionales
son engañosos
le sirve
negativamente
al cuidado
de pacientes, a la investigacion
clinica
y a la
educacion
de las
profesionales
de la salud”Source: N Engl J Med 2000, vol 342, p 1889
8
Setting / Study
Efficacy Clinical Trial
Individual Observational/Cohort Effectiveness
Population Observational/Cohort Effectiveness
9
Groups ComparedEfficacy Individual A-Treated
Individual B-Untreated
Individual Individual A-TreatedEffectiveness Individual B-Untreated
or Within individuals -
Pre-treatment compared to Post-treatment
Population Population A when most illEffectiveness are treated
compared toPopulation A when none
are treated
10
ComparabilityEfficacy Randomization
Individual Stratification/Regression; Effectiveness Inverse Probability Weights
Population Groups with similar Effectiveness times at risk
or Groups with similar markers at start of eras
11
Public Health Impact
Efficacy Results under controlled conditions
Individual Supplements (confirms) Effectiveness results of clinical trials
Population Complements clinical trials; Effectiveness Measures reduction of
disease burden resulting from treating subgroups at highest risk
12
Drawbacks
Efficacy Not necessarily applicableto “real world”
Individual Residual confounding due Effectiveness to selection by indication
Population Possible ecological Effectiveness fallaciousness
14
Patterns of the Hazard of Death after AIDS through the
Evolution of Antiretroviral Therapy: 1984 -2004
Muñoz, Gange, Jacobson. AIDS; 2000
Schneider, Gange, …, Muñoz.
AIDS; 2005
15
Objective of analysis:In individuals with the same years since AIDS diagnosis and of the same age, type of AIDS and CD4 cell count at AIDS diagnosis, the hazard of death has quantitatively
and
qualitatively
changed in the HAART eras.
Objective
Years after AIDS diagnosis time axis
calendar exposure Age, Type of AIDS, CD4 cell count at diagnosis confounders
Death event
16
Progression of HIV-1 Infection prior to HAART
Muñoz, Xu. Stat Med ’96; Enger
et al. JAMA ’96; Jacobson et al. AJE ’93 -
Update
2.7
10.3
8.9 median1.3
5.3 5 percent
SC AIDSCD4≤
200 DEATHAIDS DEATH
1199/21371093/13131213/1620
Origin Event1.0
0.8
0.6
0.4
0.2
0.0
50 10 15Time in Years
Prop
ortio
n
October 1998
17
Calendar Time
Perc
ent R
ecei
ving
The
rapy
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0102030405060708090100A
Calendar Time
Perc
ent R
ecei
ving
The
rapy
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0102030405060708090100
Study Began in October 1994
B
MonotherapyCombination
Therapy
PI-HAART
NonPI-HAART
NonPI-HAART
PI-HAARTCombination Therapy
Monotherapy
Multicenter AIDS Cohort Study
Women’s Interagency HIV Cohort Study
Schneider, Gange,…, Muñoz. AIDS 2005
18
Periods
July 1984 –
Dec 1989 no/mono therapy
Jan 1990
–
Dec 1994 mono/combination
(dual)
Jan 1995
–
June 1998 HAART introduction
July 1998 –
June 2001 Short-term stable HAART
July 2001 –
Dec 2003 Moderate-term stable HAART
19
Number Seen
Incident AIDS Deaths
Period
1984.5 –
1990.0 633 100% 61%
1990.0 –
1995.0 660 100% 67%
1995.0 –
1998.5 472 100% 23%
1998.5 –
2001.5 496 29% 14%
2001.5 –
2004.0 464 12% 9%
Descriptive Statistics by Period
20
Person Visits
Median CD4
Median HIVRNA
Adherence <95%
Period
1984.5 –
1990.0 690 65 169,105 NA
1990.0 –
1995.0 1,173 37 175,198 NA
1995.0 –
1998.5 1,555 242 10,000 10%
1998.5 –
2001.5 2,131 317 2,274 24%
2001.5 –
2004.0 1,600 349 927 21%
Descriptive Statistics by Period (cont.)
21
Data Analysis Methods
Survival methods for staggered entries (i.e., at the start of a given period, those diagnosed in previous periods and alive enter “late” to the given period)
Calendar as external time-dependent exposure
Relative times : parametric (e.g., weibull) regression
1984.5 - 1990.0
0 2 4 6 8
020
040
060
01990.0 - 1995.0
0 2 4 6 8
020
040
060
0
1995.0 - 1998.5
0 2 4 6 8
020
040
060
0
1998.5 - 2001.5
0 2 4 6 8
020
040
060
0
2001.5 - 2004.0
0 2 4 6 8
020
040
060
0
1984.5 - 1990.0
0 2 4 6 8
020
040
060
0
633= 245 + 388 660= 215
+ 445
472= 363
+ 109
496= 425
+ 71
464= 420
+ 44
: 57 per 100 PY: 49 per 100 PY
: 14 per 100 PY
: 6 per 100 PY: 4 per 100 PY
23Years from AIDS diagnosis
Cum
ulat
ive
Mor
talit
y af
ter A
IDS
(%)
0
20
40
60
80
100
Kaplan-MeierWeibull
1st QuartileMedian
3rd Quartile
0 2 4 6 8
June 1984 –
Dec 1989($=0.55; F=0.90)
Jan 1990 –
Dec 1994($=0.70; F=0.81)
Jan 1995 –
June 1998($=2.48; F=1.53) July 1998 –
June 2001($=2.96; F=1.22)
July 2001 –
Dec 2003($=3.25; F=1.31)
Schneider, Gange,…, Muñoz. AIDS 2005
24Years from AIDS Diagnosis
Haz
ard
of d
eath
afte
r AID
S (%
)
0
20
40
60
80
100
0 2 4 6 8
1st QuartileMedian
3rd Quartile
Therapy Era 1st
Quartile Median
Jun 1984 –
Dec 1989 0.6 years 1.2 years
Jan 1990 –
Dec 1994 0.7 years 1.5 years
Jan 1995 –
Jun 1998 1.8 years 6.8 years
Jul 1998 –
Jun 2001 4.2 years 12.3 yearsJul 2001 –
Dec 2003 5.0 years 16.0 years
Schneider, Gange,…, Muñoz. AIDS 2005
25
A General Gamma-Based History of Survival after AIDS: 1984-2004
Cox C, Chu H, Schneider M, Muñoz
A. Parametric Survival Analysis and
Taxonomy of Hazard Functions for the Generalized Gamma Distribution.
Stat Med 2007;26:4352-74.
26
Parametric Models for Survival Data
•
Exponential, Weibull, lognormal, gamma, log logistic
• Do not require proportionality of hazards
•
Maximum likelihood for data with right, interval and left censoring; and data with late entries (truncation)
• Complete description of hazards
• Relative times and relative hazards
27
Challenges in Parametric Analysis of Survival Data
• Choosing among many alternatives
• Avoid being too restrictive (e.g., Weibull)
•
Family should incorporate hazards which are:
– increasing
– decreasing
– arc-shaped
– bathtub-shaped
28
• λ
→ 0 corresponds to Lognormal• λ
= 1 corresponds to Weibull
• Time at which p% fail:
Proportion surviving greater then time units:
where 1
0
1( ; )( )
z uz u e duγγγ
− −Γ =Γ ∫ is the Gamma c.d.f. with
mean and variance equal to
2 / 2
2 / 2
( ) 1 [ ( ) ; ] 0
[ ( ) ; ] 0GGS t e t
e t
β λ σ
β λ σ
λ λ λ
λ λ λ
− − −
− − −
= − Γ >
= Γ <
γ
Cox, Chu, Schneider, Muñoz. Stat Med 2007
Generalized Gamma Parameters: location, scale, shape $, F , 8
( )log[ ] log[ ( )]t p t pλβ σ= +where t ( ) is the pth percentile of the standard (0,1, )p GGλ λ
t
29
0 1 2 3
-1
0
1
2
3Shape (λ)
Scale (σ)
λ = σ
Gamma
Weibull
Lognormal
Ammag
Inverse Gamma 8 = -F
λ = 1/σ
Inverse Weibull
Inverse Ammag
8
= -1/F
Cox, Chu, Schneider, Muñoz. 2007 Stat Med
30Years from AIDS diagnosis
Cum
ulat
ive
Mor
talit
y af
ter A
IDS
(%)
0
20
40
60
80
100NonparametricGeneralized Gamma
0 1 2 3 4 5 6 7 8
July 1984 -
Dec 1989
Jan 1990 -
Dec 1994
Jan 1995 -
June 1998 July 1998 -
Dec 2003
Fit of the General Gamma vs. Nonparametric
31Cumulative Percent Deceased
Rel
ativ
e Ti
me
(Deciles for Period 1 : 0.23, 0.47, 0.72 years)
Period 2 to Period 1 Period 3 to Period 1 Period 4 to Period 1
Period 2: Jan1990-
Dec1994
Period1: Jul1984-
Dec1989
Period 3: Jan1995-
Jun1998
Period1: Jul1984-
Dec1989
Period 4: Jul1998-
Dec2003
Period1: Jul1984-
Dec1989
Relative Times Cox, Chu, …, Muñoz. Stat Med 2007
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
32Cumulative Percent Deceased
Rel
ativ
e Ti
me
(Deciles for Period 1 : 0.23, 0.47, 0.72 years)
Period 2 to Period 1 Period 3 to Period 1 Period 4 to Period 1
Period 2: Jan1990-
Dec1994
Period1: Jul1984-
Dec1989
Period 3: Jan1995-
Jun1998
Period1: Jul1984-
Dec1989
Period 4: Jul1998-
Dec2003
Period1: Jul1984-
Dec1989
Relative Times Cox, Chu, …, Muñoz. Stat Med 2007
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
33Cumulative Percent Deceased
Rel
ativ
e Ti
me
(Deciles for Period 1 : 0.23, 0.47, 0.72 years)
Period 2 to Period 1 Period 3 to Period 1 Period 4 to Period 1
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
5
10
15
1
2
3
4
10 20 30
Period 2: Jan1990-
Dec1994
Period1: Jul1984-
Dec1989
Period 3: Jan1995-
Jun1998
Period1: Jul1984-
Dec1989
Period 4: Jul1998-
Dec2003
Period1: Jul1984-
Dec1989
Relative Times Cox, Chu, …, Muñoz. Stat Med 2007
34
Software for the Doerswww.statepi.jhsph.edu/software
• STATA : streg
anc
and anc2
options allow general regression, not simply conventional
•SAS : PROC LIFEREG Conventional
PROC NLMIXED General
• S-plus: censorReg
nlminb
Cox, Chu, Schneider, Muñoz. 2007 Stat Med
35
Conclusions•
Parametric survival analysis based on the General gamma–
allows for different shapes of the hazard function from increasing to decreasing; from arc-type to bathtub.
–
provides a means to characterize the HISTORY of the survival after AIDS through the evolution of antiretroviral therapies.
–
enhances the power of the cohort design by linking epidemiology to public health via the assessment of population effectiveness.
–
directly estimates relative times, a measure of effect that facilitates wider communication of findings.
Reflecciones
con cara
al futuro4El SIDA como
enfermedad
cronica.
4La necesidad
imperativa
de ofrecer tratamiento
a individuos
infectados
con VIH e
indicados
a recibir
tratamiento
4Oportunidad
de un tratamiento
de tener efectos
de prevencion
pero
al mismo
tiempo
hacer
mas
dificil
elucidacion
de eficacia
de vacunas.
4El interes
de caracterizar
el envejecimiento con tratamiento
prolongado
contra el VIH.
TROVASPor
Sigifredo
Ospina;1991
Vinieron
de todas
partesdizque
a darsen
la buena
vidapero
resultaron
metidosen la epidemiologia
del sida
Se escogio
muy
bien
el sitiocon disenos
especialeses
un lugar
excelentepa' retiros
espirituales
Vinieron
las
presentacionesmucha
gente
conocidaprimer factor de riesgoen la epidemia
del sida
Empezo
pues
la academiano hubo
ningun
debatetodo
iba
muy
biencon carrasquilla
y alzate
CIAT, Palmira
SOspina, FdelaOz, VEspitia, LMGomez, EPMuñoz, AConcha, RCorral, JRodriguez, RRodriguez, FDelgado, RCarvajal, JMuñoz, MIGutierrez, MConcha,
Incidencia
y Prevalencia
El profe
antes de hablarde un poco
de garabatoslogro
mariar
todo
el grupomoviendo
los acetatos
Si era poco
importanteno pudimos
sabero era tan importanteque
no lo dejaba
ver
Pero
de todas
manerashabia
cosas
muy
raraslas
variables de reinaldoeran
muchisimo
mas
claras
Y a alguien
le dieron
titulolo bautizaron
el biologofrente
al microcomputadorya
quedaba
microbiologo
Para promover
la BUSQUEDA
Oh, Oh
Reinaldo
Carvajal
“no tiene
sentido
biologico”
Y llego
el segundo
diamuñoz
comenzo
primerono hubo
ni
una
preguntaquedamos
viendo
un chispero
Vino
un cambio
refrescantepues
todo
estaba
al revesqueriamos
que
nos
hablaran
claroy nos
hablaron
en ingles
Preguntaron
quien
no entiendey alguien
dijo
con enojosyo
le entiendo
muy
bien
al profepero
no le entiendo
a estos
patojos
El trabajo
se hizo
en grupostrabajamos
en tripletay por
mas
que
lo intentamosno pudimos
encontrar
el beta
Ken Nelson, el otro
“blanco”
Yours truly or “el paisa”
Estudiantes
preguntando
en ingles de juanchito
β
Las presentaciones
de sidano fueron
cosa
cualquieracon la primera
presentacionse armo
pues
la garrotera
Que
es
culpa del ministerioque
problemas
de tiempo
atrasmentiras
que
es
problemadel que
recibe
por
detras
Hablaron
bastante
pajaatacaron
sin compasiony a proposito
de pajaes
un factor de proteccion
Y en los casos
de sida
en Valleuna
verdad
muy
duradizque
no hay ni
un solo casoen el municipio
de Buenaventura
OR= 4.3
OR= 0.67
Mucha
paja?
Nadie
podia creeralguien
sabe
que
paso?alguien
tuvo
la respuestael colera
los mato
Chucho
alcanzo
la estrellay tomo
la iniciativanos
puso
a ver
en egretla epidemiologia
del sida
Hablaron
tambien
muy
claroy sin muchos
disimulosnos
entregan
la tareacasi
nos
vamos
de culos
Y ya
para
terminaryo
quiero
decir
adiosespero
volverlos
a veren el proximo
llamado
de dios.
Riesgos
Competitivos
Chucho
Rodriguez
A los 25 años
de la maestria en epidemiologia, UniValle:
2008.450