Supplementary Appendix for the article
This material supplements but does not replace the content of the peer-reviewed paper published
in AIDS.
INTRODUCTION
The analyses reported in the main manuscript utilize the Hepatitis C Cost-Effectiveness
(HEP-CE) model of screening, care, and treatment for chronic (or acute) hepatitis C virus (HCV)
infection. HEP-CE is a detailed micro-simulation of HCV epidemiology, screening, linkage to
care, treatment, and outcomes. The HEP-CE model tracks clinical outcomes including number of
detected HCV antibodies, number linked to HCV care, number of cases of chronic HCV
identified, number of HCV-infected people initiating therapy, number achieving sustained viral
response (SVR), HCV-related mortality, quality-of-life (QoL), undiscounted life expectancy, and
discounted quality adjusted life expectancy (QALE). In addition, the model generates estimates
of discounted, lifetime medical costs from the health system perspective.
To incorporate outcomes and costs associated with HIV-infection, we used the CostEffectiveness of Preventing AIDS Complications (CEPAC) model, a computer simulation of the
clinical management and economics of HIV infection, to provide estimates of mortality, costs
and quality-of-life in a cohort of HIV-infected individualsreceiving U.S. guideline-concordant
HIV care [1,2]. The CEPAC model generated estimates of HIV-related, as well as non-HIV/nonHCV related mortality (competing risk mortality), QoL, and medical costs in the simulated
cohort of HIV-infected individuals. These estimates reflected the average experience of HIVinfected individuals without HCV infection. The HEP-CE model then used the CEPAC output as
1
model inputs and simulated HCV-related mortality, QoL, and costs.We used a standard
multiplicative assumption to estimate utility functions for joint health states reflecting both HIV
and HCV disease.
CEPAC HIV disease model
The CEPAC Model is a Monte Carlo simulation of the progression and outcomes of HIV
disease in a hypotheticalcohort of patients. The model follows each individual’s clinical course
from the time of entry into the model until death and keeps a running tally of clinical events, the
length of time spent in each health state, costs, and QoL. Upon the patient’s death, summary
statistics arerecorded and a new patient enters the model. This process is repeated over a
largenumber of patients (statistical convergence can typically be achieved with cohort sizes of2
to 5 million), at which point overall performance measures such as average projected
lifeexpectancy, quality-adjusted life expectancy, and cost are computed.
The CEPAC model definesthree general categories of health states: chronic HIVinfection, acute AIDS-related events, and death. Patients generally reside in the chronic HIVinfection health state. In the absence of antiretroviral therapy (ART),HIV disease progresses,
with characteristic immunesystem deterioration (CD4 count decline). The model treats HIV
RNA as the primarydriver of immune system deterioration, with current RNA
leveldeterminingthe rate of CD4 decline.
Incidence of AIDS-related events is a function of CD4 count and prior AIDS-related
events. Patients who develop an acute AIDS-related complication temporarily move to an acute
healthstate, where quality-of-life is lower and both resource consumption and mortalityare
2
higher. Deaths mayoccur while in a chronic or an acute state and can beattributed to a particular
opportunistic infection (OI), chronic AIDS, or non-AIDS-related causes.
The model includes 6 lines of ART modeling efficacy as the probability of achieving an
HIV RNA less than the level of detection typically by week 24 of therapy (Supplemental Table
1). With suppressed HIV RNA, CD4 count rises as a function of time on ART, with an initial
rapid rise, followed by two periods of progressively slower CD4 count change. Individuals on
suppressive ART face a monthly risk of experiencing virologic breakthrough related to poor
adherence. With rising HIV RNA, CD4 rapidly falls to the individual’s CD4 count nadir prior to
initiating therapy. The model distinguishes between the “true” time of virologic failure, and the
time that such failure is clinically identified. With two successive office visits documenting
elevated HIV RNA, the individual is advanced to the next line of ART.
Input data for the CEPACmodel are from published reports and public use datasets.
Model inputs have been described and published previously [1,3].
We used the CEPAC model to project life expectancy, quality of life, and medical costs
in a cohort of HIV-infected individuals with demographic characteristics matching those of the
cohort we planned to model for the cost-effectiveness analysis of HCV therapy (same mean age,
sex distribution, prevalence of injection drug use). CEPAC simulated the lifetime of each cohort
member, and provided output tables, stratified by sex and simulation month. For example, the
first table was a “lifetable” that provided the monthly probability of death from HIV-attributable,
as well as non-HIV and non-HCV related causes, conditional on being alive at the beginning of
the simulation month, stratified by sex and time in the simulation. This CEPAC output table
became an input table for the HEP-CE model, informing the monthly probability of death from
causes other than HCV. Similarly, CEPAC generated a table of the cohort’s average medical cost
3
in each simulation month, stratified by sex. HEP-CE used this “cost table” to inform non-HCV
related costs in every month. Finally, CEPAC generated a table of the cohort’s average quality of
life in every simulation month, again stratified by sex. HEP-CE used this “quality of life table” to
look up the cohort’s quality of life in each month, to which HEP-CE then applied utility weights
related to HCV infection.
HEP-CEMODEL
The HEP-CE model is a Monte Carlo simulation of HCV infection designed to estimate
the outcomes and costs associated with various strategies of screening, treatment, and improving
HCV care. The model includes 5 components: 1) HCV epidemiology, 2) HCV screening and
linkage to care, 3) HCV disease progression, 4) HCV therapy, and 5) Non-HCV mortality
(Supplemental Figure 1).
HCV epidemiology
The incidence of HCV is a model input. In each monthly cycle, individuals in the model
who are not HCV-infected face a risk of acquiring HCV that is a function of their age and HCV
risk behaviors (typically interpreted as current injection drug use versus no current use). Once
HCV-infected, individuals begin the simulation of HCV disease progression (see below). For the
purpose of this analysis, which simulates the progression of a group of chronically HIV/HCV coinfected individuals, all individuals were HCV-infected at the start of the simulation.
HCV disease progression
4
Chronic HCV is characterized in the model by 3 linear stages of fibrosis progression:
mild-to-moderate fibrosis, cirrhosis, and decompensated cirrhosis. At all stages of fibrosis, HCVinfection is associated with increased costs and decreased QoL. Mortality attributable to HCV
infection does not occur until an individual reaches his/her unique time of developing cirrhosis.
Once an individual reaches the cirrhosis stage, (s)he faces a probability of liver-related mortality
that remains constant unless s(he) initiates treatment and attains SVR.
To estimate critical disease progression event times, including the time at which an
individual develops cirrhosis (tcirrhosis) and decompensated cirrhosis (tdecomp), the model uses a
stochastic process to draw 4 critical parameter values: 1) ageat HCV infection (AgeHCV); 2) age
at simulation base line (Agebase); 3) months from infection to cirrhosis (mcirrhosis); and 4) months
from cirrhosis to first liver decompensation event (mdecomp).
For those with prevalent HCV infection at simulation baseline:
tcirrhosis = mcirrhosis – (Agebase - AgeHCV)
tdecomp = tcirrhosis + mdecomp
For those who contract incident HCV during the simulation:
tcirrhosis = mcirrhosis
tdecomp = tcirrhosis + mdecomp
As a result of the stochastic process that defines time to cirrhosis and decompensation,
not all HCV-infected patients in the model develop cirrhosis, and not all patients with cirrhosis
die of HCV-attributable causes.
5
With spontaneous clearance or successful HCV treatment, disease progression ceases,
and QoL, costs, and mortality risk return to that of HCV-uninfected individuals of the same age,
sex, and HCV risk behaviors. Re-infected individuals resume the course of fibrosis progression
at the stage reached during prior infection.
HCV screening and linkage to care
The model distinguishes between an individual’s true infection status and clinical
awareness of infection. Only individuals with identified HCV infection are eligible for HCV
therapy. The model provides flexibility to define various strategies for screening based on
different screening modalities and intervals. All screening tests are characterized by a cost,
sensitivity, and specificity. Test characteristics are a function of time since infection. The
analysis of HCV therapy in HIV/HCV co-infection did not utilize the HCV screening features of
the model as this cohort has already been identified as chronically infected with HCV.
HCV Therapy
The model provides flexibility to simulate a variety of algorithms for HCV therapy,
including distinct treatment protocols for acute and chronic HCV infection, as well as for
genotype 1 vs. genotype 2/3 infection. Treatment efficacy is modeled using 6 key parameters,
each of which is stratified by cirrhosis status:
1. Probability of developing HCV RNA >1,000 copies/ml = Pfail
2. Probability of attaining HCV RNA < assay at treatment week W4 = Prvr
6
3. Probability of attaining HCV RNA < assay at treatment week W12 = Pervr
4. Probability of attaining HCV RNA< assay 24 weeks after therapy end = P SVR
5. Probability of withdrawing from therapy for any cause = Pwithdraw
6. Proportion of all therapy withdraws that are attributable to major treatment toxicity =
Ptox
The user can define an algorithmic response to events such as virologic failure, attaining
both 4 and 12 week HCV RNA < assay (extended rapid virologic response), or having HCV
RNA detectable, but less than 1,000 copies/ml. For example, when modeling response guided
therapy using peginterferon, ribavirin, and an HCV protease inhibitor, having an HCV RNA >
1,000 copies/ml leads to treatment cessation, having detectable HCV RNA that is <1,000
copies/ml at either treatment week 4 or 12 leads to the course of therapy being extended from 24
to 48 weeks. Having undetectable HCV RNA at both weeks 4 and 12 leads to a 24-week total
treatment course.
In every month that a simulated individual is taking HCV therapy, (s)he accrues costs
associated with physician visits, as well as the cost of HCV medications and additional
medications used to manage non-treatment ending toxicities such as mild to moderate anemia,
neutropenia, and rash (Supplemental Table 2). The approach to modeling non-treatment ending
toxicity is to determine the cost of managing a given complication, as well as an estimate of the
proportion of patients in a given month who have a specific non-treatment ending toxicity type.
The “average” cost of HCV therapy is taken as the sum of all antiviral medications, plus the
weighted average of non-treatment ending toxicity costs, using the proportion of each toxicity
type in the cohort as the weight.
7
In every month while on HCV treatment, individuals face a probability of withdraw for
toxicity, inability to tolerate adverse effects of interferon, or non-adherence. Among those who
withdraw for any reason, the model determines whether that withdraw from therapy was the
result of treatment-ending toxicity. Those who experience treatment-ending toxicity have a onemonth decrement in quality of life related to the toxicity, as well as a one-time cost associated
with managing the toxicity (Supplemental Table 3). Individuals who stop treatment due to
toxicity accrue a cost that reflects the cost of an “average” toxicity event. This cost is calculated
by identifying the most common toxicity event types, estimating the cost of managing each type
of toxicity event, and taking the weighted average cost of all toxicity event types using the
proportion of toxicity events attributable to each toxicity type as the weight.
Non-HCV morbidity and mortality
In addition to the HCV-attributable causes of death, in every month individuals in the
model are exposed to risk of death from causes other than HCV, as well as costs associated with
non-HCV related healthcare. Competing risk mortality is stratified by age and sex, and is
informed by U.S. vital statistics data [4]. Competing risk mortality is adjusted using standardized
mortality ratios, to reflect elevated competing risks of death in persons with a history of HCV
risk-behaviors such as IDU [5].
Costs
8
The model generates estimates of discounted, lifetime medical costs from the health
system perspective. Component costs include those of HCV-related physician and nursing visits,
HCV medications, medications used for management of treatment toxicity, and the cost of
additional medical visits and hospitalizations associated with HCV-infection. Costs include
estimates of age and sex stratified “background” costs, which reflect healthcare utilization
unrelated to HCV.
Quality of life
Quality of life in HEP-CE is a function of HCV fibrosis stage. To simulate HIV/HCV coinfection for this analysis, the HEP-CE model integrated quality of life functions for both HIV
and HCV-infections. The CEPAC model provided an actuarial table of the mean QoL of the
cohort in each simulation month. HEP-CE combined the CEPAC estimated of average HIVassociated QoL, with HCV health state-specific estimates of QoL, using a multiplicative
assumption[6-10].
9
Supplemental Table 1. HIV drug costs and efficacies
Regimen
Base case
Range
Reference(s)
Efficacy
86.0%,
73.1%-98.9%
[11]
Monthly cost
$1,500
$800-$2,300
[12]
Efficacy
73.3%,
62.3%-84.3%
[13]
Monthly cost
$2,100
$1,100-$3,200
[12]
Efficacy
61.3%,
52.1%-70.5%
[13]
Monthly cost
$2,000
$1,000-$3,000
[12]
Efficacy
64.5%,
54.8%-74.2%
[14]
Monthly cost
$2,800
$1,400-$4,300
[12]
Efficacy
40.0%
34.0%-46.0%
[15,16]
Monthly cost
$4,800
$2,400-$7,200
[12]
First line
Second line
Third line
Fourth line
Fifth line
10
Regimen
Base case
Range
Reference(s)
Efficacy
15.0%
12.8%-17.3%
[15]
Monthly cost
$1,900
$1,000-$2,900
[12]
Sixth line
11
000251659264251660288251661312251662336Supplemental Figure 1. HEP-CE model flow
diagram
Individual enters HEP-CE model
Random numbers assign unique values for:






Age
Sex
HCV genotype
Age of infection
Time to cirrhosis
Time from cirrhosis to first
decompensation event
No
Does the individual complete therapy?



HCV disease progression
No chance for future
therapy
Chance of death from
competing mortality
Yes
No
Does the individual achieve SVR?
Yes


HCV uninfected
Chance of death from competing
mortality
12
Supplemental Table 2. HCV therapy costs (2011 US $)
Drug
Standard dose
Frequency
Cost per month
Reference(s)
Standard HCV treatment a,b
1200 mg
1/day
$1,371
[17,18]
Anemia (dose reduction) c
600 mg
1/day
$685
[17]
Standard HCV treatment d
180 mcg/ml
1/week
$2,097
[17,18]
Neutropenia (dose reduction)
135 mcg/ml
1/week
$1,572
[17]
750 mg
3/day
$15,154
[18]
300 mcg/ml
2/week
$1,900
[17], Expert opinion
150 g
1/month
$160
[17,18], Expert opinion
Ribavirin
Peginterferon alfa-2a
Telaprevir
Standard HCV treatment e
Filgrastim
Neutropeniad
Clobetasol propionate
Rash f
13
Note: the monthly cost estimate is based on the average wholesale price[12], less 23%[19],
assuming a standard dosage and number of pills for 1 month.12.0%[20] incur an additional cost
of 19.75[21]for a nurse visit to treat adverse event.
a
18.0% receive ribavirin dose reduction to 600 mg/day to treat anemia (months 1-3 only)
b
8.5% receive ribavirin dose reduction to 600 mg/day to treat anemia (months 4-12 only)
c
8.5% receive ribavirin dose reduction to 600 mg/day to treat anemia (all months)
d
6.5% receive peginterferon alfa-2a dose reduction from 180 to 135 mcg/ml to treat
neutropeniaand 300 mg of filgastrim twice per week (all months)
e
For those on triple therapy(months 1-3 only)
f
28.0%receive 150 g of 0.05% clobetasolpropionate to treat rash (months 1-3 only)
14
Supplemental Table 3.Costs of treatment ending toxicity events (2011 US $)
Resource utilized
CPT code
Unit cost
# Utilized
Total cost
Source(s)
Anemia
Epoteinalfa
N/A
485.10
1
485.10
[12,22], Expert opinion
Complete blood count
85025
10.94
3
32.82
[22,23], Expert opinion
Bilirubin
82247
7.06
1
7.06
[22,23], Expert opinion
Uric acid
84550
6.36
1
6.36
[22,23], Expert opinion
Reticulocytes
85045
5.63
3
16.89
[22,23], Expert opinion
Transfusion
96365
54.37
1
54.37
[22,23], Expert opinion
Packed red blood cells
P9021
231.50
2
463.00
[22,24], Expert opinion
Clinic visit and injection
99211
19.74
2
39.48
[21,22], Expert opinion
Moderately complex clinic visit
99214
104.16
2
208.32
[21,22], Expert opinion
Proportion with anemia (dual
therapy)
__
__
__
0.176
[25]
Proportion with anemia (triple
therapy)
__
__
__
0.097
[20]
Total cost of anemia
__
__
__
$1,313.40
Rash
15
Resource utilized
Clobetasol propionate cream
(30g)
CPT code
Unit cost
# Utilized
Total cost
Source(s)
N/A
35.86
1
35.86
[12,22], Expert opinion
Simple clinic visit
99213
70.46
2
140.92
[21,22], Expert opinion
Dermatology consultation
99243
119.81
1
119.81
[21,22], Expert opinion
Proportion with rash (dual
therapy)
__
__
__
0
[25]
Proportion with rash (triple
therapy)
__
__
__
0.387
[20]
Total cost of rash
__
__
__
$296.59
Moderately complex clinic visit
99214
104.16
2
208.32
[21], Expert opinion
Complete blood count
85025
10.94
1
10.94
[23], Expert opinion
Liver function test
80069
12.22
1
12.22
[23], Expert opinion
Other SAE
Proportion with other SAE (dual
therapy)
__
__
__
0.824
[20,25]
Proportion with other SAE (triple
therapy)
__
__
__
0.516
[20]
Total cost of other SAE
__
__
__
$231.48
16
Resource utilized
CPT code
Unit cost
# Utilized
Total cost
Total cost of treatment endingtoxicity for dual therapya
$421.90
Total cost of treatment endingtoxicity for triple therapya
$361.36
Source(s)
CPT = current procedural terminology; SAE= Serious Adverse Event
a
Total costs equal weighted average of toxicity costs:
(Proportion with anemia*total cost of treating anemia+ proportion with rash*total cost of treating rash + proportion with other SAE *
total cost of treating other SAE)
17
Supplemental Table 4. Incremental cost-effectiveness ratios of projected cost-containing strategies utilizing IFN-free treatment
Strategy
% Attaining SVR
Undiscounted
Life Expectancy
Discounted
Cost ($)
QALY
Incremental
Cost ($)
QALY
CER ($/QALY)
No treatment
0
13.240
198,700
6.760
---
---
---
Dual therapy
30.8
13.761
221,900
7.599
23,200
0.839
27,700
PEG/RBV trial*
88.7
14.728
306,500
9.233
84,600
1.634
51,800
IL28B triage*
89.6
14.737
312,800
9.284
6,300
0.051
123,700
IFN-free therapy
90.3
14.745
325,300
9.323
12,500
0.039
322,200
HIV: human immunodeficiency virus; HCV: hepatitis Cvirus; IFN: interferon; QALY: quality-adjusted life year; CER:
cost-effectiveness ratio; PEG/RBV: Peginterferon and Ribavirin; IL28B: interleukin-28B
Note: all costs and QALYs are lifetime and discounted at an annual rate of 3%. Costs are in 2011 US dollars and rounded to the nearest $100. All
life-years and QALYs are rounded to the nearest thousandth
*These strategies utilize PEG/RBV and IFN-free therapy instead of PEG/RBV/TVR in the base case
18
REFERENCES
1.
Freedberg KA, Losina E, Weinstein MC, Paltiel AD, Cohen CJ, Seage GR, et al. The
cost effectiveness of combination antiretroviral therapy for HIV disease. N Engl J Med
2001,344:824-831.
2.
Paltiel AD, Weinstein MC, Kimmel AD, Seage GR, 3rd, Losina E, Zhang H, et al.
Expanded screening for HIV in the United States--an analysis of cost-effectiveness. N
Engl J Med 2005,352:586-595.
3.
Paltiel AD, Walensky RP, Schackman BR, Seage GR, 3rd, Mercincavage LM, Weinstein
MC, et al. Expanded HIV screening in the United States: effect on clinical outcomes,
HIV transmission, and costs. Ann Intern Med 2006,145:797-806.
4.
Centers for Disease Control and Prevention. United States Life Tables, from:
http://www.cdc.gov/nchs/products/life_tables.htm. 2008.
5.
Arendt M, Munk-Jorgensen P, Sher L, Jensen SO. Mortality among individuals with
cannabis, cocaine, amphetamine, MDMA, and opioid use disorders: a nationwide followup study of Danish substance users in treatment. Drug Alcohol Depend 2011,114:134139.
6.
Chong CA, Gulamhussein A, Heathcote EJ, Lilly L, Sherman M, Naglie G, et al. Healthstate utilities and quality of life in hepatitis C patients. Am J Gastroenterol 2003,98:630638.
7.
Grieve R, Roberts J, Wright M, Sweeting M, DeAngelis D, Rosenberg W, et al. Cost
effectiveness of interferon alpha or peginterferon alpha with ribavirin for histologically
mild chronic hepatitis C. Gut 2006,55:1332-1338.
19
8.
Paltiel AD, Scharfstein JA, Seage GR, 3rd, Losina E, Goldie SJ, Weinstein MC, et al. A
Monte Carlo simulation of advanced HIV disease: application to prevention of CMV
infection. Med Decis Making 1998,18:S93-105.
9.
Schackman BR, Goldie SJ, Freedberg KA, Losina E, Brazier J, Weinstein MC.
Comparison of health state utilities using community and patient preference weights
derived from a survey of patients with HIV/AIDS. Med Decis Making 2002,22:27-38.
10.
Stein K, Dalziel K, Walker A, McIntyre L, Jenkins B, Horne J, et al. Screening for
hepatitis C among injecting drug users and in genitourinary medicine clinics: systematic
reviews of effectiveness, modelling study and national survey of current practice. Health
Technol Assess 2002,6:1-122.
11.
Gallant JE, DeJesus E, Arribas JR, Pozniak AL, Gazzard B, Campo RE, et al. Tenofovir
DF, emtricitabine, and efavirenz vs. zidovudine, lamivudine, and efavirenz for HIV. N
Engl J Med 2006,354:251-260.
12.
Micromedex 2.0. Drug Topics Red Book Online. Accessed Feburary 1st, 2013. from:
http://www.micromedexsolutions.com.
13.
Johnson M, Grinsztejn B, Rodriguez C, Coco J, DeJesus E, Lazzarin A, et al. Atazanavir
plus ritonavir or saquinavir, and lopinavir/ritonavir in patients experiencing multiple
virological failures. AIDS 2005,19:685-694.
14.
Grinsztejn B, Nguyen B, Katlama C, Gatell J, Lazzarin A, Vittecog D, et al. Safety and
efficacy of the HIV-1 integrase inhibitor raltegravir (MK-0518) in treatment-experienced
patients with multidrug-resistant virus: a phase II randomised controlled trial. Lancet
2007,369:1261-1269.
20
15.
Nelson M, Arasteh K, Clotet B, Cooper DA, Henry K, Katlama C, et al. Durable Efficacy
of Enfuvirtide Over 48 Weeks in Heavily Treatment-Experienced HIV-1–Infected
Patients in the T-20 Versus Optimized Background Regimen Only 1 and 2 Clinical
Trials. JAIDS 2005,40:404-412.
16.
Lalezeri J, Goodrich J, De Jesus E, Lampiris H, Gulick R, Saag M. Efficacy and safety of
maraviroc plus optimized background therapy in viremic ART experienced patients
infected with CCR5-tropic HIV-1: 24-week results of a phase 2b/3 study in the US and
Canada.
17.
Physician's Desk Reference. from: http://www.pdr.net. 2011.
18.
Food and Drug Administration. Telaprevir package insert. from: http://www.fda.gov.
2012.
19.
Levinson D. Medicaid drug price comparisons: Average manufacturer price to published
prices; Department of Health and Human Services. from:
http://www.oig.hhs.gov/oei/reports/oei-05-05-00240.pdf. 2005.
20.
McHutchison JG, Lawitz EJ, Shiffman ML, Muir AJ, Galler GW, McCone J, et al.
Peginterferon alfa-2b or alfa-2a with ribavirin for treatment of hepatitis C infection. N
Engl J Med 2009,361:580-593.
21.
United States Department of Health and Human Services Center for Medicare Services.
Physician Fee Schedule. Accessed January 15th, 2013. from:
http://www.cms.gov/Medicare/Medicare-Fee-for-ServicePayment/PhysicianFeeSched/index.html?redirect=/PhysicianFeeSched. 2011.
22.
Gao B. Interaction of alcohol and hepatitis viral proteins: implication in synergistic effect
of alcohol drinking and viral hepatitis on liver injury. Alcohol 2002,27:69-72.
21
23.
United States Department of Health and Human Services Center for Medicare Services.
Clinical Diagnostic Laboratory Fee Schedule. Accessed February 1st, 2013. From:
http://www.cms.gov/ClinicalLabFeesched/02_clinlab.asp. 2011.
24.
Whitaker B, Sullivan M. The 2005 Nationwide Blood Collection and Utilization Survey
Report.
25.
Torriani FJ, Rodriguez-Torres M, Rockstroh JK, Lissen E, Gonzalez-Garcia J, Lazzarin
A, et al. Peginterferon Alfa-2a plus ribavirin for chronic hepatitis C virus infection in
HIV-infected patients. N Engl J Med 2004,351:438-450.
22