୯ҥᆵεᏢϦӅፁғᏢଣ
ࢬՉੰᏢᆶႣٛᙴᏢࣴز܌ғނᙴᏢीಔ റγፕЎ
Division of Biostatistics
Graduate Institute of Epidemiology and Preventive Medicine College of Public Health
National Taiwan University Doctoral Dissertation
ᒿᐒၸำᔈҔܭႣٛЃߎහМੱ Hoehn-Yahr ϩᜪ ੯ੰϐჴຑ
Evidence-based Evaluation of Preventing Progression of Hoehn-Yahr-stage-based Parkinson’s Disease with
Stochastic Process
㵍ᛏ
Chiung-Jung Wen
ࡰᏤ௲Ǻ ഋذᅚ റγ Supervisor: Hsiu-Hsi Chen, Ph.D
ύ҇୯ 104ԃ 01 Д Jan, 2015
ी ी ी ी ी ी ी ी ीಔ ಔ ಔ ಔ ಔ ಔ ಔ ಔ
i
ᇞᖴ!
ૈֹԋፕЎǴനाགᖴޑࢂഋذᅚ௲ӭԃٰޑࡰᏤǴذᅚԴৣόӧךၶډ֚
ᜤ܈ᓍਔǴਔޑЇሦךှ،ୢᚒǴӧךЈҞύ׳ࢂന٫ޑᏢೌࣴزڂጄǶགᖴ
ቅֻ፵௲ӧԛઓϣࡰᏤǴаϷୖᆶઓᏢᑔᔠޑϡύᙴৣǵഋԿӄᙴ
ৣǵྕዝᆺᙴৣǵഋၲϻᙴৣǶᖴᖴα၂ہ୯ਕ௲ǵำ௲ǵዐߞؼ௲
ǵЦහቺ௲ǵᝄܴޱ௲๏ϒךӭᝊޑཀـǶ!
!
གᖴذᅚԴৣکࣴزი໗ᓉൟԴৣǵܴޱԴৣǵДཨԴৣǵҥܹԴৣӧፕЎޑी
کኗቪ๏ϒࡐӭޑࡰᏤǴᡣ೭ҽፕЎૈճֹԋаϷว߄Ƕ!
གᖴୖᆶ୷ໜፄӝԄᑔᔠϷၗԏޑ܌ԖπբΓǶགᖴךޑӕᏢॺکךଆࡋ
ၸࣴز܌ޑਔӀǶ!
!
നࡕǴགᖴךޑৎΓ೭ࢤਔ໔܌๏ךޑЍᆶႴᓰǴᜫգॺᆶךଆϩ٦೭ҽ
ޑ഻৹Ǽ!
! !
ৣ
ৣ
ৣό
ৣό
ৣόӧӧӧӧӧӧӧӧךၶךၶךၶךၶךၶךၶךၶךၶډ֚ډ֚ډ֚ډ֚ډ֚ډ֚ډ֚ډ֚֚
ೌࣴࣴࣴࣴࣴزڂزڂزڂزڂزڂزڂጄǶڂጄǶጄጄጄጄጄǶጄǶགᖴགᖴགᖴགᖴགᖴགགᖴ
ii
ύЎᄔा
ࣴ
ࣴزङඳ!!ЃߎහМੱࢂಃΒதـޑଏϯ܄੯ੰǴനಖᏤठيᡏфૈΠफ़аϷ
෧ϿტڮǶϷԐݯᕍёаۯ੯ੰޑаϷۯߏӸࢲਔ໔ǴࡺϷԐບᘐϷݯᕍ
วᡉளख़ाǶՠਥᏵЃߎහМੱޑ౦፦܄ࡌᄬځ੯ੰԾฅў٠ԐයບᘐЃ
ߎහМੱޑਏޑࣴزϝࡐϿـǶӢԜǴҁጇፕЎޑࣴزҞޑЬाԖΟ;!2/ճҔ
ঁаޗࣁ୷ᘵޑဂǴКၨЬୀෳᆶୀෳЃߎහМੱޑਏ!3/ࡌҥ
а Ipfio.Zbis!ϩᜪ੯ੰ໘ࢤ)ᙁᆀ I.Z!tubhf*ࣁ୷ᘵޑЃߎහМੱ੯ੰԾฅ
ўǴ٠Ъஒёૈቹៜ੯ੰԾฅўᙯ౽ೲޑӢηуΕኳԄύ!4/ՉԐයୀෳЃ
ߎහМੱޑԋҁਏϩǶ!
ᆶБݤ!!ҁࣴزၗٰྍࣁ 3112 ԃ୷ໜޗЃߎහМੱᑔᔠޑၗǶಃ
ϩࣴزӃКၨٿᅿୀෳЃߎහМੱޑБԄ)Ьୀෳᆶୀෳ*ޑಕᑈୀෳ
Ϸ I.Z!tubhf ޑϩѲǶಃΒϩࣴز߾ճҔߚྗϯޑੰٯ.Шжीܭа I.Z!
tubhf ࣁ୷ᘵޑЃߎහМੱ੯ੰԾฅўǶךॺճҔΟ໘ࢤϷϖ໘ࢤޑଭёϻኳԄ
ѐࡌᄬа I.Z!tubhf ࣁ୷ᘵޑЃߎහМੱ੯ੰԾฅўǶךॺஒёૈቹៜ੯ੰԾฅ
ўޑӢηӵ୷ҁၗǵғࢲಞᄍаϷ१ಞᄍΨԵቾܭ੯ੰԾฅўύǶനࡕǴך
ॺճҔ܌ࡌᄬϖ໘ࢤޑଭёϻኳԄǴኳᔕ 71 ྃаޑࣴزШж!-ӧଓᙫ 31 ԃࡕ
ӧόӕᑔᔠౣΠ܌ளϐЃߎහМੱᑔᔠޑԋҁਏϷԋҁਏҔϩǶ੯ੰ
ޑୖኧளԾಃΒϩளԾჴၗ܌ीޑԾฅੰўǴԋҁϷਏ֡ש 4&Ǵӧ
ኳᔕ 2111 ঁኬҁελࣁ 21111 ΓޑШжဂ܌ளӧόӕᑔᔠౣΠޑಕᑈԋҁ ᡏф
ᡏф ᡏф ᡏф
ᡏфффффૈΠૈΠૈΠૈΠૈΠૈΠૈΠૈΠफ़аफ़аफ़аफ़аफ़аफ़аफ़аफ़аϷϷϷϷϷϷϷϷ ϷԐԐԐԐԐບᘐບᘐບᘐບᘐບᘐບᘐᘐϷݯϷݯϷݯϷݯݯݯݯᕍᕍᕍᕍᕍ
iii
ਏ0ਏҔࡕǴϩձ٩ӭቚуԃϷӭቚуғࢲࠔ፦ਠ҅ԃ܌ሡቚуޑԋ
ҁǵӧόӕޑбཀᜫΠޑԋҁਏ0ਏҔௗڙԔጕǵϷԋҁ.ਏ0ਏҔණթკ
ՉКၨǶ!
่
่݀! Ьୀෳکୀෳϩձୀෳр 2:3 ঁаϷ 9: ঁЃߎහМੱੰΓǶЬ
ୀෳКୀෳऊёפр 2/9 ७ӭ):6&DJ;!2/5.3/4*ޑЃߎහМੱੰΓǴ٠Ъӧ
ЬୀෳಔύԐය)I.Z!tubhf!J-JJ*ޑঁਢ܌՞ޑКٯᡉӭܭୀෳಔ
)91/5&!wt/!72/6&-!q>1/15*!Ƕ!
! ่݀ᡉҢ੯ੰԾฅўӧΟ໘ࢤޑଭёϻኳԄύǴঁ 71 ྃаޑΓวғЃ
ߎහМੱޑԃวғࣁίϩϐ 9/3ǶவёୀෳයԿᖏයޑᙯ౽ 1/6:46!
):6&!DJ;!1/5441.1/8652*Ǵջځѳ֡ᅉ੮යࣁ 2/79 ԃǶӧϖ໘ࢤޑଭёϻኳԄ
ύǴวғЃߎහМੱޑԃวғࣁίϩϐ 8/9ǶவёୀෳයޑԐයډఁය)I.Z!
tubhf!JJJ,*ޑԃᙯ౽ೲࣁ 1/35:9!):6&!DJ;1/2531.1/4687*<வёୀෳޑԐය
ډᖏԐයޑԃᙯ౽ೲࣁ 1/4:93):6&!DJ;!1/3675.1/64::*Ƕவёୀෳఁයډ
ᖏఁයޑԃᙯ౽ೲ 3/2338):6&!DJ;!1/621:.4/8457*ǶԵໆόӕޑ܄ჹϖ
໘ࢤኳԄύӚރᄊ໔ᙯ౽ೲޑቹៜǴ่݀ᡉҢቚу 21 ྃǴวғЃߎහМੱ
ޑ࣬ჹӒᓀ܄ࣁ 2/8: ७!):6&!DJ;!2/43.3/55*!ЪуೲᑔᔠёୀෳԐයډё
ୀෳ㴣යޑೲǴځ࣬ჹӒᓀ܄ࣁ 6/19 ७):6&!DJ;!2/:5.24/3:*Ƕեֿለᐚࡋ
ӕኬࣁЃߎහМੱޑӒᓀӢηǴځ࣬ჹӒᓀ܄ࣁ 2/65 ७Ȑ:6&!DJ;!2/15.
ԃ܌ሡ܌ሡ܌ሡ܌ሡ܌ሡቚуቚуቚуቚуቚуቚуቚуቚуቚуޑԋޑԋޑԋޑԋޑԋޑԋޑԋޑԋޑԋ
00ਏਏҔਏҔਏҔਏҔਏҔਏҔਏҔਏҔණթණթණթණթණթණթණթණթკკკკკკკკ
iv
3/39*Ƕ௲ػำࡋၨଯޣځԾᑔᔠёୀෳԐයډఁයޑೲҭၨଯ!)SS>25/76-!
:6&!DJ;!3/:5.65/64*Ƕ!
ਥᏵኳᔕޑၗ܌ளډόӕᑔᔠ໔႖ޑٰ่݀ຑᑔᔠޑਏύǴԃᑔᔠ
ёаफ़ե 82&!):6&DJ;!76.91&*ԿఁයЃߎහМੱޑԭϩКǴऩᑔᔠ໔႖ϩ
ձࣁٿԃǵΟԃǵѤԃ܈ϤԃǴ߾ёаफ़եԿఁයЃߎහМੱޑԭϩКϩձ
ࣁ 65&!):6&!DJ;!56.73&*ǵ54&!):6&DJ;!43.63&*ǵ46&!):6&DJ;!34.56&*ǵаϷ
36&!):6&!DJ;23.47&*Ƕ!
җ،ۓ܄ኳԄ܌ளԋҁਏϷԋҁਏҔϩޑ่݀ϩձࣁቚуΓԃሡ
ӭ%227: ډ%2915<ቚуғࢲࠔ፦ਠ҅ԃሡӭ%2826 ډ%3717Ƕӧቚу
Γԃϐбཀᜫ%31-111 ΠǴԃᑔᔠёளډനεޑృਏ)%391-798*Ǵځԛࣁ
Βԃԛ)%391-622*ϷΟԃԛ)%391-527*کคᑔᔠȐ%391-224*Ǵ࣬ӕޑ
ᖿ༈ҭϸࢀӧағࢲࠔ፦ਠ҅ԃޑృਏ่݀Ƕ!!
!!!!!ᐒ܄ԋҁਏϩޑ่݀ᡉҢбཀᜫ%31-111 ΠǴӧୖᆶࣁ 211&Ϸ
71&Πڀԋҁਏޑᐒࣁ 7:.8:&Ϸ 75.85&<࣬ჹచҹΠڀԋҁਏҔޑᐒϩձ
ࣁ 73/7&.81/3&ک 69/3&.73/7&Ƕ!
่ፕ!!ЬୀෳКୀෳޑБԄёୀෳр൳Яٿ७ޑঁਢǴ٠Ъёа෧Ͽ 5:&
ޑఁයЃߎහМޑੰΓǶ่݀ᡉҢǴঁ 71 ྃаޑΓѳ֡வᑔᔠёୀෳය
ډᖏයޑਔ໔ऊࣁԃъǶᑔᔠޑ໔႖ຫஏǴёа෧ϿԿ I.Z!tubhf
ఁයޑКຫεǶ٩ I.Z ੯ੰރᄊѐࡌᄬੰΓࢂឦܭᑔᔠёୀෳޑԐයډᖏ )S
)S )S )S
)SSSSS>S>252525252525252525/7/7/7/76777776-6-6-66-6-6-6-!
ύύύύύǴǴǴǴǴǴԃᑔԃᑔԃᑔԃᑔԃᑔԃᑔԃᑔԃᑔԃᑔᔠᔠᔠᔠᔠᔠ
v
ය܈җᑔᔠёୀෳޑఁයډᖏයޑ੯ੰԾฅўኳԄǴёගٮЃߎහМੱ
Ԑයୀෳޑޜ໔ǶᔈҔ೭٤ރᄊᙯ౽ޑୖኧǴҁፕЎᇥܴΑཇஏޑᑔᔠ໔
႖ёफ़եఁයЃߎහМੱޑКٯཇεǴڀԋҁਏޑᐒΨཇεǶ!
!
!
ᜢᗖӷ;ЃߎහМੱǵԐයᑔᔠǵԋҁਏǵIpfio.Zbis ϩᜪ!
ගٮٮٮٮٮЃߎЃߎЃߎЃߎЃߎߎߎߎߎහМහМහМහМහМහМහМහМහМੱੱੱੱੱੱੱੱੱ
ཇ ཇ ཇஏ ཇஏ
ཇஏޑޑޑޑޑޑޑޑᑔᔠᑔᔠᑔᔠᑔᔠᑔᔠᑔᔠᑔᔠᑔᔠ໔໔໔໔໔໔໔໔ Ƕ
vi
Abstract
Background Parkinson’s disease (PD) is the second most common degenerative
disorder which will eventually cause functional decline and reduce lifespan. The
development of therapies that slow disease progression and improve survival makes
early detection and treatment of PD especially important. Besides, the characteristics of
heterogeneity in natural history and the uncertainty in the decision analysis of early
detection of PD prevention have not been fully investigated. The aims of this thesis
consist of three parts: (1) the first was to to use a community-based cohort to compare
the detection methods for active detecting PD. (2) the second was to elucidate the
temporal natural history of Hoehn-Yahr-stage-based PD with a Markov process with
and without the incorporation of covariates into different transitions corresponding to
the natural history model and the third part was to evaluate the cost-effectiveness
analysis.
Material and Method First part of data were derived from a community-based
screening survey for PD in 2001. Cumulative detection rate and Hoehn-Yahr (H-Y)
stage distribution of both the active and passive detection groups were estimated and
compared.
In the second part, we use a non-standard case-cohort design for modelling the
natural history of H- Y stage-base PD. We built a three-state and a five-state Markov g
g g g
geneneenenerererererererrratatatatatatatativivivivivive vve eeee
span.nnnn.TTTTTheheheheh
vii
models for the H-Y stage-based natural history. Variables such as baseline characteristic,
life style and dietary habit were collected and were incorporated into the model to
assess the effect of each covariate on respective transitions.
In the final part, the Markov decision analysis was envisaged to estimate the cost-
effectiveness and cost-utility of active screening for PD in the community setting for
residents aged 60 years or older over a 20-year period. We used a five-state Markov
model to simulate the progression of PD and the sequel afterwards. The cumulative cost
under different strategies was also collected. Parameters of disease progression followed
the empirical estimates of the temporal natural history in the second Part. The main
outcome measure was cost per life-year gain and per quality-adjusted life-year (QALY)
gained with a 3% annual discount rate. The scattered cost-effectiveness plane (CE
plane) and acceptability curve was presented given a 1000 Monte Carlo simulated
samples for running 10,000 trials.
Results One hundred and ninty-two IPD cases and 89 IPD were detected by the active
and passive detection methods, respectively. The active method detected approximately
1.8-fold (95% confidence interval: 1.4-2.3) the IPD cases of the passive method. Early
H-Y stage (stage I and II) IPD cases were statistically significantly higher in the active
method than in the passive method (80.4% vs. 61.5%, p=0.04).
inee e e e chchchchcharararararararrracacacacacacacacactteteteteteteteteririririririririristststststststststicicic, ,
h h h h
he e e e emmomomomomomomomodedededededededelllllllltotototototototo
viii
Base on a three-state homogeneous Markov model, annual incidence rate of being
susceptible to PD for subjects aged 60 years or older was 8.2 per 1000 person-years.
Annual transition rate from screening detectable (SD) phase to clinical detectable (CD)
phase was 0.5935 (95% CI: 0.4330-0.7541), which yielded 1.68 years of mean sojourn
time staying in the SD phase. In a five-state homogeneous Markov model, the estimate
incidence of SD phase PD was similar to that estimated from the three-state model, 7.8
per 1000. The transition rate from H-Y I/II to H-Y III+ in the SD phase was 0.2498
(95% CI: 0.1420-0.3576). The transition rates from SD to CD for early stage (H-Y I/II)
and late stage (H-Y III+) were 0.3982 (95% CI: 0.2564-0.5399) and 2.1227 (95% CI:
0.5109-3.7346), respectively. Considering the effects of patient specific covariate on the
transitions in the five-state model, the results of multivariable analysis on multiple
transition shows that advancing age led to an increased 10 years risk of developing PD
(aRR=1.79, 95% CI: 1.32-2.44) and faster transition from HY I/II to HY III+ before
surfacing to CD phase (RR=5.08, 95% CI: 1.94-13.29). Low level of uric acid also
played the role of risk factor in the incidence of PD (RR=1.54, 95% CI: 1.04-2.28).
High level of education strongly affected the transition from HY I/II to HY III+ before
surfacing to CD phase (RR=14.65, 95% CI: 2.94-54.53).
In the simulated results for effectiveness of different screening interval, annual
screening reduced 71% (95% CI: 64-77%) reduction of advanced stage (H-Y stage III+) nceeeee rrrrraaatateeeeeeeeofofofofofofofofofbbbbbbbbbeieieieieieieieieingngngngngnngngng
p p p p
peeeere ssosososososososonnnnnnnn---yeyeyeyeyeyeyeyeararararararararssss.s.sss.
deteteeeectctcctctctababababablelell ((((CDCDCDCDCCDCD)))))
ix
cases compared to no screen. When the inter-screening intervals were 2-yearly, 3-yearly,
4-year, or 6-yearly, reduction of advanced H-Y stage cases was 54% (95% CI: 45-62%),
43% (95% CI: 32-52%), 35% (95% CI: 23-45%), and 25% (95% CI: 12-36%),
respectively.
The results from deterministic Markov decision analysis of the cost-effectiveness
and cost-utility analysis shows that the incremental cost-effectiveness ratios (ICER) of
PD screening with different inter-screening intervals compared to no screen ranged from
$1169 to $1804 per life-year gained. The incremental cost-utility ratio ranged from
$1715 to $2606 per quality-adjusted life-year gained. The annual screen had the greatest
net monetary benefit (NMB) ($280,687) in terms of life-year gained, followed by
biennial ($280,511), triennial ($280,416) screen, and no screen ($280,113). The same
trend was observed for the NMB in terms of QALY gained.
The results of the probabilistic Markov decision models shows that the
probability of screening programs being cost-effective at $20,000 of willingness-to-pay
(WTP) was 69-79% and 64-74% given 100% and 60% of attendance rates, respectively.
The corresponding figures in the cost-utility analyses were 62.6%-70.2% and 58.2-
62.6% given 100% and 60% of attendance rates, respectively.
Conclusion The active method detected almost two times the PD cases as the passive
method and also reduced 49 % (95% CI: 4%-73%) the IPD cases classed in H-Y stage III 2-yeyeyeyeyeaaaararlylylylyyyyyy,,,333333333--yeyeyeyeyeyeyeyeareararararararararlylylylyly, ,
5%
5%
5%
5%
5%CCCCCCCCI:I:I:I:IIII 444444445555555---6666666622222222%)%)%)%)%)%)%)%),
2-366666%)%)%)%)%)% ,,,,,
x
or greater. Our results reveal that an individual aged 60 year or older who is susceptible to
PD and entered the SD phase would progress to CD, on average around 1.5 years. The
progression from the SD to the CD by H-Y stage had been quantified with detectable
window for the identification of early H-Y stage before the transition to late H-Y stage
which form the bases of the best-case estimates for the disease progression of PD in the
absence of screening. With the application of these transition parameters, this thesis
demonstrates that if the intensive screening for PD is offered, the large the reduction in
late H-Y PD would be achieved and the probability of being cost-effective could be high.
Keywords: Parkinson’s Disease, Early Screening, Cost-Effectiveness, Hoehn-Yahr Stage ho o o o oisisisisisssususususususususu cececececececececeptptptptptptptptptibibibibibibibibiblelelllll tttttooooo
d d d d
d 1111.5.55555555yeyeyeyeyeyeyeyearararararararars.s.s.s.s.s.s..TTTTTTTThehehehehehehehe
with hhhhdededededed tetetetetectcctcabableleleeee
xi
CONTENTS
ᇞᖴ………i!
ύЎᄔा!………..ii!
ABSTRACT………vi.
Chapter 1 Introduction ...18
1.1 Impact of Parkinson’s Disease...18
1.2 Temporal Natural History Based on Hoehn and Yahr stage ...19
1.3 The Importance of Active Detective Method for Parkinson’s Disease Classified by Hoehn and Yahr Stage ...19
1.4 Effectiveness of Early Detection and Treatment for Parkinson’s Disease ...20
1.5 Cost-effectiveness Analysis of Early Detection for Parkinson’ Disease ...21
1.6 Motivation and Aims of the Study ...22
Chapter 2 Literature Review ...24
2.1 Burden of Parkinson’s Disease ...24
2.1.1 Clinical characteristics of Parkinson’s Disease ... 24
2.1.2 Incidence... 24
2.1.3 Prevalence... 24
2.2 Natural History of Parkinson’s Disease with Hoehn-Yahr Stage ………...26
2.3 Stochastic Models for Disease Natural History ...28
2.3.1 Introduction of Markov Model ... 28
2.3.2 Three-state Homogeneous Markov Model for Disease Natural History ... 30
2.3.3 Three-state Model with Weibull Distribution ... 31
2.3.4 Incorporation of patient specific covariates ... 33
…
…
…
…
…… … … … … … … …… … … … … … … …… … … … … … … …… … … … … … … …i i i i i i i i
……… … … … …… … … … …… …. . . . .. . .i i i i i ii i i i i i
xii
2.3.5 Bayesian inversion for a non-standard case-cohort design ... 33
2.3.6 Five-state non-homogeneous stochastic model ... 35
2.3.7 Semi-Markov Model ... 37
2.4 Covariates associated with the progression of Parkinson’s Disease ...39
2.4.1 Risk Factors ... 39
2.4.2 Protective Factors ... 42
2.5 Quality of Life by Hoehn-Yahr Stage ...45
2.6 Cost-effectiveness Analysis in Parkinson’s Disease ...47
2.6.1 Cost Analysis of Parkinson’s Disease ... 47
2.6.2 Cost Effectiveness Analysis of Parkinson’s Disease ... 48
Chapter 3 Study Design and Data Source ...50
3.1 Study Cohort ...50
3.2 Study design ...51
3.2.1 Cross-sectional survey ... 51
3.2.2 Natural History of Parkinson’s Disease with Hoehn-Yahr Stage with Stochastic Process Based on Case-cohort Design ... 53
3.2.3 Data Collection ... 54
3.2.4 Homogeneous Markov model incorporated with covariates associated with the transition rates ... 56
3.2.5 Cost-effectiveness analysis for early detection of Parkinson’s disease ... 56
Chapter 4 Hoehn-Yahr stage-based natural history of PD with Stochastic Process ...58
4.1 Homogeneous Markov model ...58
4.1.1 Model Specification ... 58
4.1.2 Likelihood ... 61
4.1.3 Estimation of parameter ... 70 ...333333 ...3535353535 ... 37373333733
rk k k k ki i i i in n n n ns s s s s so o o o o o o on n n n n n n n’ ’ ’ ’ ’ ’ ’s s s s s s s s s
39
xiii
4.2 Incorporation of patient specific covariates ...70
4.3 Simulation for the effect of screening policy ...71
4.4 Cost-effectiveness Analysis ... 73
Chapter 5 Results ...79
5.1 Part I: Compare the two detection methods for detecting Parkinson’s disease...79
5.2 Part II: To Elucidate the temporal natural history of Hoehn-Yahr- stage-based Parkinson’s disease with stochastic process ...81
5.2.1 Three-state Markov model ... 81
5.2.2 Five-state Markov model... 84
5.2.3 Incorporation of patient specific covariates for the five-state Markov model ... 85
5.3 Part III: Cost-effectiveness of Population-based Screening for PD89
5.3.1 Simulation for the effect of screening policy ... 895.3.2 Results of deterministic cost-effectiveness and cost-utility analysis ... 90
5.3.3 Results of probabilistic cost-effectiveness and cost-utility analysis ... 91
Chapter 6 Discussion ...94
6.1 Part I: Compare the two detection methods for active detecting Parkinson’s disease...94
6.2 Part II: Natural History of Parkinson’s Disease by Hoehn-Yahr Stage 97 6.3 Part III Cost-effectiveness Analysis of screening of PD ... 102
Chapter 7 Conclusion ...107
FIGURES
Figure 3-1 Simulated randomized controlled trial study design ... 108Figure 3-2-1 Decision tree of Parkinson’s disease screening ... 109
Figure 3-2-2 Decision tree of Parkinson’s disease screening (continue) ... 110
... . . . .. . .. . . . .. . .... .. . . .. . . . .. .... . . . .. . . . .. . . ... . . .7 7 7 7 70 0 0 0 0 .
. . .
.. . .. .. .. .. . . . . . . . .. . . . .. . .. . . .. . . . . .. . . . .. . .. . . .. . . . . .. . . .. . . .. . . . . . .. . . .. . .. . . . . . . .. . .. . .. .7 7 7 7 7 7 7 7 71 1 1 1 1 .
. . .
.. . . .. . .. .. .. . . . . . . ... ... . . . . . .. . . .. .. . .. . . . .. . . . . . .. . . . . . . .. . . . . .. .. . . . . . .. . . . . .. . . . . .. . . . . .. 7 7 7 7 7 7 7 73 3 3 3
... .. . . . .. . .. . . .. . . .. . . . .. . . .. . . .. . . . .. . . . .. . . .. . . .. . .. . .7 7 7 7 7 79 9 9 9 9
xiv
Figure 3-2-3 Decision tree of Parkinson’s disease screening (continue) ... 111
Figure 5-1-1 Study Flow Chart ... 112 Figure 5-1 2 Cumulative detection rate of two methods of detecting Parkinson’s
disease. ... 113
Figure 5-2-1 Study flow chart include participants age 60 and older for analysis.
... 114 Figure 5-2-2 Cumulative risk for the SD and CD from free of PD in three-state
model ... 115 Figure 5-2-3 Cumulative risk of surfacing to the CD from the SD in three-state
model ... 116 Figure 5-2-4 Cumulative risk for the SD and CD from free of PD in three-state
model (sampling fraction) ... 117 Figure 5-2-5 Cumulative risk of surfacing to the CD from the SD in three-state
model (sampling fraction) ... 118 Figure 5-2-6 Predict 20-year risk of being early and advanced H-Y stage ... 119 Figure 5-2-7 The predicted 20-year risk of PD by Hoehn-Yahr stage ... 120
Figure 5-3-1 Scattered incremental cost-effectiveness analysis for 1-year vs. no screening ... 121 Figure 5-3-2 Scattered incremental cost-effectiveness analysis for 2-year vs. no
screening. ... 121 Figure 5-3-3 Scattered incremental cost-effectiveness analysis for 3-year vs. no
screening. ... 122 Figure 5-3-4 Acceptability curve for cost-effectiveness analysis for various inter- screening intervals ... 122 Figure 5-3-5 Scattered incremental cost-effectiveness analysis for 1-year with
100% attendance rate vs. no screening. ... 123 Figure 5-3-6 Scattered incremental cost-effectiveness analysis for 2-year with
100% attendance rate vs. no screening. ... 123 Figure 5-3-7 Scattered incremental cost-effectiveness analysis for 3-year with
100% attendance rate vs. no screening. ... 124 Figure 5-3-8 Acceptability curve for cost-effectiveness analysis for various inter- screening intervals with 100% attendance rate. ... 124 Figure 5-3-9 Scattered incremental cost-effectiveness analysis for 1-year with
60% attendance rate vs. no screening. ... 125 Figure 5-3-10 Scattered incremental cost-effectiveness analysis for 2-year with
contntntttinininininueueuueue))))))... 11111111111
. .
...1111111111122 te
teee
teccctctcininnnnnnnnggg ggg ggPaPPPaPPaPaP rkrkrkrkrkrkrkrkininininininininsosososososososon’n’n’n’n’n’n’s’s sss ...11111111111111333333
xv
60% attendance rate vs. no screening. ... 125 Figure 5-3-11 Scattered incremental cost-effectiveness analysis for 3-year with
60% attendance rate vs. no screening. ... 126 Figure 5-3-12 Acceptability curve for cost-effectiveness analysis for various
inter-screening intervals with 60% attendance rate. ... 126 Figure 5-3-13 Scattered incremental cost-utility analysis for 1-year vs. no
screening. ... 127 Figure 5-3-14 Scattered incremental cost-utility analysis for 2-year vs. no
screening. ... 127 Figure 5-3-15 Scattered incremental cost-utility analysis for 3-year vs. no
screening. ... 128 Figure 5-3-16 Acceptability curve for cost-utility analysis for various inter-
screening intervals ... 128 Figure 5-3-17 Scattered incremental cost-utility analysis for 1-year with 100%
attendance rate vs. no screening. ... 129 Figure 5-3-18 Scattered incremental cost-utility analysis for 2-year with 100%
attendance rate vs. no screening. ... 129 Figure 5-3-19 Scattered incremental cost-utility analysis for 3-year with 100%
attendance rate vs. no screening. ... 130 Figure 5-3-20 Acceptability curve for cost-utility analysis for various inter-
screening intervals with 100% attendance rate. ... 130 Figure 5-3-21 Scattered incremental cost-utility analysis for 1-year with 60%
attendance rate vs. no screening. ... 131 Figure 5-3-22 Scattered incremental cost-utility analysis for 2-year with 60%
attendance rate vs. no screening. ... 131 Figure 5-3-23 Scattered incremental cost-utility analysis for 3-year with 60%
attendance rate vs. no screening. ... 132 Figure 5-3-24 Acceptability curve for cost-utility analysis for various inter-
screening intervals with 60% attendance rate. ... 132
Figure 6-2-1 The predicted 20-year risk of PD by Hoehn-Yahr stage assuming Weibull distribution for transitions ... 133
TABLES
Table 4-6-1 Estimate and distribution of parameters ... 134 ... 12121212125 sisisisisisfffffororrrrrrrr 33333333---yeyeyeyeyy arararararwwwwwwwwititititititititith h hhhhh
. .
...1212222222266 ys
ys ys ys
ysisisii fffffffffororororororororvvvvvvvararararararararioioioioioioioiousususususususus ... 12121212121112666666 1-yearvvvvvsssss.nnnnoooo
16
Table 5-1-1 Annual Incidence of PD in Active Detection Group ... 136 Table 5-1-2 Annual Incidence of PD in Passive Detection Group ... 137 Table 5-1-3 Baseline characteristics of two groups of those with idiopathic
Parkinson’s disease by detection method. ... 138 Table 5-1-4 Distribution of Hoehn-Yahr (H-Y) stage for cases of idiopathic
Parkinson’s disease (IPD) detected by the active or passive method. ... 139 Table 5-1-5 Crude and adjusted relative risk for active and passive detection
methods for Parkinson’s disease. ... 140
Table 5-2-1 H-Y stage distribution in screen-detective case and clinical-detective case ... 141 Table 5-2-2 Estimated transition rates with three-state model ... 142 Table 5-2-3 Estimated transition rates with a three-state model using a case-
cohort design sampling fraction ... 143 Table 5-2-4 Estimated transition rates with a five-state model using a case-
cohort sampling fraction design ... 144 Table 5-2-5 Distribution of characteristics of subjects ... 145 Table 5-2-6 Relative risk on transition rate of normal to SD early phase of five-
state Markov model of Parkinson’s disease ... 146 Table 5-2-7 Relative risk on transition rate of SD early to SD late phase of five-
state Markov model of Parkinson’s disease ... 147 Table 5-2-8 Relative risk on transition rate of SD early to CD early phase of
five-state Markov model of Parkinson’s disease ... 148 Table 5-2-9 Relative risk on transition rate of SD late to CD late phase of five-
state Markov model of Parkinson’s disease ... 149 Table 5-2-10 Multivariate analysis on transition rate of normal to SD early phase
... 150 Table 5-2-11 Multivariate analysis on transition rate of SD early phase to SD
late phase ... 152 Table 5-2-12 Multivariate analysis on transition rate of SD early phase to CD
early phase ... 153 Table 5-2-13 Multivariate analysis on transition rate of SD late phase to CD late
phase ... 155 Table 5-2-14 Covariate in transition of five state model with hypothesis testing
... 157 Table 5-2-15 Multivariate analysis for the multiple transition in the five-state
Markov model ... 159 Table 5-2-16 Multivariate analysis for the multiple transition in the five-state
up... 13131313136 ouuuuupppppp...13131111111 77777 wiii
wi
witttththiiiiiiiiididddidididdopopopopopopppatatatatatatatathihihihihihihiiccc ccccc ...13131313131383388 s ooooof f f f f ididididi ioioioioioioioioiopapapapapapapapapathththththththththiciicicicicicicic ive meeethththththodododododo ...11139
17
Markov model with further adjustment of vegetable and fruit intake ... 160
Table 5-3-1 The simulated results of PD cases by HY stage at diagnosis with 1-, 2-, 3-, 4-, and 6-yearly screening in 12 years for a hypothetical cohort of 9829 elderly people aged 60 at entry ... 162 Table 5-3-2 Incremental cost-effectiveness ratio (ICER) and cost-utility ratio
(ICUR) among screening strategies by attendance rate ... 163 Table 5-3-3 The distribution of cost, effectiveness, and net monetary benefit 164
ABBREVIATION NOTE……….165 REFERENCE………166
fruuuitititititiiiiintntntntntakakakakakaake... 16161616160
t t t t
t dddiddiaagagagagagagagagnononoooosisisisisisisisissssssssswiwiwiwiwiwiwiwithththththththth11---, he
he he he
hetiticcacacacacacacacalllll lllcccococcococ hohohohohohohohortrtrtttttooooooooffff f ff ...16161616161116222222 cost-utititititilililililitytytytytyt rrrratatatatatioioioioio
18
Chapter 1 Introduction
1.1 Impact of Parkinson’s Disease
Parkinson’s disease (PD) is the second most common degenerative disorder in the
aging brain. It affects approximately 6.3 million people worldwide. As the disease
progress, it will affect motor, autonomic, cognitive and emotional function and
eventually reduce lifespan.1, 2The cardinal symptoms of PD such as tremor, rigidity,
bradykinesia and postural instability involve motor control. Disability in PD derives
predominantly from progressive motoric disturbance which may lead the patient
become wheelchair-bound or bedridden. Such heath consequence results in a
considerable burden of illness associated with PD. Although PD is still not curable, the
advent of the levodopa raise the hope of improving both motor disability and survival in
PD.3Before the introduction of Levodopa, previous epidemiological studies report that
patients with PD had a shorter survival than the general population.4Hoehn and Yahr
reported a mortality ratio 2.9 times higher in PD patients than that of the general
population after adjustment for age, sex and race.5
19
1.2 Temporal Natural History Based on Hoehn and Yahr stage
The severity of PD is usually classified by Hoehn and Yahr stage (H-Y stage).5In
the absence of treatment, the disease severity will progress to H-Y stage IV and V in 9.0
± 7.2 and 14.0 ± 3.4 years.5Previous study reported that H-Y stage at baseline were
greater in PD patients who had died during follow-up compared with that of survivors.6
Besides, patients with H-Y stage greater than III reported the impaired quality of life
and more non-motor symptoms.7This implies that H-Y stage plays an important role in
the natural history of PD for assessing both disease progression and prognosis of H-Y
stage.
In addition, those covariates associated with each transition rate between
consecutive stages were also with high interest to use them into the natural history
model to reduce the heterogeneity and also provide the information.
1.3 The Importance of Active Detective Method for
Parkinson’s Disease Classified by Hoehn and Yahr Stage
However, most studies detected PD cases by medical record review or service-
based detection, which usually detected PD case with syndrome at the late stage rather
than early stage.8-14Therefore, the incidence and prevalence of PD in door-to-door
an n n n nd d d d d Y Y Y Y Y Y Y Y a a a a a a a a ah h h h h h h h hr r r r r r r r
(H-Y YYYYstststststagagagagage)e)).55555555InInInInIIn
20
survey were higher than those in record-based studies.9This discrepancy implies that
outreaching surveys can yield accurate PD prevalence and incidence rates. A study in
Taiwan showed that a community-based screening program identified more early stage
PD with H-Y stage I or II than that was performed in a clinical series.15Such active
method suggested the possibility of detecting PD at early stage, and accompanied with
the effectiveness of levodopa in delaying the progression of PD, the life expectancy and
the quality of life would be expected to be improved. While temporal natural history of
H-Y-stage-based PD was proposed by Hoehn and Yahr, early detection of PD was not
envisaged at that time. In the era of preventive medicine in the 21 century, it seems
feasible as a result of effective early treatment. Screening for PD has become feasible as
Liou et al has already done in such an active detection.15With the advent of screening
for PD, PD with H-Y stage can be further divided into the screening detectable (SD)
phase and clinical detectable (CD) phase. In my thesis, the temporal natural history of
PD with H-Y stage will be classified into the SD and the CD phase for estimating the
parameters of disease progression.
1.4 Effectiveness of Early Detection and Treatment for Parkinson’s Disease
cyyyyiiiiimpmmmpmplilililililiiiiesesesesesesesessttttttttthahahahahahahahahat tttttttt
t t t t
teeeese . A AAA A A A AAststststststststududududududududy y y y y y y y inininininininin
more eeeeeaeaeaeaearlrlrlrlry yyy ystsagagagagaga eee e e
21
Progression of disability on the H-Y stage has become slower with the introduction
of levodopa treatment. The progression to severe PD would be rapid for those patients
with delayed administration of levodopa therapy. 16, 17The development of therapies that
slow disease progression and improve survival makes early detection and treatment of
PD especially important. The elucidation of temporal natural history of H-Y-stage-based
PD also provide a pseudo-control group for evaluation for preventive strategy such as
screening for early PD. It has been shown that screening for early PD can lead to 51%
reduction for advanced stage of PD, and 25% mortality reduction.18Thus, early
detection could relieve medical burden from PD not only for patients themselves, but
for family members, and even the society.
1.5 Cost-effectiveness Analysis of Early Detection for Parkinson’ Disease
There are many economic evaluations for treatment of PD, but cost-effectiveness
analysis for PD screening has been scarcely addressed. Most economic evaluation
articles in PD were performed by deterministic approach although the uncertainly in
natural history of PD and also in treatment of PD was well-known in this field. Since
the advance in methodology of cost-effectiveness analysis has increasingly gained
attention over the past decades, stochastic process in decision tree and using Bayesian h ttttheheheheheiintntntntntntntttrorororororororooduddudududududuductctctctctctctctctioioioioiioioioi n n
r r r r
r tttthohooooooooseseseseseseseseppppppppatatatatatatatatieieieieieieieientntntntntntntnts ssss sss
t offfffttttttheheheheheh rarararar pipipippesesssttttthahahahahat tt tt t
22
approach with probabilistic sensitivity analysis has also gained popularity to alleviate
concerns related to the dynamic changing of quality of life depending on disease status
and the uncertainty related to treatment and cost.
1.6 Motivation and Aims of the Study
There are few studies to depict the panorama of the natural history of PD based on
H-Y stage from various perspectives on epidemiological, clinical, and economic
aspects. Besides, the characteristics of heterogeneity in natural history and the
uncertainty in the decision analysis of early detection of PD prevention have not been
fully investigated.
The aim of this thesis includes four parts based on the principle of evidence-based
medicine.
Part I: To make use of a population and community-based cohort study to compare the two detection methods for active detecting Parkinson’s disease.
Part II: To elucidate the temporal natural history of Hoehn-Yahr-stage-based Parkinson’s
disease with stochastic process in relation to early detection of PD based on empirical data
from Part I.
Part III: To identify H-Y stage-specific factors responsible for various transitions.
Part IV: Perform cost-effectiveness and cost-utility analysis for early detection of ity y y y y tototototoaaaaaaaallllllllllllllllleveveveveveveveveviaiaiaiaiaiaiaiaiatetetetetetetetete
on on on on
onddiisisisisisisisi eaeaeaeaeaeaeaeasesesesesesesesessssssssttttttttatatatatatatatatusususususususus
23
Parkinson's disease through population-based screening.
24
Chapter 2 Literature Review
2.1 Burden of Parkinson’s Disease
2.1.1 Clinical characteristics of Parkinson’s Disease
Idiopathic Parkinson’s disease (IPD) is the second most common degenerative
disorder in the aging brain, after Alzheimer’s dementia. The cardinal signs of motor
dysfunction of Parkinson’s disease (PD) include resting tremor, bradykinesia, rigidity
and postural reflex impairment. The pathological finding of the motor deficits in PD is
degeneration of the dopaminergic neurons of the nigrostriatal pathway.
Catecholaminergic and serotoninergic brain-stem neurons may also degenerate. These
mechanisms may include protein misfolding, protein aggregation, mitochondrial
dysfunction, oxidative stress and inflammation.19-26
2.1.2 Incidence
Overall, the incidence rates for PD in all groups ranged from 1.2 to 22 per 100,000
person-years. If restricted to older populations (age above 55 or 65 years), the incidence
rates were increased between 410 and 529 per 100,000 person-years.11, 27, 28A systemic
review of incidence studies of PD reported that the age-standardized incidence rates
between 16 and 19 per 100,000 person-years.29
2.1.3 Prevalence
25
Unlike the few incidence studies, there are plenty of prevalence studies of PD. Von
Campenhausen et al reported the prevalence rate range from 65.6 to 12,500 per 100,000
in European countries. Alves et al reported overall prevalence rate in door-to-door
studies ranged from 167 to 5,703 per 100,000 worldwide.30Though previous two
studies in China reported low prevalence rate of PD,31, 32Zhang et al directly examined
29,545 individuals reported a prevalence of 1,300 per 100,000 in individuals above 55
years.14The two door-to-door survey in Ilan and Kimen also reported the prevalence
were 119 and 130 per 100,000 after calculate age-standardized prevalence proportions
using the US population in 1970 as standard, 33, 34which were similar to the prevalence
in European countries.10, 13, 35-37Thus, the low prevalence in China may resulted from
difference in methodology, rather than ethnic differences.
Although there are large variation in incidence and prevalence of PD, outreach
surveys such as door-to door surveys usually reported higher incidence and prevalence
compared to registry-based studies of ascertainment. To the best of our knowledge, no
population-based data are available to compare different case-finding methods in PD
detection.
udididididieseseeses ooooooff f ff f f ffPDPDPDPDPDPDPDPDPD. .VoVoVoVoVoVoVoVoVon n
, , , ,
,555505000000 0000pepepepepepepeper r r r rrrr10101010101010100,0,0,0,0,0,0,0,000000000000000000 000 000
oor-tototototo-o--dodododod oror
26
2.2 Natural History of Parkinson’s Disease with Hoehn-Yahr
Stage
Margaret M. Hoehn and Melvin D.Yahr first introduce the H-Y stage based on the
clinical disability of PD in 1967.5The comparable clinical disability of each stage are as
follows:
Stage I- Unilateral involvement only, usually with minimal or no functional impairment.
Stage II- Bilateral or midline involvement, without impairment of balance.
Stage III- First signs of impaired righting reflexes. This is evident as the patient turns or is
demonstrated when he or she is pushed from standing equilibrium with the feet together
and eyes closed.
Stage IV- Fully developed, severely disabling disease; the patient is still able to walk and
stand unassisted but is markedly incapacitated.
Stage V- Confinement to bed or wheelchair unless aided.
Hoehn and Yahr evaluate the total 183 patient of primary parkinsonism and provided the
mean duration of each stage of illness was 3.0, 6.0, 7.0, 9.0, and 14.0 in stage I, II, III, IV
and V, respectively. Progression of disability on the H-Y stage has become markedly
slower with the advantage of levodopa treatment and studies from the post-levodopa era
have found latencies to reach H-Y stage IV or V of up to 40 years.38Hely et al reported a
o o o o
oe e e e e eh h h h h h hn n n n n n n n n-Y Y Y Y Y Y Y Y a a a a ah h h h h h h hr r r r r r r r
27
cohort of 146 PD patient with 10-year follow up data and found median time to reach H-Y
stage IV and V was around seven years.39Different rates of progression of PD between
studies might be due to differences in patient cohorts studied. In addition, progression of
motor impairment is likely non-linear in PD with severe declines in early stage versus late
stage of the disease, which was compatible with the exponential decline of neuronal cell
counts in the substantia nigra in the brain.40This is supported by the observations of faster rates of progression of unified Parkinson’s disease rating scale (UPDRS) in the first versus
the 10th year of disease.41Liou et al. reported the average duration in H-Y stage I, II and
III was estimated as 2.83, 6.62 and 1.41 years, respectively by proposing a five-state
Markov model.15These different rates of progression in PD between studies also
suggested heterogeneity in the natural history of PD.
To model the natural history of Parkinson’s disease is often complicated by issues
of diagnostic accuracy, heterogeneity of different forms of the disease and the
confounding effects of age related comorbidities. The H-Y stage is used for evaluation
the progression of PD. The H-Y model assumes that PD is a progressive disease,
evolving from H-Y stage I to H-Y stage V. Since the introduction of L-dopa, detailed information about how a patient’s disease progressed form H-Y scale I to scale V for
untreated PD are unlikely to be quantifiable. The stochastic model was therefore
proposed. Stochastic models have been used in modelling the disease natural history of n titititiimememmm tttttttoo o o oo oorrererererererereacacacacacacacachch h hhhhhhH-H-H-H-YYYYY
n n n n
n ooooff fPDPDPDPDPDPDPDPDbbbbbbbbetetetetetetettweweweweweweweweenenenenenenenen
on, ppppprororororor grgrgrgrgreesee sis ononononono ooooof f f f ff
28
multi-state chronic diseases.42, 43Liou et al proposed a five-state Markov model
according to the disease severity by H-Y stage.15The H-Y model assumes that PD is a
progressive and irreversible disease. It means that an individual diagnosed as stage V is
supposed that he or she has transited from normal, through stage I, II, III and IV at entry
of study. (see the figure below)
However, the Markov model used to assume a homogeneous process that a
constant hazard rate with time for progression for state to state. This may be unrealistic
in medicine and biology.
2.3 Stochastic Models for Disease Natural History 2.3.1 Introduction of Markov Model
A sequence of random variables {ܺఈ,Ƚ= 0,1,…} is called a Markov chain if, for every
collection of integers, ߙ ൏ ߙଵǡ ൏ ڮ ൏ ߙ ൏ ߚ, the conditional distributions of
ܺఉsatisfy the relation:
ܲ൛ܺఉ ൌ ݅ఉหܺఈబǡ ǥ ǡ ܺఈൟ ൌ ܲ൛ܺఉ ൌ ݅ఉหܺఈൟ, for ݅ఉ
The outcome in the future (ܺఉൌ ݅ఉሻ݅ݏ݈݊݊݃݁ݎ݀݁݁݊݀݁݊ݐݑ݊ݐ݄݁ܽݏݐݏݐܽݐ݁
ov mmmmmoododododelelelelelelelele
me me me me
mes tththththththththatatatatatatatatPPPPPPPPD D D D D D D D isisisisisisisisaaaaaaaa
sed ddddasasasasasa ssssstatatatagegeggg VVVVVVViiiiisss s s
29
(ܺఈబǡ ǥ ܺఈషభሻ
For each ܺఈ, the absolute probability is denoted by ܲሼܺఈ ൌ ݅ఈሽ ൌ ܽഀ
For every pair of random variables, ఈandܺఉ, the conditional probability is denoted by
ܲሼܺఉ ൌ ݅ఉŽܺఈ ൌ ݅ఈሽ ൌ ܲഀǤഁġ
The joint probabilities of ܺఈǡ ܺఉǡ ܺఊ, for Ƚ ൏ Ⱦ ൏ ɀ, are given by
ܲ൛ܺఈൌ ݅ఈǡ ܺఉ ൌ ݅ఉǡ ܺఊൌ ݅ఊൟ ൌ ܽഀܲഀǡഁܲഁǡംǡ ܽ݊݀ ܲ൛ܺఈൌ ݅ఈǡ ܺఉ ൌ ݅ఉൟ ൌ ܽഀܲഀǡഁ
Therefore, for any collection of integers Ƚ ൏ Ⱦ ൏ ڮ ൏ Ɂ ൏ ɂ, the joint probabilities are
ܲ൛ܺఈൌ ݅ఈǡ ܺఉ ൌ ݅ఉǡ ǥ ǡ ܺఋൌ ݅ఋǡ ܺఌ ൌ ݅ఌൟ ൌ ܽഀܲഀǡഁǥ ܲഃǡച
A Markov chain with state space being the set of all the non-negative integers is
completely determined by the initial absolute probability distribution
ܲሼܺ ൌ ݅ሽ ൌ ܽబǡ݅ ൌ ͳǡʹǡ… and the transition probabilities
ܲሼܺఈାଵ ൌ ݅ఈାଵȁܺఈൌ ݅ఈሽ ൌ ܲഀǡഀశభ , ݅ఈǡ ݅ఈାଵ ൌ ͳǡʹǡ ǥ for Ƚ=0,1,…
The transition probabilities of a time homogeneous chain is denoted by
ܲሼܺఈାଵ ൌ ݆ȁܺఈ ൌ ݅ሽ ൌ ܲ
The transition probability ܲ for a three-state Markov model can be arranged in the form
of a matrix
P=൭
ܲ ܲଵ ܲଶ
ܲଵ ܲଵଵ ܲଵଶ
ܲଶ ܲଶଵ ܲଶଶ൱
ഀ
ഀ
ഀ
ഀ
ഀ
bilititititity y y y yy isisisisisddddenenotototttttedededededbbbbbbyyyyy
30
2.3.2 Three-state Homogeneous Markov Model for Disease Natural History
Chen et al applied a three-state Markov model to estimate sojourn time in chronic
disease screening without data of interval cases.43They model the disease with a
continuous-time Markov process in which X(t), the state of an individual at time t, is a random variable with a state space Ω={0,1,2}, where 0 represents no disease, 1 represents
preclinical screen detective disease (PCDP) and 2 represents clinical phase (CP). The
clinical phase in this model is an absorbing state in Markov processes language because
the natural history cannot be estimated beyond diagnosis due to the effect of therapy. They
also assume this is a progressive model.
The transition rates in the three-state model can be expressed as an intensity matrix,
൭െߣଵ ͲͲ ߣଵ
െߣଶ Ͳ
Ͳ ߣଶ
Ͳ
൱ (2-1)
ߣଵ represents the transition rate from no disease to the PCDP, ߣଶ represents the transition
rate from the PCDP to the clinical phase.
Given the transition intensity matrix above, transition probabilities for a three-state model
can be expressed as
൭ܲሺݐሻ
ͲͲ ܲଵሺݐሻ
ܲଵଵሺݐሻ Ͳ
ܲଶሺݐሻ
ܲଵଶሺݐሻ ͳ
൱ (2-2)
N N N N
Na a a a at t t t tu u ur r r r r r r ra a a a a a a a al l l l l l l l l
timmmmmeeeeee ininininincccchrhrononnnnnicicicicc
31
ܲሺሻ ൌ ݁ିఒభ௧
ܲଵሺݐሻ ൌߣଵሺ݁ିఒభ௧െ ݁ିఒమ௧ሻ ሺߣଶെ ߣଵሻ
ܲଶሺሻ ൌ ͳ െఒఒమషഊభ
మିఒభ ఒఒభషഊమ
మିఒభ (2-3)
ܲଵଵሺݐሻ ൌ ݁ିఒమ௧
ܲଵଶሺݐሻ ൌ ͳ െ ݁ିఒమ௧
The likelihood function based on the prevalent screen in a cohort with N individuals is
ܮ
ଵሺǤ ሻ ൌ ෑ ൬ ܲ
ଵሺݒ
ሻ
ܲ
ሺݒ
ሻ ܲ
ଵሺݒ
ሻ ൰
௫
ൈ ൬ ܲ
ሺݒ
௩ሻ
ܲ
ሺݒ
ሻ ܲ
ଵሺݒ
ሻ ൰
ଵି௫
ே
ୀଵ
ݒ represents age at fist screen for mth subject
ݔ ൌ ͳ when the mth subject is detected as a positive case
ݔ ൌ Ͳ otherwise.
However, as the previous mention above, the Markov model used to assume a
homogeneous process that a constant hazard rate with time for progression for state to
state. This may be unrealistic in medicine and biology.
2.3.3 Three-state Model with Weibull Distribution
In order to deal with the non-constant hazard in the stochastic model, Chen et al
propose a non-homogeneous three-state model for the disease natural history of oral
cancer.44They model the time of transitions from normal to leukoplakia and leukoplakia
to invasive carcinoma with two Weibull distributions. The transition probabilities for
staying in a no disease state (state 0), transitions from normal to leukoplakia (state 1) (2 (2 (2 (2 ( --3)3)3)3)3)
32
and from normal to invasive carcinoma (state 2) in a given time interval [t1, t2] are
expression as follows:
ܲሺݐଵǡ ݐଶሻ ൌ ͳ െ න ݂ଵ
௧మ ௧భ
ሺݑሻ݀ݑ
ܲଵሺݐଵǡ ݐଶሻ ൌ ݂௧௧మ ଵ
భ ሺݑሻ ቀͳ െ ݂௧మ ଶ
௨ ሺݒሻݒቁ ݀ݑ(2-4)
ܲଶሺݐଵǡ ݐଶሻ ൌ න ݂௧మ ଵ
௧భ
ሺݑሻ න ݂௧మ ଶ
௨
ሺݒሻ݀ݒ݀ݑ
f1(t) and f2(t) are the probability density function of Weibull distributions for time of
transition from states 0 to 1 and from state 1 to 2. The two Weibull distributions are
denoted as W1(ߣଵ,ߛଵ)and W2(ߣଶ, ߛଶ). ߣଵ andߣଶ are scale parameters and ߛଵ and ߛଶ are shape parameters for the two corresponding transitions. The transition rates as a
function of time are expressed as follows:
ߣ ൌ ߣߛݐఊିଵ where i=1 or 2
The probability of remaining in state i-1 in time t is
ܵሺݐሻ ൌ ቄെ ߣ௧ ߛݑఊିଵݑቅ ൌ ሺെߣݐఊሻ (2-5)
The corresponding probability density function is
݂ሺሻ ൌ ߣߛݑఊିଵሺെߣݐఊሻ
The transition probabilities for staying in state 1 and state 2 were also denoted as
follows:
ܲଵଵሺݐଵǡ ݐଶሻ ൌ ͳ െ ݂௧మ ଶ ௧భ ሺݑሻݑ
ܲଵଶሺݐଵǡ ݐଶሻ ൌ ݂௧మ ଶ
௧భ ሺݑሻݑ (2-6) al [[[[[ttttt11111, tttt ]22222]]]]]]]]ararararararararareeeeeeee
33
The natural history from state 1 (leukoplakia) to state 2 (invasive carcinoma) is usually
unobservable due to the interruption of medical treatment. We can only estimate
parameters via equation (1), P00, P01and P02.
2.3.4 Incorporation of patient specific covariates
The effect of patient specific covariates, say x, on the three-state stochastic model was
assessed by the exponential regression model that treats scale parameter in the Weibull
distribution as a function of patient-specific covariates. It is expressed as follows:
ߣ ൌ ߣሺߚ߯ሻ
ߣ : the scale parameter of Weibull distribution for state i
߯ : a vector of covariates for subject m
ߚ : corresponding regression coefficient
2.3.5 Bayesian inversion for a non-standard case-cohort design
For an n-state disease natural history, n sets of random samples for each transition were
selected in case-cohort study design in Chen et al. Let S denoted an indicator of whether
a subject was sampled (S=1). For individual i, let ߨ௧ be sampling fractions for state j
at time ti . ߨ௧ was denoted as follows:
ߨ௧ ൌ ሺ ൌ ͳȁͲ ՜ ݆Ǣ ݐሻ
nomommmma)aa)a)a iiiiiiss s ss sssuususususususususuauauauauauauauaualllllllllllllllllly y yyyyyyy
y y y
yeeeeestiimimimimimimimimatatatatatatatateee e e e
34
The sampling fractions for state j can be expressed as ߨ if we assume that sampling
fractions are independent of the individual. Using Bayesian inversion, the probability of
transition of being state j at time tigiven a subject was sampled is P(0՜ ݆Ǣ ݐȁܵ ൌ ͳሻ
= ሺୗୀଵȁ՜Ǣ௧ሻሺ՜Ǣ௧ሻ
σೕసభ൫ ൌ ͳหͲ ՜ ݆Ǣ ݐ൯ሺ՜Ǣ௧ሻ = σ గೕሺ՜Ǣ௧గ ሻ
ೕሺ՜Ǣ௧ሻ
ೕసభ = σ గೕగబೕሺ௧ሻ
ೕబೕሺ௧ሻ
ೕసభ (2-7)
The transition probabilities P0j(ti) are derived from equation (1).
Likelihood function, parameter estimation and model validation
The data on the first oral examination were used to estimate the parameters relate to the
disease natural history. This yields three possible observed transitions before the first
examination: staying in normal (state 0 Æ 0), normal to leukoplakia (state 0Æ 1) and
normal to invasive carcinoma (state 0 Æ 2). According to the above equation, P(0՜ ݆Ǣ ݐȁܵ ൌ ͳሻ
= ሺୗୀଵȁ՜Ǣ௧ሻሺ՜Ǣ௧ሻ
σೕసభ൫ ൌ ͳหͲ ՜ ݆Ǣ ݐ൯ሺ՜Ǣ௧ሻ = σ గೕሺ՜Ǣ௧గ ሻ
ೕሺ՜Ǣ௧ሻ
ೕసభ = σ గೕగబೕሺ௧ሻ
ೕబೕሺ௧ሻ
ೕసభ (2-8)
The likelihood function for the normal-leukoplakia-invasive carcinoma cohort with
three covariates is ς ൬σగబగబబሺ௧ሻ
ೕబೕሺ௧ሻ
మೕసబ ൰బ൬σగభൈగబభሺ௧ሻ
ೕబೕሺ௧ሻ
మೕసబ ൰భ൬σగమൈగబమሺ௧ሻ
ೕబೕሺ௧ሻ మೕసబ ൰మ
(2-9)
where ݊, ݊ଵ, and ݊ଶ were counts of normal, leukoplakia and invasive carcinoma at
age i of the first examination.
thahahahahattttt sasaaaaaampmpmpmpmpmpmpmpmplililililililililingngngngngngngngng
t t t t
thhhheheppppppppprorororororororobababababababababibibibibibibibililililililililitytytytytytytytyoooooooofffffff
