青光眼量性預防醫學與成本效益評估

Quantitative Aspects of Prevention and Economic Evaluation of Glaucoma

楊佳靜(Chia-Ching Yang)

指導教授 ： 陳秀熙

國立臺灣大學/公共衛生學院/預防醫學研究所/博士(2008年)

https://doi.org/10.6342/NTU.2008.03311

摘要

前言
青光眼在特定族群的疾病負擔與發生率、其臨床病程動態變化、用何種工具進行青光眼篩檢、此篩檢工具是否符合成本效益等問題較少被討論；而且年輕時曾接受雷射原位角膜成型術(LASIK)之患者其術後眼壓測量值之降低會對上述問題會產生何種效應也須被考慮。
研究目的
1.估計台灣地區23縣市在西元1999年至2003年間青光眼之年齡別與性別盛行率與發生率。
2.觀察已在接受青光眼降壓治療之患者其眼壓(IOP)、視神經盤及視野等動態變化來量化青光眼的臨床病程，並評估此臨床病程在各類型青光眼的異同。
3.考慮眼壓及其他預測因子之預測模型，並應用於族群青光眼篩檢。
4.應用多層次模式(Multi-level model)校正雷射原位角膜成型術(LASIK)術後被低估之眼壓值。
5.比較依現行青光眼一般臨床照護方式、積極介入更嚴格控制眼壓、進行青光眼篩檢並積極介入更嚴格控制眼壓等三種方案之成本效益來做經濟評估。
研究方法
本研究蒐集三個資料庫：以1999至2003年之全民健保資料庫進行上述研究目的(1)之研究；以在2002至2004年間至馬偕醫院就診之3360位青光眼患者隨機抽樣107位、收集病歷資料進行目的(2)之研究；以2003年馬祖地區1350位接受青光眼篩檢居民資料進行目的(3)之研究。另從文獻回顧產生三組電腦模擬數據及收集兩組醫院資料進行目的(4)中雷射原位角膜成型術(LASIK)術前術後眼壓差異之分析。綜合上述四部份研究得到之參數，與結合眼壓、前房深度、水晶體厚度、生活型態及生化因子產生之青光眼預測模型進行研究目的(5)中比較三種青光眼照護策略之經濟評估。
就方法學而言，第一部份是用布瓦松回歸模型(Poisson regression model)探討年齡、性別、及地理區域對青光眼盛行及發生個案的影響。用四階段馬可夫模型(Markov model)估計四階段間之瞬間轉移速率，其三個轉移狀態(transient state)為低、中、高三組不同眼壓狀態(分別表示眼壓值< 15mmHg、15到18 mmHg、≧18 mmHg)；唯一收斂狀態(absorbing state)為惡化之青光眼，乃以下列三種病況第一次出現之時間點來定義：視神經盤凹陷比率大於零點六為CD1，視神經盤凹陷比率大於零點七為CD2，視野平均缺損(mean defect, MD)大於零為VF。此為第二部份臨床病程之研究。第三部份乃應用邏輯式回歸模型做ROC (Receiver Characteristics curve)分析來決定有效的青光眼預測模型並應用於族群篩檢。多層次模式則被用在第四部份以後設研究(meta-analysis)做雷射原位角膜成型術(LASIK)術後被低估之眼壓值的校正。最後則擷取前四部份得到之參數為基礎置入馬可夫循環決策樹(Markov cycle tree)進行經濟評估之決策分析。
研究結果
1.盛行率和發生率
從健保資料庫分析，台灣青光眼總盛行率為千分之6.27，經以WHO2000年標準人口數校正後之年齡標準化盛行率為千分之5.99，台灣青光眼總發生率為千分之2.44，計算盛行率與年發生率比值為2.59年。依地區分類，超高度都市化地區之青光眼被診斷的機會為低度都市化地區的五倍。高度及中度都市化地區相較於低度都市化地區則分別為2.51倍及1.74倍。都市化程度對青光眼新個案診斷之地區差異也有相同影響。
(1)青光眼惡化之臨床病程觀察
馬偕醫院3360位青光眼患者中，原發性隅角開放性青光眼(primary open-angle glaucoma, POAG)佔34.70% (1166/3360)，原發性隅角閉鎖性青光眼(primary angle-closure glaucoma, PACG)佔36.31% (1220 /3360)，正常眼壓性青光眼(normal tension glaucoma, NTG)或低眼壓青光眼(low tension glaucoma, LTG)佔15.42% (518/3360)。
青光眼臨床病程之動力學變化觀察如下：當以第一次觀察到視神經盤凹陷比率大於零點六(CD1)為收斂狀態時，所有107位青光眼患者眼壓從低眼壓狀態(< 15 mmHg)轉移至中等眼壓狀態(15 to 18 mmHg)之升高速率為每人月0.1769，從中等眼壓狀態轉移至高眼壓狀態(≧18 mmHg)之升高速率為每人月0.2647；中等眼壓狀態至低眼壓狀態的降低速率為每人月0.2461，高眼壓狀態至中等眼壓狀態的降低速率為每人月0.3313。眼壓狀態間的降低轉移速率大於升高轉移速率可看出青光眼治療可有效阻止眼壓惡化升高。患者發生青光眼惡化的危險以中等眼壓狀態最高，為每人月0.45%。若以第一次發生視野平均缺損(MD)大於零來定義青光眼惡化之收斂狀態時，青光眼惡化危險以低眼壓狀態最高，其次為高眼壓狀態，中等眼壓狀態最低。
依青光眼診斷加以分類分析，PACG患者之每月眼壓降低速率稍大於升高速率。若以CD1及CD2定義收斂狀態，則進行至惡化之青光眼的轉移速率以中等眼壓狀態最高，在CD1為每人月0.48%。若以VF定義收斂狀態，則高眼壓狀態的青光眼惡化速率最快、中等眼壓次之、低眼壓最慢。
POAG患者之眼壓降低速率高於眼壓升高速率。但低眼壓與高眼壓狀態轉移至CD1青光眼惡化的危險較中等眼壓狀態者高。若以CD2定義收斂狀態時，低眼壓狀態的POAG青光眼惡化風險甚至遠高於中等眼壓與高眼壓。以VF定義的青光眼惡化風險則以高眼壓狀態最高、其次為中等眼壓、低眼壓最小。
若以CD1定義青光眼惡化，則LTG患者之眼壓上升速率接近但稍高於眼壓降低速率，且中等眼壓之青光眼惡化危險較低眼壓狀態高。以VF定義青光眼惡化時，低眼壓狀態之上升速率遠大於中等眼壓狀態降低速率，且低眼壓狀態之青光眼惡化危險較高。
依據這些升高及降低的每月轉移速率，以CD1、CD2、及VF定義青光眼惡化狀態時，所有患者發生此狀態的兩年風險各為8%，6%，及25%。
低眼壓狀態產生以CD1定義的青光眼惡化之風險以POAG最高，其次是LTG，PACG最低。中等眼壓狀態之青光眼惡化風險在POAG與LTG患者相似，均高於PACG之惡化風險。高眼壓狀態之POAG患者較PACG患者青光眼惡化風險大。若以VF定義青光眼惡化，低眼壓及中等眼壓狀態以LTG惡化風險最大、有約40%在兩年時會達到VF(MD>0)之收斂狀態；POAG風險其次為25%、LTG風險最小約15%。
用動力學曲線來看，歷經兩年追蹤治療後，停留在低眼壓、中等眼壓、高眼壓狀態的比率約分別為50%，25%，及20%；約有5%達到以CD1定義的青光眼惡化狀態。以CD2為定義時可觀察到相似情形。若以視野定義青光眼惡化，則兩年停留在低眼壓、中等眼壓、高眼壓狀態的比率約分別為40%，20%，及20%；約有20%達到以VF定義的青光眼惡化狀態。
(2)青光眼評分系統之形成
只用眼壓當青光眼唯一預測因子時，ROC分析得到的眼壓最佳切點值為16.25 mmHg，此時可得到29.7%的敏感度與87.1%的精確度。若改以考量眼壓、前房深度(anterior chamber depth, ACD)、水晶體厚度(lens thickness, LT)、生活型態因子與生化指標等因素產生的預測模型來產生青光眼評分時，此分數的最佳切點值為-3.05分，此時可達到52.7%的敏感度與82.2%的精確度。與僅用眼壓值相較，用考量較多危險因子的預測模型來預測青光眼會得到較大的ROC曲線下面積(0.733，相較於只用眼壓值的0.596)。
(3)應用多層次模式(Multi-level model)校正LASIK術後被低估之眼壓值
使用多層次模式估計出接受LASIK手術患者，其術後眼壓相較於術前會降低4.60mmHg，此數值將被用於經濟評估之計算。
(4)青光眼之經濟評估
依現行青光眼一般臨床照護方式之平均失明年為1.35年、積極介入更嚴格控制眼壓為1.226失明年、進行青光眼篩檢並積極介入更嚴格控制眼壓為0.876失明年。與依現行青光眼一般臨床照護方式之策略相較，若進行青光眼篩檢並積極介入控制眼壓時，每減少一失明年需花費新台幣48,696元(incremental cost)，而只做積極介入控制眼壓每減少一失明年所需增加之花費為新台幣120,766元。以隨機性成本效益分析可以得到相似結果。若以新台幣60,000元作為願意負擔之上限(willing to pay, WTP)，篩檢並積極介入控制眼壓相較於現行照護方式是符合成本效益的；但若單用積極介入控制眼壓而不進行篩檢，只有40%之ICER(incremental cost-effectiveness ratio)值會在新台幣60,000元以下。

關鍵字

青光眼 ； 眼壓 ； 盛行率 ； 發生率 ； 馬可夫模型 ； 成本效益分析 ； 雷射原位角膜成型術 ； Receiver Operating Characteristics Curve (ROC)

並列摘要

Introduction
Prevention of glaucoma is a challenging task because quantitative evidence from empirical data is so meager, including burden and incidence rate of a specified underlying population, dynamic changes of clinical course of glaucoma, how and whether it is cost-effective for population-based screening for glaucoma making allowance for the underestimation of IOP among subjects who had underwent LASIK at young age.
Objectives
1.Estimate age- and sex-specific prevalence rate and incidence rate of glaucoma between 1999 and 2003 in 23 areas, Taiwan.
2.Estimate types of glaucoma and elucidate clinical course of glaucoma in relation to dynamic changes of IOP, optic disc, and visual field for patients already receiving intraocular pressure-lowering treatment.
3.Determine the optimal cut-off point for glaucoma screening using a predictive model with the incorporation of IOP plus other attributes.
4.A multilevel model was applied to calibrating IOP for subjects administered by LASIK.
5.Economic appraisal for the effectiveness and cost-effectiveness of active IOP control and screening compared with routine glaucoma care.
Methods
Three empirical data sources were collected, including nationwide health insurance between 1999 and 2003 for the purpose (1); 3360 glaucoma patients visited Mackay Memorial Hospital between 2002 to 2004 enrolled and 157 subjects random sampled and clinical data gathered to achieve the purposes (2); 1350 subjects with the uptake of screening for glaucoma from Matsu in 2003 to implement the purpose (3). Three simulated datasets from literature review and two hospital-based datasets collecting IOP before and after LASIK are analyzed to fit in with the purpose (4). By the application of parameters from four parts, the framework of economic appraisal was done for comparisons of the proposed population-based screening by dint of a predictive model with the incorporation of IOP, anterior chamber depth, lens thickness, life style factors and biochemical factors and active IOP control with glaucoma care. Poisson regression model was used to model the impact of geographic area on prevalence and incident cases of glaucoma in part I. A four-state Markov model was proposed to estimate transition parameters related to three transient states between IOP status (low, intermediate, and high IOP status, representing IOP<15 mmHg, 15 to 18 mmHg, and ≧18 mmHg respectively) and to one absorbing state of the first occurrence of aggravated disease, defined by cup-disc ratios larger than 0.6 (CD1), cup-disc ratio larger than 0.7 (CD2), and mean defect of visual field larger than 0 dB (VF) in Part II. Receiver Characteristics Curve analysis was applied to develop an efficient predictive model for glaucoma by using logistic regression model for part III. The multi-level model was proposed to do a meta-analysis of calibrating IOP after the administration of LASIK in part IV. The final part was to do a decision analysis underpinning a Markov cycle tree based on parameters collected from part I to part IV.
Results
Part I Prevalence rate and incidence rate
The overall prevalence rate was 6.27 per thousand and age-standardized rate was 5.99 per thousand, increasing with age from less than 0.1% for age 0-9 years to approximately 4% for the elderly people aged 70 years or older. By geographic areas, the areas with very high urbanization level were five times more likely to be detected than those in the areas with low level. The corresponding figures were 2.51-fold for high level and 1.74-fold for moderate level.
The overall incidence rate was 2.44 per thousand. The ratio of prevalence to annual incidence rate was 2.57 yrs. The disparity in detecting new cases across area is also similar to the results as observed in prevalent cases.
Part II Clinical course of glaucoma progression
Data from Mackay Memorial Hospital found primary open angle glaucoma (POAG) accounts for 34.70% (1166/3360) of total glaucoma. Angle-closure glaucoma (ACG) is responsible for 36.31% (1220/3360), including 919 primary angle-closure glaucoma (PACG) composed of 869 chronic angle-closure glaucoma (CACG), 50 acute angle-closure glaucoma attacks, and 301 primary angle-closure (PAC). PAC explains 8.96% (301/3360). Normal tension glaucoma (NTG), the same denotation as low tension glaucoma (LTG), accounts for 15.42% (518/3360).
For total glaucoma, by using the first occurrence of cup-disc ratios larger than 0.6 (CD1) as the definition of aggravated glaucoma, namely the absorbing state, the monthly ascending rates per person were 0.1769 for the low IOP status (< 15 mmHg) to intermediate IOP status (15 to 18 mmHg) and 0.2647 for the intermediate IOP status to high IOP status (≧18 mmHg). The monthly descending rates were 0.2461 for the intermediate IOP status and 0.3313 for the high IOP status. The higher rates in the descending process strongly suggest that the deterioration of IOP can be stopped as a result of clinical treatment. The risk of progressing to aggravated glaucoma was highest in intermediate IOP status with the order of 0.45% per person-month. If aggravated glaucoma, the absorbing state, was defined by the first occurrence of visual field MD>0, the risk of progression was highest in low IOP status, followed by high IOP status, and intermediate IOP status.
For PACG, monthly descending rates were slighter greater than monthly ascending rates. The risk of progressing to aggravated glaucoma using CD1 and CD2 was highest in intermediate IOP status with the order of 0.48% per person-month in CD1 definition. Using VF as the definition of aggravated glaucoma, the highest rate for VF was seen in high IOP status followed by intermediate IOP, and low IOP status.
In POAG group, the descending rates were higher than the ascending rates. However, low IOP and high IOP status had higher risk for having CD1 than intermediate IOP status. The low IOP status had even higher risk for CD2 than intermediate or high IOP status. The risk for having VF is highest in high IOP status, then intermediate IOP and low IOP status.
For LTG group, monthly ascending rates in low IOP status was close to monthly descending rate but intermediate IOP status had higher risk for aggravated glaucoma than low IOP status. By using VF as aggravated glaucoma, monthly ascending rates in low IOP status was higher than monthly descending rates in intermediate IOP status and also low IOP status had higher risk for having VF.
Based on these monthly ascending and descending rates, for total glaucoma, two-year risk for CD1, CD2, and VF was around 8% and 6% and 25%. In patients with low IOP status the highest risk for deteriorating into aggravated glaucoma defined by CD1 was POAG followed by LTG and lowest in PACG.. For patients with intermediate IOP, the risk was identical between POAG and LTG, both of which had higher risk than PACG.. For patients with high IOP, POAG had higher risk than PACG.. By using VF as the definition of aggravated glaucoma, for patients with low IOP and intermediate IOP, the highest risk was found in LTG, 40% during two-year follow-up followed by POAG (25%) and PACG (15%).
Using kinetic curves, after two-year follow-up, the remaining patients with low IOP, intermediate IOP, and high IOP were approximately 50%, 25%, and 20%, respectively. Around 5% developed into aggravated glaucoma defined by CD1. Similar findings were noted for the outcomes defined by CD2. Using VF as the definition of aggravated glaucoma, the corresponding figures were 40%, 20% and 20%. The proportion of VF-defined aggravated glaucoma was 20%.
Part III Glaucoma scoring system development
By using IOP as only one predictor, the threshold of IOP value was 16.25 mmHg in ROC analysis. Given the optimal cut-off, sensitivity and specificity for detection of glaucoma were 29.7% and 87.1%. By using the glaucoma score developed from a predictive model with the inclusion of IOP, anterior chamber depth (ACD), lens thickness (LT), life-style factors and biochemical variables, given optimal cutoff is equal to -3.05, sensitivity and specificity were improved to 52.7% and 82.2%. The predictive model with the incorporation of more risk factors against the sole use of IOP, the area under curve will increase from 0.596 for using IOP only to 0.733 (95% CI: 0.676-0.790) for using informative predictive model.
Part IV Multi-level model to estimate the underestimation of IOP after LASIK
By using multi-level predictive model, the mean value of predicted postoperative IOP among patients undergoing LASIK procedure was 4.60 mmHg lower than preoperative IOP, which can be used for correcting IOP in the following economic appraisal.
Part V Economic evaluation
On average, blindness-years are 1.35 for glaucoma care, 1.226 for active IOP control, and 0.876 for active IOP control plus screening. The incremental cost for reducing an additional blind year was NTD 48,696 for active IOP control plus screening and NTD 120,766 for active IOP control compared with glaucoma care.
Similar results for probabilistic cost-effectiveness analysis. By using NTD 60,000 as the threshold of willingness to pay (WTP), it is cost-effective for screening plus active control IOP against glaucoma care. However, only around 40% ICER values are below the threshold of NTD 60,000 for active control IOP compared with glaucoma care.

並列關鍵字

Glaucoma ； Intraocular Pressure (IOP) ； Prevalence ； Incidence ； Markov Model ； Cost-Effectiveness Analysis (CEA) ； Laser in Situ Keratomileusis (LASIK) ； Receiver Operating Characteristics Curve (ROC)

參考文獻

1. Quigley HA. Number of people with glaucoma worldwide. Br J Ophthalmol 1996;80:389-93.

2. Foster PJ, Buhrmann R, Quigley HA and Johnson GJ. The definition and classification of glaucoma in prevalence surveys. Br. J. Ophthalmol 2002;86:238-42.

3. Katz J. Scoring systems for measuring progression of visual field loss in clinical trials of glaucoma treatment. Ophthalmol 1999;106:391–395.

4. Asman P, Heijl A. Diffuse visual field loss and glaucoma. Acta Ophthalmol Scand 1994;72:303–8.

5. The Eye Diseases Prevalence Research Group. Prevalence of open-angle glaucoma among adults in the United States. Arch Ophthalmol 2004;122:532-53.