Introduction
In the era of antiretroviral therapy (ART), an emerging epidemic of cardiometabolic diseases (CMDs) (i.e. hypertension, diabetes, dyslipidemia, and cardiovascular diseases; CVDs) has become a critical issue [1–5] [6–8] economic burden of CMDs remains scarce.
Elevated comorbid CMDs in the HIV-infected population are complicated by multiple factors including persistent immune activation and inflammation due to HIV itself which can promote atherosclerosis, adverse effects of ART on lipid metabolism, and some traditional risk factors (e.g. smoking) [5,9,10] [11–13] [8,14–18]
Despite the benefits of timely ART initiation at HIV diagnosis as highlighted by the recent guidelines [19,20] [21,22] [23,24] [25]
Against these backgrounds, we aimed to investigate the economic burden associated with recently emerging CMD epidemics from an Asian HIV-infected population. We further assessed the roles of early ART scale-up (i.e. aggressive or nonaggressive use) and ADI occurrence (i.e. present or absent) at HIV diagnosis, as clinical indicators for HIV management, on this emerging burden to provide clinical and policy implications for the early scale-up of ART at HIV diagnosis for early prevention of incident CMDs.
Methods
Data sources
A population-based longitudinal cohort of HIV/AIDS population in the period 2000–2011 was extracted from the National Health Insurance Research Database (NHIRD), a claims-based database under the National Health Insurance (NHI) program which enrolls over 99% of Taiwanese population and covers all medical services that are provided by the healthcare institutions [26] [27,28]
Our study cohort was assembled from an earlier period in 2000–2011, in which the early scale-up of ART at HIV diagnosis was not fully recommended yet and thus the real-world variations in ART use existed. During 2000–2011 in Taiwan, ART was recommended to initiate for HIV-infected patients with CD4+ cell counts less than 500 cells/μl before year 2008 but for those with CD4+ cell counts less than 200 cells/μl after year 2008. By utilizing the study cohort derived from this earlier period, we were able to observe different levels of ART use and further assess their associated cost impacts.
Study cohort selection
Patients diagnosed with HIV during 2003–2008 were identified, which allowed for at least a 3-year follow-up observation until 2011. Patients with any CMDs at the year before or at HIV diagnosis were excluded. CMDs include hypertension, diabetes, dyslipidemia, and CVDs (cerebrovascular diseases, acute myocardial infarction, heart failure, arrhythmia, ischemic heart disease, and cardiogenic shock).
The study cohort was further stratified by two variables measured at the first year of HIV diagnosis. First, the medication possession ratio (MPR), calculated by summing the days of prescriptions of ART in the first year of HIV diagnosis divided by 365 days, was used to quantify the ART utilization of each individual. The MPR is a common measure for medication utilization, which is associated with physicians’ prescribing and patients’ health behaviors (e.g. medication refill or adherence); an MPR value at least 0.8 is typically considered as a cutoff for good or optimal medication adherence in claims-based research [29,30] https://links.lww.com/QAD/B664
Study variables
The operational definitions of ADIs, CMDs, and comorbidities were provided in Supplementary Table 2, https://links.lww.com/QAD/B664
Statistical analyses
Descriptive analyses were performed for patient characteristics. The differences between study groups were tested by analysis of variance and chi-square test or Fisher's exact test, as appropriate. Incidence rates of CMDs were calculated as the total numbers of patients with newly-developed CMDs divided by total followed person-years. In addition, cumulative incidence curves were plotted by cumulative hazard function (SAS LIFETEST procedure), and the between-group difference in the cumulative incidence was tested by the log-rank test. Furthermore, Cox proportional hazard model was applied to evaluate the risk of developing CMDs beyond the first year of HIV diagnosis between study groups, with adjustment for between-group differences at the first year of HIV diagnosis, including demographics, comorbidities and CMDs.
A generalized linear model was used to estimate economic burden of CMDs. We modeled the annual medical cost as a function of patients’ demographics, HIV duration, comorbidities and CMDs by fitting generalized estimating equations (GEE), which account for the dependence of yearly measured cost data within each patient during the follow-up. HIV duration and disease conditions were measured yearly and treated as time-dependent variables in the model. We further tested the relative cost impacts associated with CMDs between study groups (SAS SLICE procedure). ART (low) and ADI (+) patients were chosen as the reference group for comparison because they were likely to be patients with the most insufficient HIV control at HIV diagnosis, as indicated by less aggressive ART and ADI occurrence.
To account for the skewed data distribution, the annual medical costs were log-transformed in the GEE analysis. The coefficients from the analysis were then exponentially transformed back to the original scale, to express the cost estimates in the form of multipliers. A cost multiplier represents the economic impact of a given patient's characteristic; for example, ‘By how much (multiplier) will the annual medical costs increase if a given event occurs compared with a nonevent?’
In this multiplicative GEE model, the base case was determined as a male diagnosed with HIV at the age of 35–49 years, without any comorbidities and CMDs, with an MPR of ART less than 0.8, and experiencing ADIs at the first year of HIV diagnosis [i.e. ART (low) and ADI (+) group]. To calculate the relative increase in annual medical costs for a given patient with characteristics other than those of the base-case patient, the annual medical cost can be calculated as the product of the annual medical cost of the base-case patient multiplied by the cost multipliers of the given demographics and disease conditions. SAS 9.4 (SAS Institute, Cary, North Carolina, USA) was used for all analyses abovementioned. A P value of less than 0.05 indicated a statistically significant difference.
Results
Table 1 shows patients’ demographics and disease conditions measured at the first year of HIV diagnosis. The average follow-up period from the overall 10 693 study patients was 4.55 years. The most common ADI at HIV diagnosis was Pneumocystis jiroveci pneumonia infection (2.96%), followed by candidiasis (2.49%), and tuberculosis (1.99%) (Supplementary Table 3, https://links.lww.com/QAD/B664
Table 1: Patients’ demographics and clinical characteristics at first year of HIV diagnosis.
Table 2 and Supplementary Table 4, https://links.lww.com/QAD/B664
Table 2: Incidence rates (event per 1000 person-years) of hypertension, diabetes, dyslipidemia, and cardiovascular diseases in the follow-up period.
Figure 1 shows the cumulative incidences of CMDs. There is no significant difference in the cumulative incidences of hypertension and diabetes between study groups, except for a significantly lower cumulative incidence of hypertension in the ART (low) and ADI (−) group compared with other groups. However, the cumulative incidence of dyslipidemia was significantly higher in the patients with ART (high). Moreover, the ART (low) and ADI (+) group had a higher cumulative incidence of CVDs compared with other groups. The trend of the Cox model results on differences in the risks of developing CMDs between study groups (Supplementary Table 5, https://links.lww.com/QAD/B664 Fig. 1 .
Fig. 1: Cumulative incidences of hypertension, diabetes, dyslipidemia, and cardiovascular disease over 5 years of follow-up after HIV diagnosis.
Table 3 shows the results of GEE analysis. The baseline cost of $1993 is interpreted as the mean annual medical cost for the base-case patient. The cost multipliers represent the magnitude of the cost impact associated with the given patients’ characteristics and disease conditions. Compared with the cost for the ART (low) and ADI (+) group, that for the ART (low) and ADI (−) group was significantly lower by 73%, and those for the ART (high) and ADI (−) and ART (high) and ADI (+) groups were significantly higher by 34 and 38%, respectively, under the base-case setting. The annual costs significantly increased by 1.31, 1.43, 1.15, and 2.27-fold when developing hypertension, diabetes, dyslipidemia, and CVDs, respectively.
Table 3: Economic burden associated with individual HIV-infected patients’ demographics and clinical conditions.
For patients with any of the characteristics or disease conditions listed in Table 3 , the annual medical cost estimate is the product of the baseline cost and the multipliers of each corresponding characteristic. For example, a male patient diagnosed with HIV infection at 30 years old, with MPR of ART more than 0.8, without ADI occurrence at the first year of HIV diagnosis [i.e. ART (high) and ADI (−) group], and without any comorbidities had a 1.26-fold (0.94 × 1.34) higher annual medical cost compared with that of the base-case patient.
Figure 2 a shows that the ART (low) and ADI (−) group had significantly lower cost burdens of hypertension, diabetes, and dyslipidemia, and an insignificantly lower cost burden from CVDs, compared with the ART (low) and ADI (+) group. For example, when hypertension occurred, its economic burden for the ART (low) and ADI (−) group was only 0.62-fold that for the ART (low) and ADI (+) group. The ART (high) and ADI (−) group had an insignificantly lower cost burden of hypertension, and the economic burden of diabetes, dyslipidemia, and CVDs for the ART (high) and ADI (−) group was significantly lower by 42, 30, and 31% compared with the ART (low) and ADI (+) group, respectively (Fig. 2 b). Moreover, the cost burdens of all components of CMDs for the ART (high) and ADI (+) group were all insignificantly lower than those for the ART (low) and ADI (+) group (Fig. 2 c).
Fig. 2: Relative conditional cost impacts of cardiometabolic diseases between study groups [the antiretroviral therapy (low) and AIDS-defining illnesses (+) group used as the reference group].
Discussion
This is the first study to provide detailed economic burden data for various CMDs among HIV-infected population. Previous studies either only focused on specific CMDs or analyzed the crude medical costs of CMDs, and thus did not reveal a whole picture of economic burden for various CMDs in HIV-infected patients, and the validity of the crude estimates may be questionable [31,32]
We found that CMDs contributed a considerable economic burden to HIV-infected population, with CVDs being the leading cost driver. Of note, the economic burden of CMDs varied significantly by the study groups stratified by the level of ART use and the occurrence of ADI at HIV diagnosis. When CMDs occurred, compared with the patients with insufficient HIV control in terms of less aggressive ART and high inflammation status [ART (low) and ADI (+) group], those with aggressive ART and low inflammation status [ART (high) and ADI (−) group] had a significantly lower cost burden associated with all components of CMDs. Hence, these indicate that the early scale-up of ART at HIV diagnosis is critical for mitigating considerable economic burden of developing CMDs.
Our results further show that the magnitude of reduced economic burden of CMDs for the ART (high) and ADI (−) vs. ART (low) and ADI (+) group was significantly larger than that for the ART (high) and ADI (+) vs. ART (low) and ADI (+) group (Fig. 2 ). For example, compared with the ART (low) and ADI (+) group, developing CVDs in the ART (high) and ADI (−) group cost 31% less, but that for the ART (high) and ADI (+) group cost only 14% less and not statistically significant. This suggests that early ART scale-up at HIV diagnosis before the ADI occurrence is crucial for ensuring the potential benefits of ART on economic burden of developing CMDs. If ART is delayed until ADIs occur, it might not be as effective in reducing the economic burden of CMDs as when it is initiated before ADIs occurrence.
In addition, we found that patients with insufficient HIV control [ART (low) and ADI (+) group] at the first year of HIV diagnosis could be at a high risk of developing CVDs (Fig. 1 and Supplementary Table 4, https://links.lww.com/QAD/B664 + cell count-guided intermittent ART arm compared with the continuous ART arm, suggesting the potential benefit of reduced CVD risk associated with viral suppression owing to continuous/persistent ART [33] economic burden of CVD epidemics in the HIV-infected population.
Despite its potential economic benefits, aggressive ART may have some downsides. First, increased ART may lead to a high upfront medical cost to healthcare system; the patients with ART (high) generally had significantly higher medical costs at baseline compared with those with ART (low) (Table 3 ). Previous studies also showed that higher healthcare costs in HIV-infected patients compared with the general population were mainly contributed by ART use [34]
Second, aggressive ART might be associated with a higher rate of developing dyslipidemia, regardless of ADI occurrence (Fig. 1 and Supplementary Table 4, https://links.lww.com/QAD/B664 [18,35] economic burden of dyslipidemia among the patients with ART (high) was not as high as that in the patients with ART (low) and ADI (+) (Fig. 2 ). This means that although increased ART use may put patients at a high risk of developing dyslipidemia, the contribution of dyslipidemia to medical costs among patients better treated with ART may be less than that for those poorly controlled by ART. This may be in part explained as follows. Owing to relatively regular follow-up or medical visits, the patients with ART (high) might get more opportunities to be monitored on lipid-metabolic profiles and thus early identified with lipid-metabolic problems. Effective preventive strategies could then be started early. This may decrease the future economic burden associated with lipid or metabolic problems once they occurred, compared with those without regular care [i.e. patients with ART (low)]. Therefore, with rising awareness of managing dyslipidemia and metabolic diseases in the HIV-infected population [27] [36,37]
Therefore, the important clinical implications from this study are that although aggressive control by ART might increase the baseline medical costs and risk of dyslipidemia in HIV-infected patients, the economic burden of CMDs is likely to be mitigated, particularly among the patients with the early scale-up of ART and without the ADI occurrence at HIV diagnosis, compared with those without aggressive ART and having ADIs. However, the lipid-metabolic abnormalities due to ART should be carefully managed. Moreover, future research is needed to determine whether the high upfront costs with the scale-up of ART can be offset by the future savings from reduced economic burden of CMDs (especially costly CVDs).
Several limitations should be acknowledged. First, indirect costs (e.g. productivity loss due to CVD-related disability) were not considered in analysis. However, if the costs of productivity loss due to CVDs were analyzed from the societal perspective, the early ART scale-up at HIV diagnosis might be even more beneficial against future CVDs-related economic burden . Second, like other studies using administrative claims data, viral load and CD4+ cell counts were lacking and thus not adjusted in the analyses. However, the ADI occurrence may be a surrogate for HIV inflammation. Moreover, the ART utilization and HIV duration could be proxies for the severity of HIV infection. Therefore, with adjustment for these variables, we minimized the unmeasured confounding bias from lacking detailed clinical data. Future research with detailed clinical information for HIV infection is needed to corroborate our findings. Third, we restricted our follow-up to 5 years after HIV diagnosis because of insufficient follow-up period in our database, which may limit our study inference to the first 5 years following the diagnosis. Fourth, we did not differentiate individual ART classes in the analyses. However, we analyzed the utilization pattern by ART class and found that the distribution of different ART classes was similar across groups (Supplementary Table 6, https://links.lww.com/QAD/B664 economic burden of CMDs vary by the level of HIV control at the first year of HIV diagnosis, and thus not designed for exploring any causal relationships between individual CMDs (e.g. increased prevalence of hypertension and risk of developing CVDs), the use of ART and incidence of CMDs, or assessing the impact of different analytic strategies on study results (e.g. different cost transformation methods). Sixth, we were unable to explore any behavioral reasons associated with different MPR values due to the lack of information of prescriber's or patient's medication-taking behaviors in claims data. Seventh, the study was based on a relatively earlier period (2000–2011). Nonetheless, the study results are still clinically relevant to current real-world settings, where patients often delay the initiation of ART at HIV diagnosis or have insufficient adherence to ART following the diagnosis, despite the early scale-up of ART at HIV diagnosis is clearly recommended nowadays. In fact, the proportion of late presenters to HIV care (defined as having any ADI before or at HIV diagnosis) remains around 30% of total HIV-infected population in Taiwan in recent years [38]
In conclusion, our results suggest that the early scale-up of ART at HIV diagnosis, especially before the ADI occurrence, along with the careful monitoring and early prevention of lipid-metabolic changes, is crucial for minimizing the future economic burden of CMD epidemics in the HIV-infected population. Future cost-effectiveness research is needed to assess the benefit of early scale-up of ART at HIV diagnosis against this emerging CMD burden.
Acknowledgements
C.T.Y., H.S.T., and H.T.O. performed the research; C.T.Y., H.S.T., C.T.L., H.T.O., and N.Y.K. designed the research study; C.T.Y., and H.T.O. analyzed the data; C.T.Y., H.T.O., and S.K. wrote the article.
Conflicts of interest
There are no conflicts of interest.
References
1. Shah AS, Stelzle D, Lee KK, Beck EJ, Alam S, Clifford S, et al.
Global burden of atherosclerotic cardiovascular disease in people living with HIV .
Circulation 2018; 138:1100–1112.
2. Triant VA, Lee H, Hadigan C, Grinspoon SK.
Increased acute myocardial infarction rates and cardiovascular risk factors among patients with human immunodeficiency virus disease .
J Clin Endocrinol Metab 2007; 92:2506–2512.
3. Feinstein MJ, Bahiru E, Achenbach C, Longenecker CT, Hsue P, So-Armah K, et al.
Patterns of cardiovascular mortality for HIV-infected adults in the United States: 1999 to 2013 .
Am J Cardiol 2016; 117:214–220.
4. Hsue PY, Waters DD.
Time to recognize HIV infection as a major cardiovascular risk factor .
Circulation 2018; 138:1113–1115.
5. Hsue PY, Lo JC, Franklin A, Bolger AF, Martin JN, Deeks SG, et al.
Progression of atherosclerosis as assessed by carotid intima–media thickness in patients with HIV infection .
Circulation 2004; 109:1603–1608.
6. Riddler SA, Li X, Chu H, Kingsley LA, Dobs A, Evans R, et al.
Longitudinal changes in serum lipids among HIV-infected men on highly active antiretroviral therapy .
HIV Med 2007; 8:280–287.
7. Bijker R, Choi JY, Ditangco R, Kiertiburanakul S, Lee MP, Siwamogsatham S, et al.
Cardiovascular disease and cardiovascular disease risk in HIV-positive populations in the Asian region .
Open AIDS J 2017; 11:52–66.
8. Echecopar-Sabogal J, D’Angelo-Piaggio L, Chanamé-Baca DM, Ugarte-Gil C.
Association between the use of protease inhibitors in highly active antiretroviral therapy and incidence of diabetes mellitus and/or metabolic syndrome in HIV-infected patients: a systematic review and meta-analysis .
Int J STD AIDS 2017; 29:443–452.
9. Benard A, Tessier JF, Rambeloarisoa J, Bonnet F, Fossoux H, Neau D, et al.
HIV infection and tobacco smoking behaviour: prospects for prevention? ANRS CO3 Aquitaine Cohort, 2002 .
Int J Tuberc Lung Dis 2006; 10:378–383.
10. Nou E, Lo J, Grinspoon SK.
Inflammation, immune activation, and cardiovascular disease in HIV .
AIDS 2016; 30:1495–1509.
11. Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the use of antiretroviral agents in adults and adolescents living with HIV. 2019. Department of Health and Human Services. Available at
https://aidsinfo.nih.gov/contentfiles/lvguidelines/AdultandAdolescentGL.pdf . [Accessed on 19 January 2019].
12. The INSIGHT START Study Group
Initiation of antiretroviral therapy in early asymptomatic HIV infection .
N Engl J Med 2015; 373:795–807.
13. El-Sadr WM, Lundgren J, Neaton JD, Gordin F, Abrams D, et al. Strategies for Management of Antiretroviral Therapy (SMART) Study Group
CD4 + cell count-guided interruption of antiretroviral treatment .
N Engl J Med 2006; 355:2283–2296.
14. Nou E, Lo J, Hadigan C, Grinspoon SK.
Pathophysiology and management of cardiovascular disease in patients with HIV .
Lancet Diabetes Endocrinol 2016; 4:598–610.
15. Thienemann F, Sliwa K, Rockstroh JK.
HIV and the heart: the impact of antiretroviral therapy: a global perspective .
Eur Heart J 2013; 34:3538–3546.
16. Cruciani M, Zanichelli V, Serpelloni G, Bosco O, Malena M, Mazzi R, et al.
Abacavir use and cardiovascular disease events .
AIDS 2011; 25:1993–2004.
17. Young J, Xiao Y, Moodie EE, Abrahamowicz M, Klein MB, Bernasconi E, et al.
Effect of cumulating exposure to abacavir on the risk of cardiovascular disease events in patients from the Swiss HIV Cohort Study .
J Acquir Immune Defic Syndr 2015; 69:413–421.
18. Tsai FJ, Cheng CF, Lai CH, Wu YC, Ho MW, Wang JH, et al.
Effect of antiretroviral therapy use and adherence on the risk of hyperlipidemia among HIV-infected patients, in the highly active antiretroviral therapy era .
Oncotarget 2017; 8:106369–106381.
19. Havlir DV, Buchbinder SP.
Ending AIDS in the United States-if not now, when? .
JAMA Intern Med 2019; 179:1165–1166.
20. Saag MS, Benson CA, Gandhi RT, Hoy JF, Landovitz RJ, Mugavero MJ, et al.
Antiretroviral drugs for treatment and prevention of HIV infection in adults: 2018 recommendations of the International Antiviral Society-USA Panel .
JAMA 2018; 320:379–396.
21. Gwadz M, Applegate E, Cleland C, Leonard NR, Wolfe H, Salomon N, et al.
HIV-infected individuals who delay, decline, or discontinue antiretroviral therapy: comparing clinic- and peer-recruited cohorts .
Front Public Health 2014; 2:81.
22. Adedinsewo DA, Wei SC, Robertson M, Rose C, Johnson CH, Dombrowski J, et al.
Timing of antiretroviral therapy initiation in a nationally representative sample of HIV-infected adults receiving medical care in the United States .
AIDS Patient Care STDS 2014; 28:613–621.
23. Lin KY, Cheng CY, Li CW, Yang CJ, Tsai MS, Liu CE, et al.
Trends and outcomes of late initiation of combination antiretroviral therapy driven by late presentation among HIV-positive Taiwanese patients in the era of treatment scale-up .
PLoS One 2017; 12:e0179870.
24. Jeong SJ, Italiano C, Chaiwarith R, Ng OT, Vanar S, Jiamsakul A, et al.
Late presentation into care of HIV disease and its associated factors in Asia: results of TAHOD .
AIDS Res Hum Retroviruses 2016; 32:255–261.
25. Low A, Gavriilidis G, Larke N, B-Lajoie MR, Drouin O, Stover J, et al.
Incidence of opportunistic infections and the impact of antiretroviral therapy among HIV-infected adults in low- and middle-income countries: a systematic review and meta-analysis .
Clin Infect Dis 2016; 62:1595–1603.
26. Hsieh CY, Su CC, Shao SC, Sung SF, Lin SJ, Kao Yang YH, et al.
Taiwan's National health insurance research database: past and future .
Clin Epidemiol 2019; 11:349–358.
27. Ou HT, Chang KC, Li CY, Yang CT, Ko NY, et al.
Intensive statin regimens for reducing risk of cardiovascular diseases among human immunodeficiency virus-infected population: a nation-wide longitudinal cohort study 2000–2011 .
Int J Cardiol 2017; 230:592–598.
28. Chen YC, Lin CY, Strong C, Li CY, Wang JS, Ko WC, et al.
Sleep disturbances at the time of a new diagnosis: a comparative study of human immunodeficiency virus patients, cancer patients, and general population controls .
Sleep Med 2017; 36:38–43.
29. Barai N, Monroe A, Lesko C, Lau B, Hutton H, Yang C, et al.
The association between changes in alcohol use and changes in antiretroviral therapy adherence and viral suppression among women living with HIV .
AIDS Behav 2017; 21:1836–1845.
30. Messou E, Chaix ML, Gabillard D, Minga A, Losina E, Yapo V, et al.
Association between medication possession ratio, virologic failure and drug resistance in HIV-1-infected adults on antiretroviral therapy in Côte d’Ivoire .
J Acquir Immune Defic Syndr 2011; 56:356–364.
31. Gallant J, Hsue P, Budd D, Meyer N.
Healthcare utilization and direct costs of noninfectious comorbidities in HIV-infected patients in the USA .
Curr Med Res and Opin 2017; 34:13–23.
32. Guaraldi G, Zona S, Menozzi M, Carli F, Bagni P, Berti A, et al.
Cost of noninfectious comorbidities in patients with HIV .
Clinicoecon Outcomes Res 2013; 5:481–488.
33. Phillips AN, Carr A, Neuhaus J, Visnegarwala F, Prineas R, Burman WJ, et al.
Interruption of antiretroviral therapy and risk of cardiovascular disease in persons with HIV-1 infection: exploratory analyses from the SMART trial .
Antivir Ther 2008; 13:177–187.
34. Ritchwood TD, Bishu KG, Egede LE.
Trends in healthcare expenditure among people living with HIV/AIDS in the United States: evidence from 10 Years of nationally representative data .
Int J Equity Health 2017; 16:188.
35. Souza SJ, Luzia LA, Santos SS, Rondó PHC.
Lipid profile of HIV-infected patients in relation to antiretroviral therapy: a review .
Rev Assoc Med Bras 2013; 59:186–198.
36. Quercia R, Roberts J, Martin-Carpenter L, Zala C.
Comparative changes of lipid levels in treatment-naive, HIV-1-infected adults treated with dolutegravir vs. efavirenz, raltegravir, and ritonavir-boosted darunavir-based regimens over 48 weeks .
Clin Drug Investig 2015; 35:211–219.
37. Sension M, Deckx H.
Lipid metabolism and lipodystrophy in HIV-1-infected patients: the role played by nonnucleoside reverse transcriptase inhibitors .
AIDS Rev 2015; 17:21–36.
38. Taiwan Ministry of Health and Welfare. HIV/AIDS 5-year control program. 2016. Available at:
https://www.cdc.gov.tw/File/Get/hG9uzlwre8PEq0P-KVT37A . [Accessed on 17 November 2019].
