https://orcid.org/0000-0002-5439-7133
Figures
Abstract
Background
From 2004 onwards, the Chinese government has freely offered complimentary Chinese herbal medicine (CHM) to Chinese HIV/AIDS patients, alongside the prescribed first line therapy of highly active antiretroviral therapy (HAART). Thus, we aimed to explore the effectiveness and safety of CHM for patients with HIV/AIDS.
Methods
The data from the Guangxi pilot database and antiviral treatment sites database have been respectively developed into two datasets in this prospective cohort real-world study, the CHM combined HAART group (the integrated group) and the HAART group. A 1:1 propensity score matching (PSM) was performed and the longitudinal data were analyzed using a generalized estimating equation (GEE) model with an autocorrelation matrix and log link function attached to the Gamma distribution.
Results
A final sample of 629 patients, 455 and 174 in the integrated group and HAART group respectively, were obtained from the full dataset. As covariates for PSM, gender, age, baseline CD4+ and CD4+/ CD8+ were assessed based on the results of the logistic regression analyses. Following PSM, 166 pairs from the full dataset were matched successfully, with 98 pairs in the baseline CD4+ > 200 subgroup, and 55 pairs in the baseline CD4+ ≤ 200 subgroup. In the full dataset, HAART group achieved higher CD4+ count (OR = 1.119, 95%CI [1.018, 1.230]) and CD4+/CD8+ ratio (OR = 1.168, 95%CI [1.045, 1.305]) than the integrated group, so did in the CD4+ > 200 subgroup. For the CD4+ ≤ 200 subgroup, the CD4+ (OR = 0.825, 95%CI [0.694, 0.980]) and CD4+/CD8+ (OR = 0.826, 95%CI [0.684, 0.997]) of the integrated group were higher than those of the HAART group. The safety outcomes showed that there were no significant differences in BUN, ALT and AST levels between the groups but Cr showed significantly higher levels in HAART groups of all three datasets.
Conclusions
Compared to HAART alone, CHMs combined with HAART had better effects in improving the immune function of HIV/AIDS in patients with baseline CD4+ count ≤ 200. The results of the two subgroups are in opposite directions, and chance does not explain the apparent subgroup effect. A study with larger sample size and longer follow-up period is warranted in order to increase study credibility.
Citation: Li J, Shen C, Liu Z-W, Pu F-L, Cao S-H, Zhang Y, et al. (2024) Chinese herbal medicine for patients living with HIV in Guangxi province, China: A propensity score matching analysis of real-world data. PLoS ONE 19(9): e0304332. https://doi.org/10.1371/journal.pone.0304332
Editor: Satish Rojekar, Icahn School of Medicine at Mount Sinai Department of Pharmacological Sciences, UNITED STATES
Received: February 3, 2024; Accepted: May 6, 2024; Published: September 6, 2024
Copyright: © 2024 Li et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because the Guangxi pilot database is currently inactive due to updates in national policies. Data are available from the Centre for Evidence Based Chinese Medicine, Beijing University of Chinese Medicine (contact via bucmxzyxzx@163.com) for researchers who meet the criteria for access to confidential data.
Funding: This work was supported by the General Project of the National Natural Science Foundation of China “Methodological study on curative effect evaluation based on registered database on Chinese medicine for HIV/AIDS: a Bayesian and propensity score analysis” (No. 81673828). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Abbreviation: AIDS, Acquired immunodeficiency syndrome; ALT, Alanine transaminase; AST, Aspartate aminotransferase; BUN, Blood urea nitrogen; CHM, Chinese herbal medicine; Cr, Creatinine; FIs, Fusion inhibitors; GEE, Generalized estimating equation; HAART, Highly active antiretroviral therapy; HIV, Human immunodeficiency virus; INSTIs, Integrase inhibitors; NFCHMP, National Free CHM HIV/AIDS Treatment Program; NNRTIs, Non-nucleoside reverse transcriptase inhibitors; NRTIs, Nucleoside reverse transcriptase inhibitors; PIs, Protease inhibitors; PSM, Propensity score matching; RCT, Randomized controlled trial; RWS, Real-world study; SATCHM, State Administration of Traditional Chinese Medicine; TCM, Traditional Chinese medicine; UNAIDS, Joint United Nations Programme on HIV/AIDS
Background
Acquired immunodeficiency syndrome (AIDS) is a chronic and fatal infectious disease caused by the human immunodeficiency virus (HIV). HIV destroys the white blood cells called CD4+ cells, weakening a person’s immunity against opportunistic infections, such as tuberculosis and fungal infections, severe bacterial infections and some cancers [1]. According to the Joint United Nations Programme on HIV/AIDS (UNAIDS), HIV/AIDS has driven 37.7 million people to be affected by the end of 2020, with 1.5 million newly HIV infected in that year. About 27.5 million people are now receiving antiretroviral therapy. which is indicated in cases of HIV [2]. Currently, there are six classes of AIDS agents in the world, namely nucleoside reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), protease inhibitors (PIs), integrase inhibitors (INSTIs), fusion inhibitors (FIs) and CCR5 inhibitors [3]. Long-term use of NRTI agents could be attributed to producing adverse reactions of hyperlactatemia and lactic acidosis, neuropathy, pancreatitis and lipoatrophy [4]. The most commonly used NNRTIs, efavirenz is known to have a significant side-effect profile that includes neuropsychiatric toxicity, rash, hyperlipidemia and elevated transaminases [5, 6]. According to the world health organization’s HIV Drug Resistance Report in 2021, more and more nations were reaching the 10% threshold of pretreatment HIV drug resistance (HIVDR) to NNRTIs and it was found up to 3 times more likely in people who had previous exposure to antiretroviral agents [7]. PIs may likewise elicit side effects encompassing metabolic abnormalities including dyslipidemia (primarily triglycerides), insulin resistance, hyperglycemia, and lipodystrophy [8].
Chinese herbal medicine (CHM) was first used to treat HIV-infected people in 1987, when traditional Chinese medicine (TCM) practitioners from China provided medical assistance in Tanzania, Africa [9]. Given its promising effectiveness and high safety profile, CHM has been used in the treatment of HIV/AIDS for more than 30 years as an alternative and complementary therapy to highly active antiretroviral therapy (HAART) [10]. The application of TCM has also been proven to significantly improve patients’ clinical symptoms and signs, improve their working capacity and quality of life, safeguard their immune system, and postpone the onset of AIDS [11–14]. Additionally, the synergistic administration of CHM and antiviral agents can reduce some adverse reactions to antiviral agents [11]. A pilot project in China, the National Free CHM HIV/AIDS Treatment Program (NFCHMP), was launched in 2004 and extended rapidly. NFCHMP was supported by the State Administration of Traditional Chinese Medicine (SATCHM) and the Ministry of Finance, and the program has provided free TCM treatment to tens of thousands of HIV/AIDS patients.
It is apparent that real-world studies (RWSs) are not constrained by the small sample size or the strict inclusion criteria, such as the exclusion of children or the elderly, as are randomized controlled trials (RCTs). As a result, it can contribute to a broad evaluation of treatment modes and external effectiveness [15]. A number of RWSs have been conducted on HIV. For instance, a multicentral RWS by Santinelli et al. assessed the real-life effectiveness, tolerability, and safety of long-term Raltegravir use in elderly HIV infected patients [16]. Similarly, Okoli et al described the actual use and effectiveness of Dolutegravir-based regimens in HIV patients treated in the United Kingdom [17]. A retrospective cohort study was conducted in Henan province on basis of the NFCHMP database and it demonstrated that CHM could decrease the disease progression, reduce the mortality of people living with HIV, and improve life expectancy. However, the predominant limitation was that they chose the contemporaneous world mortality rate as a comparison [18]. A 7-year observational study indicated long-term utilisation of CHM could keep up or impede the pace of CD4+ cell counts declining. However, this study did not address bias, potential confounders and the possibility that results may have occurred by chance [19]. RWSs employing TCM to treat HIV are, whereas, limited in small sample size, do not address confounding factors, or ignored individual disease progression.
RWS, on the other hand, may be accompanied by more confounding factors than RCTs. Thus, propensity score-based approaches have been developed to reduce or remove the factors [20]. The propensity score represents the probability of assigning treatment conditions on observed baseline attributes. Furthermore, Liang and Zeger proposed the generalized estimating equation (GEE) to analyze real-world data (RWD), which was developed just on the basis of the generalized linear model (GLM) and enhance GLM to accommodate the modelling of correlated data. GEE is appropriate for complete data or missing data at random [21]. Therefore, we aimed to analyze the longitudinal data using propensity score matching (PSM) and GEE to explore the effectiveness and safety of CHM for patients with HIV/AIDS.
Methods
Ethics approval and consent to participate
This study was approved by the ethics committee of the Beijing University of Chinese Medicine (BZZYYDX-LL20160215).
Data source
The ethics committee of the Beijing University of Chinese Medicine approved this study before data collection began (BZZYYDX-LL20160215). The prospective cohort study was based on two registration databases, the Guangxi pilot database of the NFCHMP (hereinafter referred to as Guangxi pilot database) and the antiviral treatment site database of Ruikang Hospital affiliated with the Guangxi University of Traditional Chinese Medicine (hereinafter referred to as antiviral treatment sites database). The study participants provided their written informed consent and permitted the use of their medical information.
Participants
The real-world data was composed of two sets, namely the CHM combined with the HAART group (integrated group for short) and the HAART group. The participants in the integrated group were sourced from the Guangxi pilot database and those in the HAART group were sourced from the antiviral treatment sites database. Over the course of 36 months, all participants were followed up every three months.
Eligible participants were those diagnosed with HIV/AIDS and receiving HAART treatment between 2004 and 2016. A complete set of included case data should be provided with all necessary information. Participants were excluded if they did not have baseline characteristics (gender, age, marital status, possible route of infection) or CD4+ baseline data. In the case of participants with all follow-up data missing within 36 months, they were excluded from the study.
Exposure factors
This study was divided into the integrated group and the HAART group based on whether participants received CHMs treatment or not. There was no limit to the duration of the exposure. CHMs were available in three forms: Tangcao Tablets, containing Geranium wilfordii Maxim., Lonicera japonica Thunb., Trichosanthes kirilowii Maxim., Bupleuri Radix, Elsholtzia ciliata, Punica granatum L., Astragalus membranaceus (Fisch.) Bunge, Glycyrrhizae Radix et Rhizoma, Bombax ceiba L., Millettia dielsiana, Carthamus tinctorius L., Oryza sativa L., Terminalia chebula Retz., Scleromitrion diffusum (Willd.) R.J. Wang, Trapa bispinosa Roxb., Ginkgo biloba L., Portulaca oleracea L., Neopicrorhiza scrophulariiflora (Pennell) D. Y. Hong, Solanum nigrum L. and Buthus martensii Karsch; Qingdu Capsules/Granules, mainly composed of Astragalus membranaceus (Fisch.) Bunge, Atractylodes Lancea (Thunb.) DC., Polyrhachis vicina Roger, Scutellaria baicalensis Georgi, Poria cocos (Schw.) Wolf, Ganoderma lucidum (Curtis) P. Karst., Andrographis paniculata (Burm. f.) Wall. ex Nees in Wallich and Gynostemma pentaphyllum (Thunb.) Makino, and Shenling Fuzheng Capsules, mainly composed of Codonopsis pilosula (Franch.) Nannf., Astragalus membranaceus (Fisch.) Bunge, Atractylodes macrocephala Koidz., Gynostemma pentaphyllum (Thunb.) Makino, Polyrhachis vicina Roger and Ganoderma lucidum (Curtis) P. Karst. A HAART regimen was administered in accordance with the National Free Antiretroviral Treatment Manual Book. The free drugs used in 2004 included domestic drugs Zidovudine, Stavudine, Didanosine, Nevirapine, and imported drugs Lamivudine, Efavirenz, and Indinavir. As of 2012, Tenofovir was included among the first-line drugs.
Outcomes
Over a period of 36 months, the primary outcome was CD4+ T cell count. Secondary outcomes were CD8+ T cell count, CD4+/ CD8+ ratio, the level of creatinine (Cr), blood urea nitrogen (BUN), alanine transaminase (ALT) and aspartate aminotransferase (AST).
Data analysis
Data processing and data analysis were undertaken using Excel 2016 and SPSS 22.0, respectively. For normally distributed quantitative data (CD4+ and CD8+), we used the Pauta criterion (values beyond ± 3S) to detect outliers. After verification, outliers and anomalies that represent accurate values were still included in the analysis. The continuous variables were statistically described by means and standard deviation, median and 95% confidence intervals (CIs), and were compared between groups using a t-test or Wilcoxon rank-sum test. Categorical variables were described by constituent ratios, and intergroup comparisons were conducted using the chi-square test or Fisher exact probability method. To explore the covariates associated with PSM, a logistic regression model was applied. The PSM was carried out using the R plug-in in SPSS. In the GEE model, an autocorrelation matrix of an autoregressive AR (1) process was selected in the study for dealing with data at different time points. Statistically significant was determined by P < 0.05.
In the logistic regression model, the outcome variable was the CD4+ cell count change. Using the relative magnitude of the difference between the CD4+ cell count at the last follow-up visit and baseline for each patient, and the mean of these differences for all patients, the change in CD4+ cell count has been expressed. In cases where the difference was greater than the mean of the differences (dominant population), 1 was used, while in cases where the difference was less than the mean of the differences (inferior population), 0 was used. Let a binary dependent variable be set as Y, then:
As possible covariates of PSM, the following variables may affect the immune function of HIV/ AIDS patients: gender, age, marital status, possible route of infection, baseline CD4+ count, baseline CD8+ count, baseline CD4+ / CD8+ T-cell ratio, and baseline level of Cr, BUN, ALT and AST. Upon considering the 11 independent variables, a logistic regression model [22], was used to calculate the probability of obtaining positive results.
A propensity score (PS) is defined as the conditional probability of assigning a research object to the treatment group when multiple covariates are present [23]. The baseline characteristic variables selected in this study were used as the matching factors, and the 1:1 matching was performed based on the principles of nearest neighbour matching and calliper matching (calliper value: 0.03) [24]. In addition, the value of CD4+≤ 200 was of important clinical significance and was considered to be in the AIDS stage. Hence, the PSM was conducted within two subgroups respectively, baseline CD4+ count > 200 and baseline CD4+ count ≤ 200.
Results
Sample characteristics
The identification and selection process of the study sample is shown in Fig 1. A final sample of 629 patients, 455 in the integrated group and 174 in the HAART group, was obtained. Baseline characteristics of the samples by groups, without PSM or with PSM, are shown in Table 1. The results of single-factor and multifactor logistic regression analysis were represented in S1 and S2 Tables, Taking into account the results of the logistic regression analyses and important clinical indicators, gender, age, baseline CD4+ count and CD4+/ CD8+ were incorporated as covariates for the PSM. PSM enabled matching of 166 pairs across the full dataset were matched successfully, along with 98 pairs in the subgroup with a baseline CD4+ count > 200, and 55 pairs in the subgroup with a baseline CD4+ count≤ 200.
CD4+ cell count
The results of CD4+ cell count over time in patients before PSM are presented in S3 Table. After PSM, Table 2 illustrates the changes in CD4+ among AIDS patients based on their treatment every three months during the follow-up period. In the full dataset, a significant difference was observed between the groups at 9, 15, 18, 30 and 33 months of follow-up and the CD4+ levels were all higher in the WM group than that was in the integrated group. According to the statistical significance in the baseline CD4+ > 200 subgroup, detected from 3 to 33 months of follow-up, the CD4+ levels in the HAART group were all higher than that in the integrated group. In the subgroup with baseline CD4+ ≤ 200, CD4+ was significantly different at the 3rd, 9th and 15th-month follow-up, with higher levels in the integrated group.
Fig 2 depict changes in CD4+ longitudinal data after PSM. Within the baseline CD4+ > 200 subgroup, the HAART group maintained higher CD4+ mean levels than the integrated group during the 36-month follow-up period. During the first 6 months of follow-up of the full dataset, CD4+ mean levels were higher in the integrated group than in the HAART group and after 6 months, they remained higher in the HAART group. As for the baseline CD4+ ≤ 200 subgroup, higher CD4+ mean levels were observed in the integrated group than in the HAART group during the 36-month follow-up period, except for the 18th and 30th~36th months. All six groups experienced rapid CD4+ increases during the first three months. After three months, the integrated group displayed an overall decrease, whereas the HAART group showed an accumulated increase.
CD4+/CD8+ cell ratio
The CD4+/CD8+ ratios of patients before and after PSM are respectively presented in S4 Table and Table 3. Statistical analysis of the full dataset revealed that the CD4+/CD8+ ratio between the two groups was statistically significant at 15, 18 and 21 months, and the HAART group had a higher ratio than the integrated group. In terms of the CD4+ > 200 subgroup, the HAART group also demonstrated statistically significant higher CD4+/CD8+ ratios than the integrated group among the 12 visits of follow-up, with an exception of the 33rd month. As for the CD4+ ≤ 200 subgroup, statistically significant higher CD4+/CD8+ ratios were witnessed at the 9th and 15th-month follow-up in the integrated group.
The changes of CD4+/CD8+ longitudinal data after PSM were visually shown in Fig 3. For the full dataset, a greater CD4+/CD8+ mean ratio was observed in the HAART group than in the integrated group during the interval of 6 to 33 months. It was also noticed that, in the CD4+ > 200 subgroup, the ratio maintained higher levels in the HAART group throughout the study follow-up, whereas the integrated group remained at higher levels within the CD4+ ≤ 200 subgroup over the entire 36 months. All six groups experienced CD4+/CD8+ ratio increases during the first 6 months.
Analysis results based on generalized estimating equation model
The GEE model was used to analyse the data after PSM, shown in Table 4, with the integrated group serving as the control group. On the basis of the full dataset, CD4+ (OR 1.119, 95%CI [1.018, 1.230]) and CD4+/CD8+ (OR 1.168, 95%CI [1.045, 1.305]) in the HAART group during the follow-up period were significantly higher than those in the integrated group. It was observed that in the subgroup with baseline CD4+ > 200, HAART group had significantly higher values of CD4+ count (OR 1.326, 95%CI [1.214, 1.449]) and CD4+/CD8+ ratio (OR 1.429, 95%CI [1.278, 1.598]) than in the integrated group, while the baseline CD4+ ≤ 200 subgroup showed the opposite, the integrated group revealing higher CD4+ (OR 0.825, 95%CI [0.694, 0.980]) and CD4+/CD8+ (OR 0.826, 95%CI [0.684, 0.997]). As far as safety outcomes are concerned, there were no statistically significant differences between the integrated and the HAART groups in the three datasets in terms of BUN, ALT and AST. Cr level was found significantly higher in the HAART group in all three datasets.
Discussion
We analyzed the real-world longitudinal data to explore the effectiveness and safety of CHM for patients with HIV/AIDS. The patients were followed up for 36 months. A 1:1 PSM was performed to balance the baseline and eliminate confounding factors. The longitudinal data were analyzed using a GEE model. A final sample of 629 patients, 455 and 174 in the integrated group and HAART group respectively, were obtained from the full dataset. Following PSM, 166 pairs from the full dataset were matched successfully, with 98 pairs in the baseline CD4+ > 200 subgroup, and 55 pairs in the baseline CD4+ ≤ 200 subgroup. In the full dataset, HAART group achieved higher CD4+ count (OR 1.119, 95%CI [1.018, 1.230]) and CD4+/CD8+ ratio (OR 1.168, 95%CI [1.045, 1.305]) than the integrated group. Higher level of CD4+ count (OR 1.326, 95%CI [1.214, 1.449]) and CD4+/CD8+ ratio (OR 1.429, 95%CI [1.278, 1.598]) were also observed in the CD4+ > 200 subgroup. For the CD4+ ≤ 200 subgroup, the CD4+ (OR 0.825, 95%CI [0.694, 0.980]) and CD4+/CD8+ (OR 0.826, 95%CI [0.684, 0.997] of the integrated group were higher than those of the HAART group. The safety outcomes showed that there were no significant differences in BUN, ALT and AST levels between the groups but Cr showed significantly higher levels in HAART groups of all three datasets.
In spite of the high external validity of real-world data, there are a number of confounding factors that can make causal inferences less accurate. To improve the internal validity of inferences, matching methods are often applied to real-world data. To improve the accuracy of the results of this study, we use PSM to determine the baseline and to balance both the internal and external validity of real-world data. Meanwhile, using CD4+ as the outcome variable, we converted longitudinal continuous variables into dichotomous variables, dividing them into dominant and inferior populations. As covariates for PSM, gender, age, baseline CD4+ and CD4+/ CD8+ were assessed based on the results of the logistic regression analyses. For a more in-depth analysis, we also divided the CD4+ baseline into two subgroups, baseline CD4+ > 200 and baseline CD4+ ≤ 200 subgroups. In the GEE model, repeated measures are fully taken into account. The GEE is also capable of analyzing a variety of types of outcome variables, as well as processing data with missing data with different observation times and time intervals for observed objects.
The limitation of this study mainly lies in the high rate of lost follow-up of data. Although an appropriate analytical model is adopted for processing, the high rate of lost follow-up will have a certain impact on the accuracy of the results. In addition, PSM can only make equilibrium adjustments for known and measurable covariables, but cannot control the effect of unknown or unmeasured covariables on the outcome effect. Only when all covariables are known and measurable can the unbiased estimation of outcome effects be truly realized.
The results of the subgroup (baseline CD4+ > 200) showed that the HAART group was superior to the integrated group to improve patients’ immune function, whereas the full dataset and the baseline CD4+ ≤ 200 subgroup revealed that the integrated group was more beneficial. The result is consistent with a previous cohort study we conducted without PSM, showing that the CD4+ counts of the integrated group remained significantly lower than those of the HAART group in the first 3 years [25]. However, the results of the two subgroups were in opposite directions, showing qualitative differences, which is relatively rare. Chance alone is unlikely to account for significant subgroup effects, which may not be genuine [26]. Data from a single study can only generate hypotheses about differences between subgroups. To enhance credibility, it is essential to replicate such findings through repeated studies.
Conclusion
The results of the study show that after three years of treatment, the difference between CHMs combined with antiviral therapy and the use of antiviral therapy alone in improving the immune function of HIV infection and AIDS patients has no clinical significance. The results of the two subgroups are in opposite directions, and chance does not explain the apparent subgroup effect. A study with a larger sample size and longer follow-up period is warranted in order to increase study credibility.
Supporting information
S1 Table. Single-factor comparison of characteristics between dominant and inferior patients.
https://doi.org/10.1371/journal.pone.0304332.s001
(DOCX)
S2 Table. Results of logistic regression analysis.
https://doi.org/10.1371/journal.pone.0304332.s002
(DOCX)
S3 Table. CD4+ of HIV/ AIDS patients before PSM (full dataset, CD4+ > 200, CD4+≤ 200) grouped by treatment methods.
https://doi.org/10.1371/journal.pone.0304332.s003
(DOCX)
S4 Table. CD4+/CD8+ of patients before PSM during follow-up (full dataset, baseline CD4+ > 200, baseline CD4+≤ 200) grouped by treatment methods.
https://doi.org/10.1371/journal.pone.0304332.s004
(DOCX)
References
- 1.
World Health Organization (WHO). HIV/AIDS. https://www.who.int/health-topics/hiv-aids#tab=tab_1 assessed on 8 March 2022.
- 2.
World Health Organieezation (WHO), 30 November 2021. HIV/AIDS. https://www.who.int/news-room/fact-sheets/detail/hiv-aids assessed on 8 March 2022.
- 3. Acquired Immunodeficiency Syndrome and Hepatitis C Professional Group, Society of Infectious Diseases, Chinese Medical Association (CMA) & Chinese Center for Disease Control and Prevention. Chinese Guidelines for Diagnosis and Treatment of Human Immunodeficiency Virus Infection/ Acquired Immunodeficiency Syndrome (2021 edition). Chinese Journal of Infectious Diseases. 2021 Dec: 715–735.
- 4. Nolan D, Mallal S. Complications associated with NRTI therapy: update on clinical features and possible pathogenic mechanisms. Antivir Ther. 2004 Dec;9(6):849–63. pmid:15651744
- 5. Rihs TA, Begley K, Smith DE, Sarangapany J, Callaghan A, Kelly M, et al. Efavirenz and chronic neuropsychiatric symptoms: a cross-sectional case control study. HIV Med. 2006 Nov;7(8):544–8. pmid:17105514
- 6. Mollan KR, Smurzynski M, Eron JJ, Daar ES, Campbell TB, Sax PE, et al. Association between efavirenz as initial therapy for HIV-1 infection and increased risk for suicidal ideation or attempted or completed suicide: an analysis of trial data. Ann Intern Med. 2014;161(1):1–10. Erratum in: Ann Intern Med. 2014;161(4):308. pmid:24979445
- 7. World Health Organization (WHO), 24 November 2021. HIV agents resistance report 2021. https://www.who.int/news-room/fact-sheets/detail/hiv-aids assessed on 8 March 2022.
- 8. Pau AK, George JM. Antiretroviral therapy: current agentss. Infect Dis Clin North Am. 2014 Sep;28(3):371–402. pmid:25151562
- 9. Micollier E, Black M. Management of the AIDS epidemic and local/global use of Chinese medicine. Chin Perspect. 2009;1(77):67–78.
- 10. Qian Z, Zhang Y, Xie X, Wang J. Efficacy and safety of traditional Chinese herbal medicine combined with HAART in the treatment of HIV/AIDS: A protocol for systematic review and meta-analysis. Medicine (Baltimore). 2021;100(52):e28287. pmid:34967361
- 11. Wen L, Liu YF, Jiang C, Zeng SQ, Su Y, Wu WJ, et al. Comparative Proteomic Profiling and Biomarker Identification of Traditional Chinese Medicine-Based HIV/AIDS Syndromes. Sci Rep. 2018 Mar 8;8(1):4187. pmid:29520099; PMCID: PMC5843661.
- 12. Wu X. Clinical Observation on the Effect of Regulating Intestinal Flora with Traditional Chinese Medicine on Improving the Immune Reconstitution Insufficiency of AIDS. China Academy of Chinese Medical Sciences. 2021.
- 13. Tan Y. Retrospective Study of Traditional Chinese Medicine in Intervening AIDS Immune Reconstruction. Hunan University of Chinese Medicine. 2020.
- 14. Tu MX. Retrospective Study on the Adverse Effects of Traditional Chinese Medicine in Intervening AIDS with Highly Active Antiretroviral Therapy (HARRT). Hunan University of Chinese Medicine. 2020.
- 15. Camm AJ, Fox KAA. Strengths and weaknesses of ’real-world’ studies involving non-vitamin K antagonist oral anticoagulants. Open Heart. 2018;5(1):e000788. Published 2018 Apr 21. pmid:29713485
- 16. Santinelli L, Ceccarelli G, Borrazzo C, Celani L, Pavone P, Innocenti GP, et al. Real word outcomes associated with use of raltegravir in older people living with HIV: results from the 60 months follow-up of the RAL-age cohort. Expert Rev Anti Infect Ther. 2020;18(5):485–492. pmid:32096433
- 17. Okoli C, Schwenk A, Radford M, Myland M, Taylor S, Barnes J, et al. Using Climate-HIV to describe real-world clinical outcomes for people living with HIV taking dolutegravir-based regimens. Int J STD AIDS. 2021;32(12):1165–1173. pmid:34156330
- 18. Jin Y, Liu Z, Chen X, Wang X, Wang D, Jiang Z, et al. Survival of people living with HIV after treatment with traditional Chinese medicine in Henan province of China: a retrospective cohort study. J Tradit Chin Med. 2014;34(4):430–6. pmid:25185360
- 19. Wang J, Liang B, Zhang X, Xu L, Deng X, Li X, et al. An 84-month observational study of the changes in CD4 T-lymphocyte cell count of 110 HIV/AIDS patients treated with traditional Chinese medicine. Front Med. 2014;8(3):362–7. pmid:25190350
- 20. Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res. 2011;46(3):399–424. pmid:21818162
- 21. Generalized Estimating Equations. (n.d.) Retrieved March 8, 2022, from https://support.sas.com/rnd/app/stat/topics/gee/gee.pdf
- 22. Li K, He J. Medical Statistics (Sixth edition). Beijing: People’s Medical Publishing House. 2013: 125–130.
- 23. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55.
- 24. Huang LH, Chen F. The Propensity Score Method and its Application. Chinese Journal of Preventive Medicine. 2019;53(7):752–755. pmid:31288349
- 25. Sun J, Jiang F, Wen B, Liu ZW, Han M, Robinson N, et al. Chinese herbal medicine for patients living with HIV in Guangxi province, China: an analysis of two registries. Sci Rep. 2019 Nov 25;9(1):17444. pmid:31767895
- 26. Buyse M E. Analysis of clinical trial outcomes: some comments on subgroup analyses. Control Clin Trials. 1989, 10(4)(suppl):187–194. pmid:2605967