Chinese Herbal Medicine Usage Reduces Overall Mortality in HIV-Infected Patients With Osteoporosis or Fractures

The survival of patients with HIV has greatly improved, due to Anti-Retroviral Therapy (ART). However, long-term HIV survivors often develop serious bone abnormalities, possibly due to the interplay of osteoblasts, osteoclasts, HIV ad ART. We evaluated in a nation-wide study in Taiwan the effect of Chinese herbal medicine (CHM) on overall mortality in HIV patients with osteoporosis or fractures. Enrollment period was between 1998 and 2011. Patients with osteoporosis or fractures before the HIV infection, and those with less than 14 days CHM use, were excluded. This left 498 patients, 160 CHM users, 338 without CHM. Univariate Kaplan-Meier and multivariate Cox regression analysis were used to compare the overall mortality in these 2 groups. Due to the nature of Chinese medicine, CHMs inevitably varied. We therefore also used rule mining and network analysis to determine which major CHM clusters were prescribed to the patients. CHM users had a much Lower mortality (hazard ratio (HR) = 0.43, 95% confidence interval (CI): 0.24–0.77, p < 0.005) and higher survival (p = 0.004, log-rank test). Although the CHMs greatly varied, network analysis identified one main cluster of strongly related CHM combinations (Chuan-Xiong-Cha-Tiao-San (CXCTS), Gan-Cao (GC; Glycyrrhiza uralensis Fisch.), Liu-He-Tang (LHT), Huang-Qin-Tang (HQT), Jia-Wei-Ping-Wei-San (JWPWS), and Dang-Gui-Long-Hui-Wan (DGLHuiW)). CHM as an additional treatment strongly improves overall survival in HIV-infected patients with osteoporosis and fractures.

Bone related abnormalities including low bone density, osteomalacia, osteonecrosis, osteopenia, osteoporosis, and fracture (Hoy and Young, 2016;Ahmad et al., 2017). Osteoporosis is a multifactorial systemic skeletal disease with low bone density, degeneration of bone architecture, bone fragility, and consequent increased risk of fracture (Ji and Yu, 2015;Sozen et al., 2017). A number of studies report that lower bone density was observed in HIV-infected patients when compared with non-infected individuals (Bruera et al., 2003;Amiel et al., 2004;Arnsten et al., 2007). The pathological mechanism between HIV and/or ART and bone related abnormalities remain to be elucidated, but are probably due to HIV and ART affecting the interactions between osteoclasts and osteoblasts. Furthermore, the loss of bone mineral density is frequently observed in HIV-infected patients with ART (Duvivier et al., 2009;van Vonderen et al., 2009;Grant et al., 2016;Hoy et al., 2017;Chisati et al., 2020b). HIV-infected patients placed on protease inhibitor (PI) regimens demonstrate bone loss in the spine, while the nucleoside/nucleotide reverse transcriptase inhibitor (NRTI) regimen is associated with bone loss at the hip (Hoy et al., 2017).
Chinese herbal medicines (CHMs) are often used to treat bone related diseases, as they show anti-inflammatory, antiosteopenia, anti-osteoporotic, and promote fracture healing activities (Chow et al., 1982;Chen et al., 2005;Li et al., 2011;Ma et al., 2011;Xiang et al., 2011;Shih et al., 2012;Wong et al., 2013;He and Shen, 2014;Mukwaya et al., 2014;Lin et al., 2015;Zhang et al., 2016;Hsiao et al., 2017;Wang et al., 2018b;Xi et al., 2018;Cheng et al., 2019a;Cheng et al., 2019b). However, none of these studies have been carried out in prospective randomized clinical trials in humans. These results encourage to analyze if CHM as additional therapy to improve osteoporosis and fractures management and survival among HIV-infected patients. We therefore analyzed in a population-based nationwide database from Taiwan, what the effect was of CHM treatment-or-not on the overall mortality in HIV-infected patients with osteoporosis or fractures.
The flowchart for the selection for HIV-infected patients with osteoporosis or fractures is shown in Figures 1,2. Patients were initially diagnosed with HIV infection followed by osteoporosis or fractures ( Figure 2). Between January 01, 1998 and December 31, 2011, 3450 HIV-infected patients with osteoporosis or fractures were identified ( Figure 1). After exclusion, there were 498 patients with osteoporosis or fractures, including 160 CHM and 338 non-CHM users ( Figure 1). The CHM users were defined as the patients who received CHMs for at least 14 days in the 12 months after osteoporosis or fractures (Figures 1,2).
Patients were classified as CHM users when they received more than 14 CHM cumulative prescription days among the first year after osteoporosis or fractures (n 160, Figures 1,2). The index date started after the 14 cumulative CHM prescription days were accomplished ( Figure 2). The CHM users received CHM therapies during the study period (Supplementary Table S2). On the other hand, the controls were classified as non-CHM users when they did not receive any CHMs for the study period (n 338). To reduce potential confounding factors, these two groups were matched for age, gender, and index duration using the propensity score matching method (1:1 ratio) ( Table 1). This resulted in 149 CHM and 149 non-CHM users (Figures 1,2; Table 1).
Among these patients, the characteristics included age, gender, index duration (from the HIV infection diagnosed date to the diagnosed date of osteoporosis or fractures), Charlson comorbidity index (CCI), and comorbidities ( Table 1). In this study, comorbidities were defined before HIV infection ( Table 1). The ART usage was defined before the diagnosed date of osteoporosis or fractures ( Table 1). The study was approved by the Institutional Review Board of the China Medical University Hospital (The ethics approval number: CMUH107-REC3-074(CR1)).
Taiwan. CHMs include single herbs and herbal formulae. A single herb is made from the flower, root, stem, or leaf of a given plant. It is also made from an organ of an animal, insect, or mineral source. The herbal formulae are mixtures of a minimum of two single herbs. The CHM composition, frequency, and usage patterns are shown in Supplementary Table S1. CHMs are produced by pharmaceutical Good Manufacturing Practice companies with in Taiwan.
Association rule mining was performed, as previously described, using SAS software (version 9.4; SAS Institute, Cary, NC, United States). This association rule mining has been applied to discover studies in the relationships of these CHM prescriptions (Chen et al., 2014;Cheng et al., 2019b;Tsai et al., 2019). Chinese herbal medicine (CHM) product X (CHM_X) and CHM product Y (CHM_Y) were shown as the "items," respectively. The CHM prescriptions were used as the "transactions," with co-occurrences of CHM_X and CHM_Y ( Table 3). This expression shows the relationship between the occurrences of CHM_X and CHM_Y. The strength of the association using this technique was expressed as support, confidence, and lift. Support is a measure of whether an association between CHM_X and CHM_Y happened by chance. The support (X) (%) value is the calculated joint probability of having both of CHM_X and CHM_Y, which is (the frequency of CHM_X and CHM_Y/total number of prescription) × 100%. Confidence is an indicator of how often  CHM_Y appeared in transactions that contained CHM_X. The confidence value (CHM_X → CHM_Y; %) is the calculated conditional probability of having a prescription of CHM_Y among those who already have the prescription of CHM_X, which is given as (frequency of CHM_X and CHM_Y/ frequency of CHM_X) x 100%. Lift is the ratio of observed support to expected support when X and Y are independent. The lift value is the confidence (CHM_X → CHM_Y) (%)/P (Y) (%) or confidence (CHM_Y → CHM_X) (%)/P (X) (%). A lift value greater than 1 indicates that the occurrences between the two CHM products are dependent and suggests a strong cooccurrence relationship between CHM_X and CHM_Y. Network analysis was performed as previously described (Cheng et al., 2019a;Cheng et al., 2019b) (Figure 3). The single herb is expressed as a green circle, and the herbal formula is shown as a red circle. The prescription frequency of the single herb or herbal formula is shown (Supplementary Table   S2) and is denoted as the circle size. The support value (%) (between CHM_X and CHM_Y) is shown in Table 3 and is expressed as the line size. The lift value is also shown in Table 3 and is represented as the line color. The connection strength between the paired CHM products is shown as the line size and line color. All data were employed using Cytoscape software (https://cytoscape.org/, version 3.7.0).

Statistical Analysis
Age was expressed as continuous data (years, mean ± SD) and categorical data (numbers (percentages)) ( Table 1). Index duration was expressed as continuous data (from the diagnosed date of HIV infection to the diagnosed date of osteoporosis or fractures) (day, Mean ± SD) ( Table 1). Gender, antiretroviral therapies (ART) usage, Charlson comorbidity index (CCI) and comorbid conditions were expressed as categorical data (numbers (percentages))  Frontiers in Pharmacology | www.frontiersin.org April 2021 | Volume 12 | Article 593434 ( Table 1). The un-paired Student t-test was applied in continuous data (Supplementary Table S2). The Chi-squared test was used in categorical data. Univariate (crude) and multivariate (adjusted) Cox proportional hazard models were employed to evaluate the risk of overall mortality ( Table 2). Multivariate-adjustments include age, gender, CHM use, ART use, and CCI ( Table 2). For survival analysis, Kaplan-Meier method and the log-rank test were performed (Figure 4; Supplementary Table S4, S5). All data and statistical analyses were employed using SAS software (version 9.4; SAS Institute, Cary, NC, United States).

Basic Characteristics
The 160 CHM users received CHM therapies during the study period (Supplementary Table S3). The other 338 patients did not use any CHM at all. The demographic characteristics of total subjects are shown in Table 1. When compared with the 338 non-CHM users, the 160 CHM users were slightly older, more often females, had a longer index duration between HIV diagnosis date and the osteoporosis or fractures date, and had more often comorbidities (p < 0.05). To prevent the effects of these confounding factors, propensity score matching (1:1 ratio) was applied to match the two groups for age, gender, and index duration. After matching, each group had 149 HIV-infected patients with osteoporosis or fractures (Figures 1,2; Table 1).

Risk of Overall Mortality
The risk of overall mortality in patients with osteoporosis or fractures was evaluated by Cox proportional hazard models ( . Female patients had a lower risk of overall mortality than male patients (HR: 0.51, 95% CI: 0.26-0.99, p 0.0465). The CHM users had a lower risk of overall mortality than non-CHM users (HR: 0.39, 95% CI: 0.21-0.70, p 0.0018). Patients with Charlson comorbidity index (CCI) ≥ 2 showed a higher risk of overall mortality than those who did not have any comorbidities (HR: 2.95, 95% CI: 1.26-6.88, p 0.0125).
The multivariate Cox proportional hazard model showed that patients had a higher risk of overall mortality per year increase in age after adjusting for gender, CHM use, and Charlson comorbidity index ( Table 2; adjusted hazard ratio (aHR): 1.02, 95% CI: 1.00-1.04, p 0.0173). The CHM users had a lower risk of overall mortality than non-CHM users after adjusting for age, gender, and Charlson comorbidity index (aHR: 0.43, 95% CI: 0.24-0.77, p 0.0047). Patients with Charlson comorbidity index (CCI) ≥ 2 showed a higher risk of overall mortality than those who did not have any comorbidities after adjusting for age,  Kaplan-Meier survival plots exhibited that there was a difference in the cumulative incidences of overall survival between the CHM and non-users (Figure 4; p 0.0036, log-rank test). The cumulative incidence of overall survival was significantly higher in CHM users.

CHM Prescription Pattern and Network Analysis
The commonly prescribed CHM products and compositions are listed for the HIV-infected patients with osteoporosis or fractures in Supplementary Table S1. LC-MS/MS analysis of active component standards and these 6 herbal extracts are also shown in Supplementary Figures S1-S6. According to the frequency of prescriptions (Supplementary Table S1), Chuan-Xiong-Cha-Tiao-San (CXCTS) was the most commonly herbal formula. The second and third formulas were Liu-He-Tang (LHT) and Jia-Wei-Ping-Wei-San (JWPWS), respectively. Gan-Cao (GC; Glycyrrhiza uralensis Fisch.) was the most commonly single herb.
Network analysis showed the CHM prescription network for patients with osteoporosis or fractures (Figure 3; Supplementary Figure S7). There were 149 patients who used 3,745 prescriptions by traditional Chinese medicine doctors ( Table 3). Network analysis showed one main CHM cluster, including CXCTS, GC, LHT, HQT, JWPWS, and DGLHuiW. Our results show that these 6 CHMs are important for HIV-infected patients with osteoporosis or fractures.

DISCUSSION
Long-term living with HIV and ART use in HIV-infected patients are associated with adverse effects including bone related abnormalities. In this study, we investigated the effect of CHMs on the overall mortality in HIV-infected patients with osteoporosis or fractures in Taiwan. We found that CHM usage reduced the overall mortality for these patients. We also described their CHM prescription network; these included CXCTS, GC, LHT, HQT, JWPWS, and DGLHuiW. CHM treatment exhibited lower risks of overall mortality for HIV-infected patients with osteoporosis or fractures in Taiwan.
Reduced bone mineral density is observed in HIV-infected patients on ART therapy (Duvivier et al., 2009;van Vonderen et al., 2009;Grant et al., 2016;Hoy et al., 2017;Chisati et al., 2020b). Furthermore, Chisati et al., reported that low bone mineral density was also associated with low levels of physical activity among these patients (Chisati et al., 2020b). Maximal strength training for physical activity improves bone mineral density for people living with HIV and receiving ART (Chisati et al., 2020a). In this study, we observed that among HIV-infected   Table S1). Support (X) (%) Frequency of prescriptions of X and Y products/total prescriptions × 100%. Confidence (X → Y) (%) Frequency of prescriptions of X and Y products/Frequency of prescriptions of X product × 100%. P (Y) (%) Frequency of prescriptions of Y product/total prescriptions × 100%.
In this study, the limitations were the lacks of laboratory tests, education, occupation, and lifestyle in the database. However, we found that CHM may reduce risk of overall mortality in patients with osteoporosis or fractures, and may be useful for future investigations in randomized controlled trials (RCT) and functional studies in bone protection. Large-scale RCTs for these CHMs in HIV-infected patients should be performed to determine their relative effectiveness and safety, and to evaluate their potential interactions during regular treatments in these patients.
HIV-infected patients with osteoporosis or fractures who used CHMs as adjunctive therapy had a better survival rate. Based on association rules mining and network analysis, CXCTS, GC, LHT, HQT, JWPWS, and DGLHuiW are potential CHMs for these patients. Further investigations may be undertaken to validate the safety and efficacy of CHMs among these patients. An investigation into the mechanism of actions of the potential compounds of CHMs are required.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by The study was approved by the Institutional Review Board of the China Medical University Hospital (The ethics approval number: CMUH107-REC3-074(CR1)). Written Frontiers in Pharmacology | www.frontiersin.org April 2021 | Volume 12 | Article 593434 8 informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
Y-JL, M-WH, and J-PL wrote the manuscript and interpreted the data. C-JC, J-SC, M-LC, C-FC, T-ML, Y-CW, T-HL, C-CL, S-MH, Y-NL, and C-HC collected, assembled, and analyzed the data. F-JT and Y-JL provided study materials. W-ML and Y-JL designed, conceived the study, and amended the manuscript.

ACKNOWLEDGMENTS
We are grateful to the Health Data Science Center at the China Medical University Hospital for providing administrative, technical, and funding support. The funding entities for this study had no roles in the study design, data collection, data analysis, interpretation, or authorship of this manuscript.