1 INTRODUCTION

Once a patient has had chickenpox or received the varicella vaccine, the varicella-zoster virus (VZV) can become latent in the sensory ganglia. A decline in cell-mediated immunity can cause the virus to reactivate and spread along the sensory nerve to the skin with the development of herpes zoster.^{1, 2} In about 10%–50% of patients with zoster, pain may persist for 90 days or more due to postherpetic neuralgia (PHN).¹ Although attenuated varicella and zoster vaccines have been available since 1995, the incidence of varicella and zoster has increased worldwide.² The global numbers of cases with varicella and zoster in 1990, 2000, 2010, and 2019 were 71.25, 73.79, 79.52, and 83.96 million, respectively.³

Older adults (with a gradual decline in immunity due to aging) and patients with cancer or autoimmune diseases on immunosuppressive drugs are high-risk groups for herpes zoster.^{2, 4} It is not uncommon that persons with type 2 diabetes (T2D) show compromised immune function due to chronic hyperglycemia. Studies have reported that persons with T2D are more likely to develop zoster and PHN.⁴ Although medications for treating zoster are available, many patients with PHN show low treatment satisfaction.⁵ PHN is a complicated neuropathic pain that can be severe and long-lasting and significantly impact the social, physical, and psychological aspects of quality of life.⁵

Metformin is a synthetic derivative of guanidine, isolated from the extract of Galega officinalis, with a prominent antidiabetic effect noted since the 1920s.⁶ The UK Prospective Diabetes Study (1998) demonstrated that metformin could significantly reduce cardiovascular events in patients with T2D.⁷ Metformin has been used worldwide since 1998, and it is by far the first-line antidiabetic drug for the management of T2D.⁶ Metformin was also used for treating malaria and influenza in the 1940s.⁶ Reports show beneficial effects of treatment for tuberculosis and sepsis.⁸ Studies show that metformin reduces the severity and mortality of COVID-19.⁹ Preclinical studies have shown that metformin can decrease the secretion of proinflammatory cytokines and improve T-cell immunity through the activation of AMP-activated protein kinase (AMPK).^8-10 Therefore, we hypothesized that metformin could diminish the risk of herpes zoster and PHN. We, herein, designed this study to compare the incidence of zoster and PHN between metformin users and nonusers in patients with T2D.

2 MATERIALS AND METHODS

2.2 Study design

We recognized newly diagnosed T2D patients between January 1, 2000, and December 31, 2017, and followed them till December 31, 2018. Diagnosis of T2D was based on ICD codes (ICD-9 code: 250, except 250.1x; ICD-10: E11. Additional File 1: Table S1) for at least 3 outpatient visits within 1 year or one hospitalization record. The algorithm of using ICD codes to define T2D was validated by a previous Taiwan study with an accuracy of 74.6%.¹² Patients were excluded (Figure 1) under the following conditions: (1) age below 20 or above 80 years; (2) missing sex or age data; (3) diagnosis of type 1 diabetes, hepatic failure, or under dialysis; (4) diagnosis of herpes zoster or PHN before the index date. (5) diagnosis of T2D before January 1, 2000 (to exclude prevalent T2D cases).

3 RESULTS

3.2 Main outcomes

In the matched cohorts, 1864 (3.93%) metformin users and 2153 (4.54%) nonusers developed herpes zoster during the follow-up period (incidence rate: 8.1 vs. 11.1 per 1000 person-years). In the multivariable model, metformin users showed a significantly lower risk of herpes zoster (aHR = 0.70, 95% CI = 0.66–0.75). Patients aged 41–60 years and 61–80 years with COPD, RA, SLE, cancer, depression, on insulin, and immunosuppressants showed a significantly higher risk of herpes zoster, but those with psychosis had a significantly lower risk of zoster (Table 2). Because sulfonylureas use was not matched well between the metformin users and nonusers (SMD = 0.514), we performed a subgroup analysis (Table S2) to observe if the effects of metformin use would be influenced by sulfonylureas use. The subgroup analysis showed that metformin use was associated with a significantly lower risk of herpes zoster in both patients with (aHR = 0.58, 95% CI = 0.51–0.66) and without sulfonylureas use (aHR = 0.75, 95% CI = 0.69–0.81). The time-dependent analysis, which adjusted impacts of the time-varying exposure of metformin, and other medications on the metformin-associated risk of herpes zoster, showed that metformin users had a significantly lower risk of herpes zoster (aHR = 0.57, 95% CI = 0.52–0.62) compared with nonusers (Table S3).

Table 2. Incidence rate and hazard ratio of herpes zoster in patients with type 2 diabetes

Variables	Herpes zoster
	n	PY	IR	cHR	(95% CI)	p value	aHRa	(95% CI)	p value
Metformin nonusers	2153	194 643	11.1	1.00	(reference)	–	1.00	(reference)	–
Metformin users	1864	230 082	8.1	0.73	(0.69, 0.78)	<0.001	0.70	(0.66, 0.75)	<0.001
Sex
female	2249	222 873	10.1	1.00	(reference)	–	1.00	(reference)	–
male	1768	201 852	8.8	0.87	(0.82, 0.92)	<0.001	0.94	(0.88, 1.01)	0.080
Age
20–40	125	38 108	3.3	1.00	(reference)	–	1.00	(reference)	–
41–60	1489	188 163	7.9	2.41	(2.01, 2.89)	<0.001	2.29	(1.90, 2.76)	<0.001
61–80	2403	198 454	12.1	3.69	(3.08, 4.42)	<0.001	3.36	(2.79, 4.05)	<0.001
Obesity	80	11 655	6.9	0.72	(0.58, 0.90)	0.004	0.89	(0.71, 1.11)	0.295
Smoking	67	7455	9.0	0.95	(0.75, 1.21)	0.674	1.07	(0.84, 1.36)	0.596
Comorbidities
Alcohol disorders	120	15 736	7.6	0.80	(0.67, 0.96)	0.016	0.97	(0.80, 1.18)	0.756
COPD	1513	129 510	11.7	1.38	(1.29, 1.47)	<0.001	1.18	(1.11, 1.27)	<0.001
Rheumatoid arthritis	35	1753	20.0	2.12	(1.52, 2.96)	<0.001	1.58	(1.11, 2.25)	0.011
SLE	9	438	20.5	2.18	(1.14, 4.19)	0.019	2.07	(1.07, 4.01)	0.031
Liver cirrhosis	134	13 352	10.0	1.06	(0.90, 1.26)	0.477	1.07	(0.89, 1.29)	0.446
Cancer	611	44 440	13.7	1.53	(1.41, 1.67)	<0.001	1.43	(1.30, 1.57)	<0.001
Psychosis	332	39 175	8.5	0.89	(0.79, 0.99)	0.034	0.80	(0.71, 0.90)	<0.001
Depression	1685	156 515	10.8	1.24	(1.16, 1.32)	<0.001	1.13	(1.05, 1.21)	<0.001
Dementia	14	1326	10.6	1.11	(0.66, 1.88)	0.687	1.16	(0.68, 1.99)	0.590
DCSI
0	1211	146 725	8.3	1.00	(reference)	–	1.00	(reference)	–
1	795	84 953	9.4	1.13	(1.04, 1.24)	0.006	0.96	(0.88, 1.06)	0.451
≥2	2011	193 047	10.4	1.26	(1.18, 1.36)	<0.001	0.96	(0.87, 1.06)	0.388
Medications
Sulfonylureas	1282	139 640	9.2	0.96	(0.90, 1.02)	0.200	1.04	(0.97, 1.12)	0.240
OAD number
0–1	3897	413 635	9.4	1.00	(reference)	–	1.00	(reference)	–
2–3	116	10 926	10.6	1.13	(0.94, 1.35)	0.208	1.07	(0.89, 1.29)	0.474
>3	4	164	24.4	2.59	(0.97, 6.91)	0.057	2.07	(0.78, 5.54)	0.146
Insulin	1812	167 861	10.8	1.26	(1.18, 1.34)	<0.001	1.15	(1.07, 1.23)	<0.001
Corticosteroids	43	4201	10.2	1.08	(0.80, 1.46)	0.604	1.05	(0.78, 1.42)	0.732
Immunosuppressants	48	2199	21.8	2.32	(1.75, 3.09)	<0.001	2.13	(1.57, 2.89)	<0.001

• Abbreviations: aHR, adjusted hazard ratio; cHR, crude hazard ratio; CI, confidence interval; COPD, chronic obstructive pulmonary disease; DCSI, Diabetes Complications Severity Index; IR, incidence rate, per 1,000 person-years; OAD, oral antidiabetic drug; PY, person-years; SLE, systemic lupus erythematosus.
• ^a Multivariable analysis, including sex, age, comorbidities, CCI, DCSI, corticosteroids, immunosuppressants, statins, aspirin, insulin, item and number of oral antidiabetic drugs, duration of diabetes, and HbA1c check-up per year.

Among those who had herpes zoster infection, 80 metformin users and 257 nonusers got PHN during the follow-up period (incidence rate: 8.9 vs. 42.53 per 1000 person-years). The multivariable analysis showed that metformin users had a significantly lower risk of PHN (aHR = 0.51, 95% CI = 0.39–0.68) in T2D patients with herpes zoster infection. Patients with SLE, psychosis, depression, and receiving insulin showed a significantly higher risk of PHN (Table 3).

Table 3. Incidence rate and hazard ratio of postherpetic neuralgia in patients with type 2 diabetes who had herpes zoster infection

Variables	Postherpetic neuralgia
	n	PY	IR	cHR	(95% CI)	p value	aHRa	(95% CI)	p value
Metformin nonusers	257	6043	42.53	1	(reference)	–	1	(reference)	–
Metformin users	80	8989	8.90	0.48	(0.36, 0.63)	<0.001	0.51	(0.39, 0.68)	<0.001
Sex
female	211	8541	24.70	1.00	(reference)	–	1.00	(reference)	–
male	126	6491	19.41	0.74	(0.56, 0.98)*	0.037	0.89	(0.66, 1.19)	0.423
Age
20–40	5	503	9.94	1.00	(reference)	–	1.00	(reference)	–
41–60	127	5709	22.24	2.24	(0.71, 7.10)	0.170	2.08	(0.65, 6.67)	0.219
61–80	205	8820	23.24	2.39	(0.76, 7.50)	0.136	2.14	(0.67, 6.90)	0.202
Obesity	7	292	23.94	1.17	(0.48, 2.84)	0.731	1.09	(0.44, 2.70)	0.846
Smoking	3	167	17.94	0.35	(0.05, 2.48)	0.292	0.29	(0.04, 2.07)	0.215
Comorbidities
Alcohol disorders	14	268	52.26	1.85	(0.95, 3.61)	0.072	2.11	(0.99, 4.49)	0.053
COPD	145	5328	27.21	1.17	(0.89, 1.54)	0.252	1.01	(0.75, 1.34)	0.973
Rheumatoid arthritis	5	117	42.74	2.29	(0.85, 6.17)	0.100	1.96	(0.63, 6.04)	0.244
SLE	4	22	180.15	8.83	(2.82, 27.61)	<0.001	14.00	(3.70, 53.19)	<0.001
Liver cirrhosis	10	396	25.22	0.91	(0.41, 2.06)	0.830	0.63	(0.26, 1.54)	0.314
Cancer	52	2227	23.35	1.12	(0.78, 1.61)	0.546	1.40	(0.94, 2.06)	0.095
Psychosis	56	829	67.59	2.55	(1.76, 3.70)	<0.001	2.05	(1.36, 3.09)	<0.001
Depression	202	5559	36.34	2.00	(1.53, 2.62)	<0.001	1.50	(1.10, 2.04)	0.010
Dementia	3	29	102.74	1.72	(0.24, 12.29)	0.587	0.85	(0.11, 6.53)	0.877
DCSI
0	83	5032	16.50	1.00	(reference)	–	1.00	(reference)	–
1	63	3072	20.51	1.23	(0.82, 1.86)	0.315	0.99	(0.64, 1.54)	0.979
≥2	191	6929	27.57	1.55	(1.12, 2.15)	0.008	1.37	(0.89, 2.11)	0.149
Medications
Sulfonylureas	106	5173	20.49	0.99	(0.75, 1.32)	0.955	1.39	(1.02, 1.88)	0.037
OAD number
0–1	327	14630.6	22.35	1	(reference)	–	1.00	(reference)	–
2–3	10	389.7	25.661	1.18	(0.56, 2.51)	0.6669	1.39	(0.64, 2.99)	0.406
Insulin	180	6096	29.53	1.37	(1.05, 1.80)	0.020	1.11	(0.83, 1.48)	0.494
Corticosteroids	3	158	18.99	1.34	(0.43, 4.20)	0.612	1.13	(0.35, 3.68)	0.841

• Abbreviations: aHR, adjusted hazard ratio; cHR, crude hazard ratio; CI, confidence interval; COPD, chronic obstructive pulmonary disease; DCSI, Diabetes Complications Severity Index; IR, incidence rate, per 1,000 person-years; OAD, oral antidiabetic drug; PY, person-years; SLE, systemic lupus erythematosus.
• ^a Multivariable analysis, including sex, age, comorbidities, CCI, DCSI, corticosteroids, immunosuppressants, statins, aspirin, insulin, item and number of oral antidiabetic drugs, duration of diabetes, and HbA1c check-up per year.

The Kaplan–Meier analysis showed that the cumulative incidences of herpes zoster and PHN were lower in metformin users than nonusers (Figure 2).

4 DISCUSSION

Although vaccines and medications for herpes zoster are available, zoster and PHN pose significant healthcare challenges worldwide. Notably, this study showed that metformin use was associated with a significantly lower risk of herpes zoster and PHN. Furthermore, a higher cumulative dose of metformin was associated with a further lower risk of zoster and PHN. Metformin may provide a suitable option to reduce the burden of zoster and PHN in patients with T2D.

Chronic hyperglycemia in patients with diabetes can aggravate cellular oxidative stress and compromise innate and adaptive immune functions.¹⁶ Studies have demonstrated that patients with diabetes have about 1.3–1.6 times higher risk of herpes zoster than those without diabetes.^{4, 17} Lai et al.¹⁷ have performed a systemic review and meta-analysis, and showed that the incidence of herpes zoster ranging from 2.11 to 9.38 per 1000 person-years, with the overall incidence was 7.22 per 1000 person-years in patients with diabetes mellitus. The calculated incidence of herpes zoster in our study was 9.45 per 1000 person-years, which was slightly higher than the upper incidence (9.38 per 1000 person-years) of Lai's study. The difference of incidences between these two studies may be due to their dissimilar study population with different ethnicity, age, and immune function. Our study also showed that increasing age, COPD, autoimmune diseases, cancers, depression, and immunosuppressant therapy or insulin use had a higher risk of zoster. This finding is consistent with previous studies.^{1, 4} Older adults may be susceptible to herpes zoster due to a decline in immune function (immunosenescence). The use of chemotherapy in cancer patients or immunosuppressants in patients with autoimmune diseases and depression can compromise patients' cellular immunity and cause zoster.^{4, 18} People who receive insulin therapy may have long diabetes duration and suboptimal glucose control, which can comprise cell-mediated immunity, resulting in zoster.¹⁶ However, this study also showed that patients with psychosis had a lower risk of herpes zoster. The reason for this result is unclear. To the best of our knowledge, our study is the first to show that metformin use may reduce the risk of herpes zoster in patients with T2D, and a higher cumulative dose of metformin use can further lower the risk of zoster than no-use of metformin.

PHN is a complicated neuropathic pain that persists after the recovery of zoster. It can manifest as burning, electric shock-like, shooting, or evoked pain. PHN is difficult to treat, and less than half of patients experience a substantial reduction in suffering.² Some patients have severe neuralgia that lasts for years and dramatically impairs their quality of life.^{2, 5} Currently, medications for pain relief are available, but there is no disease-modifying treatment for PHN.⁵ Few studies have reported risk factors for PHN. Studies have shown that severe and widespread zoster is more likely to produce PHN.¹ Reports suggest that diabetes can also predispose to PHN.⁵ Our study showed that patients with autoimmune diseases, mental illness, and insulin users were more likely to develop PHN, probably because these people were more prone to herpes zoster or severe zoster that may cause PHN.^{4, 5} This study showed that metformin use was associated with a significantly lower risk of PHN, and a higher cumulative dose of metformin was associated with a more reduced risk of PHN.

Studies show that metformin can reduce the risks of pneumonia, tuberculosis, and sepsis; metformin can also decrease the severity and risk of mortality due to COVID-19 infection.^{8, 9} The mechanisms for metformin to decrease the risk of zoster and PHN may be multifactorial. (1) Metformin can reduce reactive oxygen species and C-reactive protein production, suppress nuclear factor-kappa B activation, decrease the secretion of interferon (INF)-α, INF-γ, tumor necrosis factor-α, interleukin (IL)−1β, IL-6, IL-8, and monocyte chemoattractant protein-1; metformin can increase the levels of IL-10 and transforming growth factor-β.^{8-10, 19} Thus, metformin may alleviate systemic inflammation, amend the imbalance of proinflammatory and anti-inflammatory cytokines, and alleviate the reactivation of the VZV. (2) By suppressing signal transducer and activator of transcription-3, metformin can inhibit monocyte-to-macrophage differentiation, improve autophagy, and restrict viral multiplication.^{20, 21} (3) Metformin can activate AMPK and suppress the mammalian target of rapamycin to improve T cell differentiation and cell-mediated immunity.¹⁰ (4) AMPK activation can potentiate the expression of genes with antiviral properties (IFNs, IFN-stimulated genes), inhibit inflammatory mediators (tumor necrosis factor-α and CCL5), and enhance the innate antiviral response in virus-infected cells.²² (5) The activation of AMPK by metformin may inhibit virus-induced glycolysis, attenuate the phosphorylation of acetyl-CoA carboxylase, resulting in the decreased source of energy and building blocks for viral replication.²²

The strengths of this study are as follows: this is a nationwide and population-based study with minimal selection bias. The observational period of this study spans 18 years with the availability of comprehensive information on demographics, comorbidities, and medications, which can contribute to the investigation of outcomes. The clinical implications of this study may help reduce the burden of zoster and PHN in patients with T2D.

This study has some limitations. First, the National Health Insurance dataset does not contain data on biochemical tests, hemoglobin A1C levels, and immune functional tests, which may prevent an accurate evaluation of immune function and T2DM severity. This database also lacks information on body weight, diet, exercise habits, smoking, alcohol drinking, and family history. We searched for obesity, smoking, and alcohol consumption through the ICD codings. We matched clinically important variables, including sex, age, comorbidities, CCI, and medications, for maximal balance between the study and control groups. We also matched the items and number of OADs, insulin use, duration of T2D, and DCSI scores to balance the severity of T2D; and used the HbA1c check-up >2 times per year as a surrogate for hospital usage. Second, persons with severe hepatic and renal abnormalities may be concerned about lactic acidosis with metformin use. We, therefore, excluded patients on dialysis and those with hepatic failure to reduce the bias of confounding by indication. Third, the main participants in this study were Chinese; therefore, the results may not be applied to other ethnic groups. Finally, a cohort study is usually accompanied by certain unknown or unobserved confounding factors that affect the outcomes. Therefore, randomized controlled trials are needed to verify our results.

Unhealthy lifestyles and population aging have contributed to the increasing prevalence of T2D worldwide. Hyperglycemia compromises cell-mediated immunity in patients with T2D, making them susceptible to zoster and PHN. This study demonstrated that metformin use was associated with lower risks of herpes zoster and postherptic neuralgia in patients with T2D. The anti-inflammatory and antimicrobial effects of metformin need more research to repurpose it for a broader range of applications.

AUTHOR CONTRIBUTIONS

Fu-Shun Yen: study concept and design, drafting the manuscript, critical revision of the manuscript for important intellectual content, study supervision. James Cheng-Chung Wei: critical revision of the manuscript for important intellectual content, technical and material support, study supervision. Hei-Tung Yip: data acquisition, analysis, and interpretation; statistical analysis. Chih-Cheng Hsu: data analysis and interpretation, drafting the manuscript, critical revision of the manuscript for important intellectual content, and study supervision. Chii-Min Hwu: study concept and design, data acquisition, analysis, and interpretation; drafting the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; funding acquisition; technical or material support; study supervision.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

ETHICS STATEMENT

This study identified patients from the NHIRD and received approval from the Research Ethics Committee of China Medical University and Hospital [CMUH110-REC1-038(CR-1)]. The identifiable information of health providers and patients was encrypted before release to protect individual privacy. Informed consent was waived by the Research Ethics Committee.