Observational Study Medicine ®

OPEN

Allopurinol use and type 2 diabetes incidence

among patients with gout
A VA retrospective cohort study
∗
Anastasia Slobodnick, MDa,b, , Michael Toprover, MDa,b, Jeffrey Greenberg, MD, MPHb,
Daria B. Crittenden, MD, MSb, Virginia C. Pike, BAa,b, Yingzhi Qian, MAc, Hua Zhong, PhDc,
Michael H. Pillinger, MDa,b

Abstract
To assess the impact of allopurinol on diabetes in a retrospective cohort of Veterans’ Affairs patients with gout.
The New York Harbor VA computerized patient record system was searched to identify patients with an ICD-9 code for gout
meeting at least 4 modified 1977 American Rheumatology Association gout diagnostic criteria. Patients were divided into subgroups
based on >30 continuous days of allopurinol, versus no allopurinol. New diagnoses of diabetes, defined according to American
Diabetes Association diagnostic criteria or clinical documentation explicitly stating a new diagnosis of diabetes, were identified during
an observation period from January 1, 2000 through December 31, 2015.
Six hundred six gout patients used allopurinol >30 continuous days, and 478 patients never used allopurinol. Over an average 7.9
± 4.8 years of follow-up, there was no significant difference in diabetes incidence between the allopurinol and non-allopurinol groups
(11.7/1000 person-years vs 10.0/1000 person-years, P = .27). A lower diabetes incidence in the longest versus shortest quartiles of
allopurinol use (6.3 per 1000 person-years vs 19.4 per 1000 person-years, P<.0001) was attributable to longer duration of medical
follow-up.
In this study, allopurinol use was not associated with decreased diabetes incidence. Prospective studies may further elucidate the
relationship between hyperuricemia, gout, xanthine oxidase activity, and diabetes, and the potential impact of gout treatments on
diabetes incidence.
Abbreviations: ACR = American College of Rheumatology, AMP kinase = 5’ adenosine monophosphate-activated protein
kinase, ARA = American Rheumatism Association, CPRS = computerized patient record system, HOMA-IR = homeostatic model of
insulin resistance, hsCRP = high-sensitivity c-reactive protein, ICD-9 = international classification of diseases, 9th ed., MPR =
medication possession ratio, NYH VA = New York Harbor Veterans’ Affairs Healthcare System.
Keywords: Crystal arthritis, diabetes, allopurinol

1. Introduction as urate-related inhibition of the enzyme AMP kinase, which may

Several studies suggest that patients with gout are at increased play a central role in the effects of high-fructose diets on both gout
risk for developing type 2 diabetes.[1–4] Although the basis for this and diabetes.[5] Hyperuricemia may also be a risk factor for the
risk is not known, possible explanations include increased development of hyperinsulinemia and insulin resistance.[6]
systemic inflammation that may impair insulin sensitivity, as well Alternatively, recent studies suggest that increased activity of

Editor: Raouf Hajji.

JG is currently the Chief Scientific Officer for CORRONA. DBC is currently employed by Terns Pharmaceuticals. MHP is supported in part by NYU CTSA grant
1UL1TR001445 from the National Center for the Advancement of Translational Science (NCATS). NIH has served as a consultant for AstraZeneca, Crealta/Horizon and
Sobi, and has received investigator-initiated grant funding from Hikma and Horizon Therapeutics.
The authors have no funding and conflicts of interest to disclose.
Supplemental Digital Content is available for this article.
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
a
Rheumatology Section, Department of Medicine, VA NY Harbor Health Care System, NY Campus, b Division of Rheumatology, Department of Medicine, NYU School
of Medicine/NYU Langone Health, c Division of Biostatistics, Department of Population Health, NYU Langone Health, NY, USA.
∗
Correspondence: Anastasia Slobodnick, Division of Rheumatology, NYU Langone Orthopedic Hospital, 301 East 17th Street, Suite 1410, NY 10003
(e-mail: As3564@nyu.edu).
Copyright © 2020 the Author(s). Published by Wolters Kluwer Health, Inc.
This is an open access article distributed under the terms of the Creative Commons Attribution-Non Commercial License 4.0 (CCBY-NC), where it is permissible to
download, share, remix, transform, and buildup the work provided it is properly cited. The work cannot be used commercially without permission from the journal.
How to cite this article: Slobodnick A, Toprover M, Greenberg J, Crittenden DB, Pike VC, Qian Y, Zhong H, Pillinger MH. Allopurinol use and type 2 diabetes incidence
among patients with gout: a VA retrospective cohort study. Medicine 2020;99:35(e21675).
Received: 3 September 2019 / Received in final form: 29 May 2020 / Accepted: 9 July 2020
http://dx.doi.org/10.1097/MD.0000000000021675

1
Slobodnick et al. Medicine (2020) 99:35 Medicine

xanthine oxidase (the enzyme that generates uric acid), Patients were assigned to the control group if they had never
independent of serum urate level, may be associated with the received an allopurinol prescription during the study period.
development of hyperinsulinemia and type 2 diabetes.[7,8] Patients who filled prescriptions for allopurinol for 30 days or
Despite the evidence that xanthine oxidase activity and/or fewer during the study period were excluded. Allopurinol and
hyperuricemia may be associated with diabetes risk, studies of colchicine use information, including dosing and dates of
xanthine oxidase inhibitor use in diabetes have yielded mixed initial prescription and refills, were obtained from the NYH
results. One brief trial randomized 73 patients with asymptom- VA electronic pharmacy records. For each patient in the
atic hyperuricemia and no diabetes to receive either allopurinol allopurinol group, the index date of observation was defined
300 mg daily or placebo. After 3 months of treatment, patients in as the date of the earliest allopurinol prescription within the study
the allopurinol group had significantly improved fasting blood period. For each patient in the non-allopurinol group, the index
glucose, fasting insulin, Homeostatic Model Assessment of date was defined as the later of either the date of gout diagnosis or
Insulin Resistance (HOMA-IR), and high-sensitivity C-reactive January 1, 2000.
protein (hsCRP) levels compared to patients in the hyperuricemic The primary outcome of the study was the incidence of new
placebo group.[9] In an observational study of a cohort of patients type 2 diabetes diagnoses in each group. Incident diabetes was
followed prospectively for 4 years, patients who used allopurinol defined either as meeting 1 criterion from the American Diabetes
and had improved serum urate levels had significantly improved Association’s definition of diabetes (a hemoglobin A1c that was
plasma fasting blood glucose levels. However, patients who newly ≥ 6.5%, a fasting plasma glucose ≥126 mg/dL, an oral
experienced spontaneously improved serum urate levels without glucose tolerance test 2 hour blood glucose ≥200 mg/dL, or a
the use of allopurinol did not experience a significant change in random blood glucose ≥ 200 mg/dL in the setting of hyperglyce-
their plasma fasting blood glucose levels, suggesting that the mic symptoms); or physicians’ notes within the study period
allopurinol effect may not have been linked directly to serum explicitly stating that the subject had a new diagnosis of diabetes.
urate lowering.[10] In contrast, a study of 41 patients with In the case of physician diagnosis, charts were reviewed for
diabetes, randomized to receive allopurinol 100 mg or placebo 3 corroborating data, and patients lacking corroborating data (e.g.,
times daily for 14 days, demonstrated no significant reduction in subsequently meeting ADA criteria, treatment with glucose
fasting blood glucose or hemoglobin A1c levels in either lowering agents, or explicit documentation of disease, and
group.[11] Furthermore, a study utilizing Taiwan’s national management plan) were excluded. Patients were also excluded if
health insurance research database found a direct and dose- chart review indicated evidence that their diabetes could
dependent relationship between allopurinol use and diabetes possibly be of a type other than type 2 (e.g., alcoholic, viral,
risk.[12] or malignant pancreas damage, evidence of diabetic ketoacidosis,
In the present study, we aimed to clarify the association or explicit physician documentation of a non-type 2 diabetes).
between allopurinol use and diabetes incidence by conducting a We hypothesized that gout patients taking allopurinol would
retrospective cohort study in a well-characterized population of experience a lower incidence of type 2 diabetes than those not
patients with gout who had received care at the New York taking allopurinol.
Harbor VA Health Care System of the United States Department Student t test (continuous variables) and Fisher exact test
of Veterans Affairs (NYH VA). (categorical variables) were used to compare baseline character-
istics. Categorical outcome variables, including the primary
2. Methods endpoint, were analyzed utilizing Fisher exact test. Continuous
outcome variables were analyzed using Cox proportional
We utilized a previously established database of subjects with
hazards regression analysis. Multivariate analysis was performed
gout, which was collected starting in 2008, and extended it for 5
using R software version 3.6.1. The remainder of the statistical
additional years.[13] For this dataset, we identified all patients
analysis was performed using Stata Statistical Software: Release
with an International Classification of Diseases, 9th ed. (ICD-9)
15. College Station, TX: StataCorp, LLC. This study was
code for gout (274.xx) and at least 1 visit documented in the
approved by the Institutional Review Board of the New York
NYH VA computerized patient record system (CPRS) between
Harbor Veterans’ Affairs Healthcare System, approval number
August 1, 2007 and August 1, 2008. Each patient’s record was
01700. Informed consent for each individual patient was waived
manually reviewed to confirm a diagnosis of gout according to
for this study as this was a retrospective analysis of data which
the 1977 American Rheumatism Association (ARA) classification
was collected for the purposes of clinical care.
criteria (with the use of urate lowering therapy accepted as a
surrogate for hyperuricemia).[14] Patients’ clinical notes, vital
signs, and laboratory results were reviewed to identify demo- 3. Results
graphics, comorbidities, and the use of steroids and statins.
3.1. Patients and demographics
Patients who carried a diagnosis of diabetes at baseline were
excluded from the study. 40,107 patients had at least 1 documented clinical visit within the
The study period was defined as January 1, 2000 through NYH VA CPRS between August 1, 2007 and August 1, 2008.
December 31, 2015. Patients were divided into groups based on From among these patients, 6442 had an ICD-9 code for gout,
whether or not they used allopurinol during the study period. and 2195 of those patients met at least 4 of the 1977 ARA gout
Patients were included in the allopurinol use group if they filled at classification criteria. Of these, 587 patients were excluded for
least 2 or more consecutive prescriptions totaling >30-days for diabetes at baseline, and 91 additional patients were excluded for
allopurinol. We chose this threshold to confirm the intent of the not meeting the inclusion criterion for allopurinol use during the
treating physician to place the patient on chronic allopurinol study period. After applying exclusion criteria, 636 patients were
therapy, to confirm at least a minimal commitment to patient included in the allopurinol group, and 478 patients were included
compliance, and to ensure at least a minimal treatment period. in the non-allopurinol group (Fig. 1).

2
Slobodnick et al. Medicine (2020) 99:35 www.md-journal.com

Paents with acve

records 2007 – 2008
(N = 40,107)

Paents with ICD-9 code

for gout
(N = 6442)
Allopurinol use ≤ 30
connuous days Paents meeng ≥4 ARA
between 1/1/2000 and gout diagnosc criteria
12/31/2015 (N=91) (N = 2195)
excluded

Allopurinol use > 30 No allopurinol use

connuous days between 1/1/2000 and Diabetes diagnosed
Diabetes diagnosed between 1/1/2000 and 12/31/2015 (N = 681)
before 1/1/2000 or <1/1/2000 or prior to
12/31/2015 (N = 1033) excluded gout diagnosis (N = 197)
prior to inial excluded
allopurinol prescripon
(N = 390)
Paents without Paents without
diabetes at baseline diabetes at baseline Type 1 diabetes or
Type 1 diabetes or (N=643) (N=484) unclear diabetes
unclear diabetes excluded excluded type (N=6)
type (N=7)

Incidence of new diabetes No new diabetes No new diabetes Incidence of new diabetes
between 1/1/2000 and diagnosis between diagnosis between between 1/1/2000 and
12/31/2015: HbA1c newly ≥ 1/1/2000 and 1/1/2000 and 12/31/2015: HbA1c newly
6.5% or clinical notes stang 12/31/2015 (N = 517) 12/31/2015 (N = 401) ≥ 6.5% or clinical notes
new type 2 diabetes stang new diabetes
diagnosis (N = 119) diagnosis (N = 77)
Colchicine use, duraon
Colchicine use, duraon of allopurinol use/follow Colchicine use, duraon
of allopurinol use/follow up, allopurinol of follow-up Colchicine use, duraon
up, allopurinol medicaon possession of follow-up
medicaon possession rao
rao
Figure 1. Flowchart depicting patient selection and analysis.

Table 1 compares the baseline demographic information, colchicine during the study period. Patients in the allopurinol
comorbidities, use of statins, steroids, and/or colchicine between group were also less likely to have chronic kidney disease (CKD),
the allopurinol and non-allopurinol groups, collected at the study presumably because of concerns about using allopurinol in
index date for each patient. Patients in the allopurinol group were patients with CKD prior to the publication of the American
more likely to be black, less likely to have used statins during the College of Rheumatology 2012 Gout Treatment Guide-
study period, and more likely to have used steroids and/or lines.[15,16] Supplementary Table 1, http://links.lww.com/MD/
E746 shows the baseline characteristics of the excluded patients.
Table 1
Baseline demographic information, co-morbidities, and medica- 3.2. Overall diabetes incidence
tion use of the study sample.
One hundred sixty-six patients were diagnosed with incident
Allopurinol Non-allopurinol
diabetes based on elevated hemoglobin A1c levels or elevated
Characteristic cohort cohort P value
fasting blood glucose, while 41 patients were diagnosed with
Number of subjects 636 478 incident diabetes based on physician diagnosis with corroborat-
Age in yr, mean ± SD 71.3 ± 12.14 71.8 ± 13.75 .47 ing data support. Four patients (2 taking and 2 not taking
Male sex, n (%) 633 (99.5) 473 (99.0) .27
allopurinol) with a physician diagnosis of incident diabetes who
Race: 356 (56.0) 283 (59.2) baseline
- White, n (%) 248 (39.0) 145 (30.3) .019
were lacking corroborating data were excluded from further
- Black, n (%) 5 (0.8) 12 (2.5) .040 analysis. Twenty patients in the allopurinol group were
- Asian/Pacific islander, n (%) 27 (4.2) 38 (7.9) .030 diagnosed with incident diabetes by physician diagnosis,
- Other/unknown, n (%) 66 (10.3) 69 (14.4) .15 compared to 21 patients in the non-allopurinol group (P = .43).
Hispanic, n (%) Among patients in the allopurinol group, we observed no
BMI, mean ± SD 29.6 ± 5.53 29.0 ± 5.62 .12 significant difference in incidence of type 2 diabetes compared to
Chronic kidney disease, n (%) 198 (31.5) 183 (38.6) .015 the non-allopurinol group, over an average 7.9 ± 4.8 years of
Medication use: 377 (59.3) 170 (35.6) <.001 follow-up (11.7 per 1000 person-years in allopurinol group
- Steroids, n (%) 274 (43.1) 301 (63.0) <.001 compared to 10.0 per 1000 person-years in the non-allopurinol
- Statins, n (%) 450 (70.3) 157 (33.2) <.001
group, P = .27) (Fig. 2A). Among patients on allopurinol who
- Colchicine, n (%)
experienced a new diagnosis of type 2 diabetes during the

3
Slobodnick et al. Medicine (2020) 99:35 Medicine

25 20 p = 0.94
20
Incidence p=0.27 Incidence 15
per 15 per
1000 1000 10
10
person- person-
years 5 years 5

0
0
Allopurinol Non-allopurinol Allopurinol + + - -
A B Colchicine + - + -

p<0.0001
= Allopurinol
p<0.0001 25 = No allopurinol
25
p<0.0001
20
20
Incidence Incidence
per 15
per 15
1000 1000 10
10 person-
person-
years years 5
5
0 0
1st 2nd 3rd 4th 1st 2nd 3rd 4th
quarle quarle quarle quarle D quarle quarle quarle quarle
C
Figure 2. (A) Overall incidence of type 2 diabetes among patients in the allopurinol and non-allopurinol groups during study period, (B) Diabetes incidence among
patients who were prescribed both allopurinol and colchicine, only allopurinol, only colchicine, and neither medication during the study period, (C) Diabetes
incidence for each quartile
∗
of duration of allopurinol use among patients in the allopurinol group, (D) Diabetes incidence for each quartile of duration of observation
among all patients. ANOVA analysis across all categories.

observation period, the mean duration of allopurinol use prior to during the study period. Patients in the longest quartile of
diabetes diagnosis was 4.0 ± 3.50 years and the mean allopurinol allopurinol use had a significantly lower incidence rate of type 2
dose was 211.9 ± 90.6 mg daily. The mean duration of diabetes than those in the shortest quartile of use (6.3 per 1000
observation prior to diabetes diagnosis in control patients was person-years vs 19.4 per 1000 person-years, P<.0001) (Fig. 2C).
4.17 ± 3.6 years (P = .76). However, when patients in the allopurinol group were divided
into quartiles based on duration of observation rather than
allopurinol treatment, we observed a similar difference in
3.3. Colchicine use diabetes incidence rate between patients in the longest quartile
Previous studies have suggested that colchicine acts on pathways of observation and those in the shortest quartile (3.3 per 1000
common to gout and diabetes, including the interleukin-1b/ person-years vs 15.2 per 1000 person-years, P<.0001) (Fig. 2D).
cryopyrin and AMP kinase pathways.[5,17,18] To evaluate for a Moreover, when patients in the non-allopurinol group were
possible confounding colchicine effect on diabetes incidence, divided into quartiles based on duration of observation, we saw a
patients were categorized into 4 groups: similar reduction in diabetes incidence rate with longer
observation. To further assess for the possibility that the duration
1) those who used both allopurinol and colchicine for >30 days of follow-up might be influencing diabetes incidence, patients in
during the study period (N = 448); the shortest 2 quartiles of allopurinol treatment duration were
2) those who used only allopurinol (N = 140); divided into those patients who had short observation periods
3) those who used only colchicine (N = 157), and (defined as less than the mean 7.9 years duration of follow-up,
4) those who used neither allopurinol nor colchicine (N = 271). n = 68) and those who had long observation periods (defined as ≥
No significant difference in diabetes incidence was found 8 years, n = 252). Diabetes incidence rate was found to be
between these 4 groups (Fig. 2B). significantly higher for the short observation group compared to
the long observation group, (23.5 per 1000 person-years
compared to 6.6 per 1000 person-years, P < .001), despite
3.4. Duration of allopurinol use
similar durations of allopurinol treatment (Fig. 3A). Conversely,
The mean duration of allopurinol use was 3.62 ± 3.33 years. The when patients with a long duration of observation (defined as ≥ 8
mean duration of observation was 9.17 ± 4.62 years for the years) were divided into those with a short duration of
allopurinol group and 6.69 ± 4.31 years for the control group allopurinol exposure (lowest 2 quartiles, n = 233) and long
(P = <.00001). To assess whether duration of allopurinol use duration of allopurinol exposure (highest 2 quartiles, n = 140), no
affected diabetes incidence, patients in the allopurinol group were significant difference in diabetes incidences was found between
divided into quartiles based on the total number of months of use the short allopurinol group and the long allopurinol group (6.3

4
Slobodnick et al. Medicine (2020) 99:35 www.md-journal.com

3.5. Allopurinol dose

p<0.001
To assess for dose effect, diabetes incidence was also compared
25
between patients on higher doses of allopurinol (average dose
20 ≥300 mg daily) and those who received no allopurinol during
Incidence the study period. There was no significant difference in diabetes
15
per incidence between patients on high doses of allopurinol and
1000 those on no allopurinol (11.6 per 1000 person-years in high
10
person-
years
allopurinol group vs 10.0 per 1000 person-years in non-
5
allopurinol group, P = .42). There was also no significant
0 difference in diabetes incidence between patients on high doses
A Short follow-up Long follow-up of allopurinol and low doses (<300 mg) of allopurinol
(incidence was 13.0 per 1000 person-years for low allopurinol
25 dose; P = 1.0) (Table 2).

20
Incidence 3.6. Allopurinol adherence
per 15
p=0.86 To account for non-adherence among patients in the allopurinol
1000
person-
10 group, diabetes incidence was stratified by quartiles of medica-
years 5
tion possession ratio (MPR). There was no significant difference
in diabetes incidence between the highest and lowest MPR
0 quartiles (10.1 per 1000 person-years for the highest quartile
B Short allopurinol Long allopurinol compared with 14.0 per 1000 person-years for the lowest
Figure 3. (A) Diabetes incidence among patients in the lowest 2 quartiles of quartile, P = .20) (Table 2).
allopurinol use, stratified by duration of observation, (B) Diabetes incidence
among patients with a long duration of follow-up, stratified by duration of
allopurinol use. 3.7. Serum urate
To evaluate for a possible effect of serum urate concentration on
diabetes development, diabetes incidence was calculated accord-
ing to baseline serum urate level, as well as by extent of serum
per 1000 person-years compared to 5.8 per 1000 person-years, urate change over the first 6 to 12 months of observation,
P = .86) (Fig. 3B). These data suggest that longer duration of regardless of allopurinol use status. There was no association
allopurinol treatment was not responsible for the lower incidence between either baseline serum urate level (OR 0.96, 95% CI
rate of diabetes observed in the long observation group. 0.87–1.06) or extent of serum urate change (OR 1.06, 95% CI
Given the possibility that new therapies and/or practice 0.93–1.2) and diabetes incidence.
guidelines may have affected diabetes incidence over the 16- Diabetes incidence was further assessed according to extent of
year course of our study period, we calculated the diabetes serum urate change 6 to 12 months after allopurinol initiation,
incidence of patients in the allopurinol and non-allopurinol for patients in the allopurinol group. There was no association
groups by calendar year. No trend in diabetes incidence was between extent of serum urate change and diabetes incidence in
found when comparing the years 2000 and 2015 (Table 2). this group (OR 1.07, 95% CI 0.94–1.22). Additionally, we
assessed whether achieving the recommended target serum urate
Table 2 level for most patients with gout (defined as 6 mg/dL) was
related to reduced diabetes incidence.[15] Patients with available
Diabetes incidence among patients taking allopurinol and those
not taking allopurinol, stratified by calendar year of diabetes
follow-up serum urate levels were divided into those who
diagnosis, allopurinol dose, allopurinol adherence and serum urate achieved the target serum urate level 6 to 12 months after
levels. allopurinol initiation (N = 119), and those who did not (N = 217).
Allopurinol Non-allopurinol
There was no significant difference in diabetes incidence between
Characteristic cohort cohort P value these 2 groups (13.8 per 1000 person-years in the achieved target
urate group vs 12.7 per 1000 person-years in those who did not
Yr
achieve target urate level, P = .78) (Table 2).
2000 4.7 8.2 .47
2015 9.4 4.9 .71
Allopurinol average dose: 3.8. Multivariate analysis
<300 mg daily 13.0 1.0
≥300 mg daily 11.6 Diabetes incidence during the study period was analyzed in
Allopurinol medication-possession multivariate logistic and Cox regressions including allopurinol
ratio: use, baseline age, race, BMI, CKD, baseline serum uric acid, statin
Lowest MPR quartile 14.0 .2 use, prednisone use, and colchicine use. The multivariate logistic
Highest MPR quartile 10.1 regression demonstrated no significant association between
Serum urate level achieved after allopurinol use and diabetes incidence. However, the Cox
allopurinol initiation: proportional hazards model demonstrated a significantly in-
6 mg/dL 13.8 .78 creased hazard ratio for diabetes incidence among subjects taking
>6 mg/dL 12.7
allopurinol compared to subjects who did not take allopurinol
MPR = medication-possession ratio. (Table 3).

5
Slobodnick et al. Medicine (2020) 99:35 Medicine

Table 3
Cox multivariate regression analysis of potential risk factors among patients diagnosed with diabetes.
Model 1 Model 2 Model 3 Model 4
Unadjusted Adjusted Adjusted Adjusted Adjusted
Variable HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value
Allopurinol use 1.43 (1.06, 1.92) .018 1.41 (1.05, 1.89) .024 1.37 (1.01, 1.87) .043 1.46 (1.06, 2.01) .019 1.70 (2.05, 2.76) .032
during study
Age 0.99 (0.98, 1.00) .018 0.99 (0.98, 1.00) .104 0.99 (0.98, 1.00) .12 0.99 (0.97, 1.00) .119
Race: -
- White
- Black 1.48 (1.11,1.98) .008 1.33 (0.99, 1.79) .062 1.33 (0.99, 1.80) .058 1.23 (0.84, 1.81) .288
- Asian/Pacific islander (0.16, 2.68) .559 0.87 (0.21, 3.56) .847 0.80 (0.20, 3.30) .763 1.10 (0.15, 8.26) .923
- Other/unknown (0.42, 1.96) .809 1.02 (0.47,2.21) .955 0.99 (0.46, 2.14) .975 0.79 (0.19, 3.29) .744
BMI, mean ± SD 1.06 (1.03, 1.08) <.001 1.06 (1.03,1.08) <.001 1.05 (1.02, 1.08) .001
Chronic kidney disease 0.87 (0.58, 1.32) .518
Medication use:
- Steroids 0.80 (0.59, 1.07) .132 0.73 (0.50, 1.07) .103
- Statins 1.42 (0.96, 2.10) .083
Baseline serum uric acid 1.03 (0.94, 1.14) .503
Colchicine use 0.95 (0.63, 1.43) .813
Model 1: n = 1127, adjusted for age and race. Model 2: n = 1036, adjusted for age, race, and BMI. Model 3: n = 637, adjusted for age, race, BMI, serum uric acid level, statin use, prednisone use, colchicine use,
and chronic kidney disease.

3.9. Gout diagnosed during study period patients with gout. Diabetes incidence did decline with longer
Patients with gout whose index date was January 1, 2000, (first duration of allopurinol use, but this effect appeared to be related
date of observation period) had gout diagnoses of indeterminate to duration of observation rather than to duration of allopurinol
time prior to that date, setting up a possible chronological bias in itself. On the other hand, we noted a significant increase in the
the study design. To address this issue, a subset of patients who hazard ratio for diabetes among subjects taking allopurinol in the
were unequivocally diagnosed with gout after January 1, 2000 Cox proportional hazards regression analysis, results which
based on electronic medical records was selected from the total contrasted with the rest of the findings of the study, and which are
sample (N = 333), and diabetes incidence was compared in not explained by the analyses of comorbidities, medication use,
patients who used allopurinol within that group (N = 156), versus and duration of follow-up that we performed. It is possible that
those who did not (N = 177). Supplementary Table 2, http://links. other factors associated with allopurinol use were not captured
lww.com/MD/E746 compares the baseline demographic infor- by the data collected for this study, and contributed to the
mation between the allopurinol and non-allopurinol groups increased hazard ratio.
within this subset. Patients in this subanalysis were similar in This study also found no association between colchicine use
most respects, except that the patients who had taken allopurinol and diabetes incidence, in spite of evidence suggesting that
were less likely to be Hispanic, and more likely to have taken colchicine acts on several pathophysiologic mechanisms involved
steroids, compared with the non-allopurinol group. No differ- in diabetes. A recent study of more than 27,000 Veterans’ Affairs
ence in diabetes incidence was found between the 2 groups (9.2 patients with gout also found no overall association between
per 1000 person-years in both groups). In addition, there was no colchicine use and risk of diabetes, although there was a trend
significant difference in the average time from gout diagnosis to toward decreased incidence with increased doses and duration of
new diabetes diagnosis between patients who used allopurinol colchicine use.[20] Since we did not analyze diabetes incidence by
and those who did not (53.9 months ± 40.93 vs 51.0 months ± colchicine dose in our study, we are unable to comment on
38.9, P = .78). As for the entire cohort, a decreased incidence rate whether using higher doses of colchicine may be associated with
of diabetes with longer allopurinol treatment was likely reduced diabetes incidence.
confounded by duration of observation (Supplementary Fig. 1, Our study had several strengths. The sample size was large, and
http://links.lww.com/MD/E745). patients were followed for a long duration of time. We rigorously
defined gout and confirmed the diagnosis with direct chart review
for each subject. The majority of patients with new-onset diabetes
4. Discussion
had documented hemoglobin A1c levels ≥ 6.5%, an objective and
Based on several biological and clinical observations, recent widely accepted measure of diabetes. The study also utilized
studies suggest that allopurinol use may be associated with allopurinol use data collected in a real-world healthcare setting,
improved insulin sensitivity and blood glucose levels. There is and was thus likely reflective of patients’ everyday patterns of
evidence that decreased urate may account for some of this allopurinol use. Indeed, our ability to individually review the
benefit, but suppression of xanthine oxidase activity may also patient’s records during initial data collection rendered our study
play a role. Xanthine oxidase activity has been linked to increased relatively protected from errors of coding.
reactive oxygen species generation in the setting of hyperglyce- There are several important limitations to this study. Our use
mia, and allopurinol has been found to prevent this effect.[19] of 1977 ARA criteria for gout, instead of the more recent 2015
Nevertheless, our study did not demonstrate an association American College of Rheumatology (ACR) classification criteria,
between allopurinol use and decreased diabetes incidence among was predicated on the fact that data collection for this data set

6
Slobodnick et al. Medicine (2020) 99:35 www.md-journal.com

was begun before 2015.[21] However, our use of ICD-9 diagnosis [5] Thottam GE, Krasnokutsky S, Pillinger MH. Gout and metabolic
syndrome: a tangled web. Curr Rheumatol Rep 2017;19:60.
as an entry criterion before applying ARA criteria ensured that
[6] Niskanen L, Laaksonen DE, Lindstrom J, et al. Serum uric acid as a
our ascertainment of gout was at least 95% specific.[22] Given the harbinger of metabolic outcome in subjects with impaired glucose
retrospective nature of our design, it is likely that we were unable tolerance: the Finnish diabetes prevention study. Diabetes Care
to capture all of the allopurinol use for all of the patients. 2006;29:709–11.
Similarly, it is impossible for us to fully assess adherence to [7] Li X, Meng X, Gao X, et al. Elevated serum xanthine oxidase activity is
associated with the development of type 2 diabetes: a prospective cohort
allopurinol therapy, although we approached this issue through study. Diabetes Care 2018;41:884–90.
several sub-analyses. The fact that allopurinol use was shorter [8] Sunagawa S, Shirakura T, Hokama N, et al. Activity of xanthine oxidase
than the observational period for allopurinol users indicates that in plasma correlates with indices of insulin resistance and liver
allopurinol use was intermittent for many patients, consistent dysfunction in patients with type 2 diabetes mellitus and metabolic
syndrome: a pilot exploratory study. J Diabetes Investig 2019;10:94–
with prior reports; whether continuous allopurinol use would
103.
have a greater impact on diabetes incidence cannot be [9] Takir M, Kostek O, Ozkok A, et al. Lowering uric acid with allopurinol
determined.[23,24] We were not able to assess the impact of other improves insulin resistance and systemic inflammation in asymptomatic
urate-lowering therapies, since their use was either negligible or hyperuricemia. J Investig Med 2015;63:924–9.
not available at the initiation of the study. Additionally, although [10] Cicero AFG, Rosticci M, Bove M, et al. Serum uric acid change and
modification of blood pressure and fasting plasma glucose in an overall
we utilized all 4 of the 2018 American Diabetes Association healthy population sample: data from the Brisighella heart study. Ann
criteria (along with physician notation of new diabetes) to Med 2017;49:275–82.
diagnose diabetes in our cohort, only hemoglobin A1c ≥ 6.5% [11] Afshari M, Larijani B, Rezaie A, et al. Ineffectiveness of allopurinol in
was commonly observed.[25] Therefore, it is possible that we did reduction of oxidative stress in diabetic patients; a randomized, double-
blind placebo-controlled clinical trial. Biomed Pharmacother
not capture all of the incident diabetes cases for our sample. It is
2004;58:546–50.
also possible that some of the patients characterized as having [12] Chang H-W, Lin Y-W, Lin M-H, et al. Associations between urate-
new diabetes in our study may have had diabetes at baseline, lowering therapy and the risk of type 2 diabetes mellitus. Shimosawa T,
which was either not documented properly or was not properly ed. PLoS One 2019;14:e0210085.
evaluated in a medical setting until a date within our study [13] Crittenden DB, Lehmann RA, Schneck L, et al. Colchicine use is
associated with decreased prevalence of myocardial infarction in patients
period. Patients on allopurinol were more likely to have with gout. J Rheumatol 2012;39:1458–64.
received steroids, which could have contributed to increased [14] Wallace SL, Robinson H, Masi AT, et al. Preliminary criteria for the
diabetes incidence and obscured any beneficial effect of urate classification of the acute arthritis of primary gout. Arthritis Rheum
lowering. 1977;20:895–900.
[15] Khanna D, Khanna PP, Fitzgerald JD, et al. 2012 American College of
Rheumatology guidelines for management of gout. Part 2: therapy and
5. Conclusion antiinflammatory prophylaxis of acute gouty arthritis. Arthritis Care Res
(Hoboken) 2012;64:1447–61.
This study found that allopurinol use was not associated with [16] Toprover M, Crittenden DB, Modjinou DV, et al. Low-dose allopurinol
decreased incidence of type 2 diabetes among patients with gout. promotes greater serum urate lowering in gout patients with chronic
No association with diabetes incidence was found when kidney disease compared with normal kidney function. Bull Hosp Jt Dis
allopurinol use was stratified by dose or by measures of 2019;77:87–91.
[17] Wang Y, Viollet B, Terkeltaub R, et al. AMP-activated protein kinase
adherence to allopurinol therapy. Diabetes incidence was not suppresses urate crystal-induced inflammation and transduces colchicine
associated with baseline serum urate or change in serum urate effects in macrophages. Ann Rheum Dis 2016;75:286–94.
level. Colchicine use was also not associated with diabetes [18] Ehses JA, Lacraz G, Giroix M-H, et al. IL-1 antagonism reduces
incidence. Whether allopurinol use could reduce diabetes hyperglycemia and tissue inflammation in the type 2 diabetic GK rat.
Proc Natl Acad Sci 2009;106:13998–4003.
incidence in a more narrowly defined subset of patients, or in
[19] Eleftheriadis T, Pissas G, Antoniadi G, et al. Allopurinol protects human
patients with hyperuricemia without gout, remains to be glomerular endothelial cells from high glucose-induced reactive oxygen
determined, most likely through prospective trials. species generation, p53 overexpression and endothelial dysfunction. Int
Urol Nephrol 2018;50:179–86.
[20] Wang L, Sawhney M, Zhao Y, et al. Association between colchicine and
Author contributions risk of diabetes among the veterans affairs population with gout. Clin
XXXX. Ther 2015;37:1206–15.
[21] Neogi T, Jansen TLTA, Dalbeth N, et al. 2015 gout classification criteria:
an American college of rheumatology/European league against rheuma-
References tism collaborative initiative. Ann Rheum Dis 2015;74:1789–98.
[22] Singh JA. Veterans affairs databases are accurate for gout-related health
[1] Kim SC, Liu J, Solomon DH. Risk of incident diabetes in patients with care utilization: a validation study. Arthritis Res Ther 2013;15:R224.
gout: a cohort study. Arthritis Rheumatol 2015;67:273–80. [23] Mantarro S, Capogrosso-Sansone A, Tuccori M, et al. Allopurinol
[2] Choi HK, De Vera MA, Krishnan E. Gout and the risk of type 2 diabetes adherence among patients with gout: an Italian general practice database
among men with a high cardiovascular risk profile. Rheumatology study. Int J Clin Pract 2015;69:757–65.
2008;47:1567–70. [24] Riedel AA, Nelson M, Joseph-Ridge N, et al. Compliance with
[3] Tung Y-C, Lee S-S, Tsai W-C, et al. Association between gout and allopurinol therapy among managed care enrollees with gout: a
incident type 2 diabetes mellitus: a retrospective cohort study. Am J Med retrospective analysis of administrative claims. J Rheumatol 2004;
2016;129:1219e17–25. 31:1575–81.
[4] Pan A, Teng GG, Yuan J-M, et al. Bidirectional association between [25] American Diabetes Association2. Classification and diagnosis of
diabetes and gout: the Singapore Chinese health study. Sci Rep diabetes: standards of medical care in diabetes-2018. Diabetes Care
2016;6:25766. 2018;41(Suppl 1):S13–27.