Background/Purpose:

Diabetes mellitus is one of the SLICC/ACR Damage Index items and a recognized risk factor for cardiovascular disease and renal failure.  In rheumatoid arthritis, hydroxychloroquine is protective against diabetes.  We examined the incidence of and risk factors for diabetes in SLE.

Methods:

For each month of follow-up for a patient in a large single-site clinical SLE cohort, we calculated the proportion of previous months in the cohort that hydroxychloroquine was prescribed.  We then determined the rate of incident diabetes in groups defined by hydroxychloroquine exposure and other patient characteristics.Diabetes was defined based on the SLICC/ACR Damage Index.

Results:

The analysis was based on 1944 SLE patients.  They were 93% female.  They were 55% Caucasian, 37% African American, and 8% other ethnicity.  At the start of follow-up there were 632 (33%) under 30 years of age, 803 (41%) between 30 and 44, 408 (21%) age 45-59, and 101(5%) 60 or older.  The mean age was 37.5 years. The mean duration of follow-up per patient was 6.6 years.  682 (35%) were followed for less than 3 years, 413 (21%) were followed for 3-6 years, 258 (13%) were followed for 6-9 years, and 591 (30%) were followed for 9 or more years.

There were 59 incident cases of diabetes out of 12,802 person-years of follow-up (rate=4.6/1000).

Table 1 shows the rates of incident diabetes in subgroups defined by demographic and medication variables.  The last column shows the rate ratios found after performing a multivariable regression adjusting for all other variables in the table.  In the unadjusted analysis hydroxychloroquine appears protective (RR for >75% vs. <25%=0.5, p=0.012).  After adjustment, there is still some evidence of a protective effect (RR=0.6, p=0.067).        

Table 1: Rates of incident diabetes, by demographic and treatment variables

 

Subgroup	Events/ Person-yrs	Rate per 1000 Person-yrs	Rate Ratio (95% CI)	P-value	Adjusted Rate Ratio3	Adjusted p-value3
Everyone	59/12,802	4.6
Age   18-39   40-49   50-59   60+	19/5920 15/3414 13/2236 12/1232	3.2 4.4 5.8 9.7	1.0 (REF) 1.4 (0.7, 2.8) 1.9 (0.9, 3.9) 3.0 (1.4, 6.3)	0.33 0.085 0.0037	1.0 (REF) 1.3 (0.6, 2.8) 1.8 (0.8, 4.0) 3.8 (1.7, 8.4)	0.44 0.17 0.0012
Sex   Female   Male	57/11,858 2/948	4.8 2.1	1.0 (REF) 0.5 (0.1, 1.8)	0.26	0.2 (0.1, 1.8)	0.16
Ethnicity   White   Black   Other	23/6921 33/5197 3/688	3.3 6.3 4.4	1.0 (REF) 1.9 (1.1, 3.3) 1.3 (0.4, 4.3)	0.18 0.67	1.0 (REF) 1.2 (0.7, 2.2) 2.2 (0.6, 7.5)	0.50 0.23
Year   1987-92   1993-98   1998-04   2005-11	10/849 15/1881 14/3562 20/6513	11.8 8.0 3.9 3.1	1.0 (REF) 0.6 (0.3, 1.3) 0.3 (0.1, 0.7) 0.2 (0.1, 0.5)	0.20 0.0042 0.0002	1.0 (REF) 0.9 (0.4, 2.5) 0.5 (0.2, 1.2) 0.3 (0.1, 0.9)	0.89 0.12 0.035
BMI   <20   20-25   25-30   30+	4/949 8/3565 16/3148 25/3557	4.2 2.2 5.1 7.0	1.8 (0.6, 6.2) 1.0 (REF) 2.3 (1.0, 5.3) 3.1 (1.4, 6.9)	0.30 1.0 (REF) 0.058 0.0055	1.8 (0.6, 6.1) 1.0 (REF) 2.3 (1.0, 5.4) 3.2 (1.4, 7.3)	0.32 0.058 0.0052
Hydroxy-chloroquine1    <25%   25-75%   > 75%	27/3941 9/2029 23/6835	6.8 4.4 3.4	1.0 (REF) 0.6 (0.3, 1.4) 0.5 (0.3, 0.9)	0.24 0.012	1.0 (REF) 0.7 (0.3, 1.6) 0.6 (0.3, 1.0)	0.42 0.066
Prednisone2   <1mg/d   1-7.5 mg/d    7.5+ mg/d	15/4264 18/4372 26/4170	3.5 4.1 6.2	1.0 (REF) 0.9 (0.4, 1.7) 1.5 (0.8, 2.8)	0.66 0.17	1.0 (REF) 1.1 (0.5, 2.2) 1.5 (0.7, 2.9)	0.89 0.28

¹ Refers to percentage of prior months of cohort follow-up in which hydroxychloroquine was prescribed.

² Refers to the mean daily dose of prednisone during prior months of cohort follow-up

³ Adjusting for all variables in the table

Conclusion:

In univariate analysis, hydroxychloroquine appears protective.  However, after adjustment, while the estimated RR did not change, the result was no longer statistically significant (RR=0.6, p=0.066).  This is likely due to the fact that there is a strong association between hydroxychloroquine and year (it is being used much more frequently in recent years) and also a strong relationship between year and incidence of diabetes.  This results in a loss of power to detect an effect of hydroxychloroquine as the regression model is trying to tease out which of the two (hydroxychloroquine vs year) are the true predictors.  Although the evidence is modest (p=.066), our data are consistent with previous findings of lower rates of diabetes among those who use hydroxychloroquine.

---

« Back to 2013 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/association-between-hydroxychloroquine-exposure-and-incidence-of-diabetes-mellitus-in-systemic-lupus-erythematosus/