Similar burden of type 2 diabetes among adult patients with sickle cell disease relative to African Americans in the U.S. population: a six-year population-based cohort analysis

Summary

Conflicting evidence exists on the epidemiology of type 2 diabetes mellitus (T2DM) among patients with sickle cell disease (SCD). This study measured the prevalence, incidence, and clinical outcomes associated with T2DM in a large U.S. population of commercially insured adults aged ≥20 years with SCD between 2009 and 2014. Among 7,070 patients with SCD, the mean age (median) was 39 (37) years and 60.8% were female. The standardized prevalence of T2DM among patients with SCD showed a modest increase from 15.7% to 16.5% (p trend=0.026) and was comparable to African-American respondents to the National Health and Nutrition Examination Survey (18.2%). Over 17,024 person-years, the crude incidence rate for T2DM was 25.4 per 1000 person-years. Incident T2DM was associated with comorbid hypertension (HR=1.45, 95%CI 1.14–1.83), and dyslipidaemia (HR=1.43, 95%CI 1.04–1.96). Compared to SCD patients without T2DM, more SCD patients with T2DM had diagnoses of nephropathy (28.0% vs. 9.5%; p<0.001), neuropathy (17.7% vs. 5.2%; p<0.001), and stroke (24.1% vs. 9.2%; p<0.001). Prevalence of T2DM in SCD patients is similar to the general African American population with an increasing trend in recent years. These trends support routine screening for T2DM in aging patients with SCD, especially those with comorbid hypertension and/or dyslipidaemia.

Introduction

The prevalence and incidence of type 2 diabetes mellitus (T2DM) in the United States (U.S.) is increasing with more than 100 million adults living with diabetes or pre-diabetes and 1.5 million new cases of diabetes diagnosed annually.(CDC 2017, Geiss, et al 2014) African Americans share a disproportionately high burden of diabetes and diabetes-related complications, with an up to two-fold increased likelihood of developing diabetes compared to non-Hispanic White adults.(Carter, et al 1996) These racial disparities in diabetes may be attributable to lifestyle differences as well as socioeconomic and genetic differences between these populations.(Abate and Chandalia 2003)

Sickle cell disease (SCD) is an autosomal recessive hemoglobinopathy that mostly affects persons of African origin in the U.S.(Piel, et al 2017) It is caused by homozygosity for a Glu6Val mutation in HBB or compound heterozygous forms with other HBB mutations. It remains unclear whether patients with SCD have similar risks for developing diabetes compared to African Americans without SCD. Iron-overload, a frequent complication induced by transfusion and chronic haemolysis in patients with SCD, could lead to long-term progressive damage to the endocrine system.(Porter and Garbowski 2013) It is hypothesized that the impaired metabolic homeostasis of glucose secondary to pancreatic β-cell destruction could contribute to the early onset of insulin resistance, carbohydrate intolerance and eventually diabetes among patients with SCD.(Smiley, et al 2008) On the other hand, SCD is associated with a shortened life expectancy,(Platt, et al 1994) so development of T2DM may be apparently diminished due to premature death among patients with SCD.

Definitive population-based evidence on the prevalence and risks for diabetes in SCD patients has yet to be documented. Historically, SCD was associated with low prevalence of diabetes. A large multi-centre study involving 31 clinical haematology centres in the U.S., Canada and the U.K. showed lower than expected prevalence of diabetes in the SCD population. Only 2% of treated SCD patients had diabetes and transfusion duration was strongly associated with T2DM.(Fung, et al 2006) Similarly, an epidemiologic study in Bahrain found relatively high prevalence of diabetes (8.3%) but it was lower than the general population after adjusting for age and sex.(Mohamed, et al 2015) In another longitudinal registry study, only 10 patients with SCD and diabetes were identified from the German-Austrian DPV (Diabetes Patienten Verlaufsdokumentation) registry and all received insulin treatment, suggestive of type 1 diabetes mellitus.(Warncke, et al 2016) These studies are limited due to relatively small sample sizes, lack of a nationally representative sample, non-contemporary study periods and shorter follow-up. Moreover, no study has documented trends in the annual prevalence, incidence and clinical outcomes of diabetes over time.

Greater insight on the burden of diabetes in patients with SCD will inform prevention strategies in this population with increasing life expectancy due to improved care. The purpose of this study was to estimate the prevalence, incidence, clinical outcomes and factors associated with T2DM in a commercially-insured population of adults with SCD using a large, nationally representative U.S. administrative claims database.

Methods

Data sources

Truven Health MarketScan® Databases

Our analyses of a retrospective cohort of patients with SCD were conducted using data from the Truven Health MarketScan® Commercial Claims and Encounters and Medicare Supplemental and Coordinated Benefits databases between 1 January 2009 and 31 December 2014. The databases include health insurance claims and enrolment data for over 100 million covered individuals from large groups of participating U.S. employers and health plans across all 50 states and Washington D.C.(Adamson, et al 2008) The databases contain individual-level records of inpatient, outpatient and pharmacy services for enrolees and their dependents. The data were de-identified and compliant with the Health Insurance Portability and Accountability Act (HIPAA) and this study was determined to be exempt by the Institutional Review Board of the University of Illinois at Chicago.

Identification of patients with SCD

International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (See Supplemental Table 1) were used to identify patients with SCD. Eligible subjects were required to have at least one inpatient SCD-related diagnosis code (282.41–42, 282.6, 282.60–64, and 282.68–69) or two separate outpatient medical claims that were at least 30 days apart. Further, a validated algorithm (RuSH) to identify SCD using administrative claims was applied to minimize misclassification or selection biases.(Paulukonis, et al 2014) The algorithm correctly identified 81%−87% of all patients with SCD ages 19 years and older from Georgia and California population-based SCD surveillance data, the sensitivity, specificity, and positive predictive value were 87%, 79%, and 87% respectively.(Snyder, et al 2017) The codes used for this validated step are summarized in Supplemental Table 2.

In addition, all adult subjects (aged ≥ 20 years, assessed at cohort entry) were required to have ≥ 6 months of continuous enrolment after cohort entry. Patients who disenrolled within 6 months (180 days) from cohort entry and/or whose age was younger than 20 years old were excluded. For each year, a base cohort was created to include patients who were continuously enrolled for the whole calendar year and had at least one medical encounter. The latter criterion was applied to exclude patients who were enrolled in more than one health plan and to ensure that the majority of the medical and pharmacy encounters were captured by the databases.

Ascertainment of diabetes diagnoses

The T2DM cases were identified through administrative health plan claims with ICD-9-CM diagnosis codes of 250.x0 or 250.x2 from one inpatient discharge summary or two outpatient claims that are 30 days apart. In addition, patients with at least one claim of oral anti-diabetic medication identified by national drug code (NDC) were included as having T2DM. All oral anti-diabetic medications and drug classes are summarized in Supplemental Table 3. Patients with insulin as the only anti-diabetic medication without T2DM-related qualifying diagnosis codes were not considered to be prevalent or incident cases to minimize misclassification of type 1 diabetes. Similarly, patients with gestational diabetes (ICD-9 code 6480), or type 1 diabetes mellitus (ICD code 250.x1, 250.x3) at any time during the study period were excluded. The patient selection flow chart is shown in Supplemental Figure 1.

T2DM cases in the National Health and Nutrition Examination Survey (NHANES) data were identified through self-reported diagnosis by the participants or fasting plasma glucose (FPG) level equal or above 126 mg/dL (7.0 mmol/L) based on the American Diabetes Association (ADA)’s recommendation on diagnosis and classification of diabetes mellitus(Association 2010), or both. Self-reported diabetes was defined as the participant responded yes to the survey question “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?”

Ascertainment of incident and prevalent diabetes cases

Incident cases of diabetes were identified through ICD-9 diagnosis codes for inpatient and outpatient encounters and/or NDC codes for outpatient dispensed anti-diabetic medications. The date of T2DM occurrence was defined as the earlier of either the first date of a qualifying diabetes diagnosis code or oral anti-diabetic medication dispensing. Patients were considered prevalent diabetes cases when T2DM was identified over the first 6 months after cohort entry or during the previous years of continuous enrolment in the database.

Estimation of prevalence and incidence rates

Prevalence and incidence rates were estimated using the number of identified prevalent and incident cases in the respective base cohorts per individual year. The annual prevalence and incidence rates were determined per 1000 person-years and per 1000 population continuously enrolled insured individuals annually. We used direct age and sex standardization to the U.S. population at the 2010 Census to describe the prevalence and incidence rates of T2DM. Prevalence and incidence rates for T2DM were adjusted to account for age and gender differences between the reference population and our population of patients with SCD. Patients were categorized into 14 strata based on age (quinquennium) and sex.

We compared prevalence rates of T2DM in our cohort of patients with SCD to subjects self-identified as non-Hispanic Black in the NHANES survey data from three cycles: 2009–2010, 2011–2012 and 2013–2014. Crude rates and age- and sex-adjusted prevalence rates for T2DM were calculated in both the SCD cohort and from NHANES subjects.

Covariates

Demographic and clinical characteristics were described for the identified SCD subjects over the 6-month baseline period. Information on SCD-specific conditions included vaso-occlusive crisis and acute chest syndrome. Comorbidities associated with diabetes, including obesity, hypertension, dyslipidaemia and chronic kidney disease, were collected using respective ICD-9 codes (See Supplemental Table 4). Deyo-Charlson comorbidity index (CCI) scores were calculated to measure the severity of comorbid conditions using all the diagnoses collected over the 6-month baseline period.(Klabunde, et al 2000)

National Health and Nutrition Examination Survey as Comparator Group

Race is not an available individual-level characteristic in the Truven databases. Given that the vast majority of patients with SCD are African-American and SCD is a rare congenital hemoglobinopathy, NHANES was used to provide a comparison group for describing T2DM rates across demographically similar patients. NHANES is a cross-sectional survey conducted by the Centers for Disease Control and Prevention (CDC)’s National Center for Health Statistics. The survey collects demographic, socioeconomic, dietary, and health-related information in addition to medical, dental, physiological and laboratory measurements through interview and medical examination on a two-year cycle. Response rates for participation range from 71% to 80%.(CDC 2018) All adult participants (aged ≥20 years) self-identified as non-Hispanic black from the NHANES data between 2009 to 2014 cycles were included as a reference population for comparison of T2DM rates with the commercially insured cohort of SCD patients.

Statistical analyses

Differences in demographic and clinical characteristics were assessed between SCD subjects with and without prevalent T2DM using descriptive statistics. Specifically, student’s t-tests were used for continuous variables and chi-square tests for associations between categorical variables.

The analyses examined data from 2009–2014 to evaluate the time trends in T2DM prevalence and incidence rates. Age- and sex-adjusted prevalence and incidence rates were calculated using direct standardization from age and sex distribution of the 2010 U.S. Census population. Specifically, patients from the Truven data were coded into the same age groups used in the US Census and the prevalence and incidence rates within those age groups were weighted by the population percentages in those age cohorts from the US Census. 95% confidence limits of prevalence rates were estimated using the Agresti-Coull method assuming a binomial distribution.(Agresti and Coull 1998) Statistical significance of trends in the number of incident and prevalent patients with T2DM over the studied time period were assessed using a generalized linear model with a Poisson link, adjusting for age and gender. The PROC GENMOD procedure in SAS 9.4 was used for trend testing.

Demographic and clinical factors associated with T2DM incidence were examined using data collected from a 6-month baseline period after patients entered the cohort. Here, patients who met the eligible diabetes criteria within the baseline period (i.e. prevalent cases) were excluded and the censoring date was the end of study period (31 December, 2014) or end of enrollment, whichever came first. Cox proportional hazard models were then used to calculate hazard ratios (HR) and 95% confidence intervals (CI) for T2DM onset. Demographic and clinical covariates with p values below 0.1 and at least 1% prevalence in initial univariate analyses of the base cohort were included in the final model for risk factor assessment, along with age, gender, and CCI scores. The proportional hazards assumptions was evaluated by visual inspection of the log minus log survival curves, and from the interaction between variables and follow-up time.(Collet 2003) We found no evidence suggesting a violation of the proportionality assumption in these models.

All analyses were conducted using SAS 9.4 (Cary, North Carolina, U.S.) and a two-tailed p value of less than 0.05 was used to determine statistical significance.

Results

Study sample

A total of 7,070 patients with SCD aged ≥20 years at the time of cohort entry in the administrative databases from 2009 through 2014 met the selection criteria and were included in the final analysis (Supplemental Figure 1). Baseline descriptive demographic and clinical characteristics are summarized in Table 1. Among these patients, 908 patients were identified to have prevalent (n=476) and incident (n=432) T2DM. Mean [median] age was greater among SCD patients with diabetes compared to SCD patients without T2DM (51 [51] vs. 37 [35] years; p<0.001).

Table 1.

Baseline demographic characteristics of patients aged 20 years and older with sickle cell disease, 2009–2014

	Total N (%)		SCD with T2DM		SCD without T2DM		P-value
	N=7070		N=908		N=6162
Sex
Male	2770	39.2%	342	37.7%	2428	39.4%	0.317
Female	4300	60.8%	566	62.3%	3734	60.6%
Age (years)
Mean (Median)	38.9	37	51.0	51	37.1	35	<.001
20–29	2132	30.2%	60	6.6%	2072	33.6%	<.001
30–39	1868	26.4%	121	13.3%	1747	28.4%
40–49	1471	20.8%	222	24.4%	1249	20.3%
50–59	1018	14.4%	276	30.4%	742	12.0%
60–69	418	5.9%	154	17.0%	264	4.3%
70+	163	2.3%	75	8.3%	88	1.4%
Region
Northeast	1415	20.0%	183	20.2%	1232	20.0%	<.001
North Central	970	13.7%	163	18.0%	807	13.1%
South	3724	52.7%	423	46.6%	3301	53.6%
West	610	8.6%	95	10.5%	515	8.4%
Unknown	351	5.0%	44	4.8%	307	5.0%
Year of cohort entry
2009	3516	49.7%	551	60.7%	2965	48.1%	<.001
2010	978	13.8%	122	13.4%	856	13.9%
2011	961	13.6%	108	11.9%	853	13.8%
2012	545	7.7%	50	5.5%	495	8.0%
2013	710	10.0%	63	6.9%	647	10.5%
2014	360	5.1%	14	1.5%	346	5.6%

SCD, sickle cell disease; T2DM, type 2 diabetes mellitus.

Clinical characteristics at baseline

Over the first 6 months baseline period following cohort entry, more SCD patients with T2DM had comorbid hypertension (49.0% vs. 18.0%; p<0.001), dyslipidaemia (26.4% vs. 4.4%; p<0.001), and chronic renal diseases (5.9% vs. 2.6%; p<0.001), when compared with SCD patients without T2DM. On the other hand, patients with T2DM had lower rates of vaso-occlusive crisis (30.6% vs. 46.5%; p<0.001), acute chest syndrome (2.0% vs. 3.7%; p=0.074), and pneumonia (7.6% vs. 10.7%; p=0.004). In general, SCD patients with T2DM had greater comorbidity as measured by CCI score, 68.6% of T2DM patients with pre-existing SCD had CCI scores of 2 or greater, compared to 23.6% among SCD patients without T2DM (p<0.001) (Table 2)

Table 2.

Clinical characteristics of patients aged 20 years and older with sickle cell disease, 2009–2014

	Total N (%)		SCD with T2DM		SCD without T2DM		P-value
	N=7070		N=908		N=6162
Baseline clinical characteristics
Hypertension	1555	22.0%	445	49.0%	1110	18.0%	<.001
Dyslipidaemia	512	7.2%	240	26.4%	272	4.4%	<.001
COPD	855	12.1%	126	13.9%	729	11.8%	0.078
Impaired fasting glycose	423	6.0%	331	36.5%	92	1.5%	<.001
Malignancies	248	3.5%	72	7.9%	176	2.9%	<.001
Overweight/Obesity	268	3.8%	59	6.5%	209	3.4%	<.001
Baseline SCD-specific conditions
Vaso-occlusive crisis	3146	44.5%	278	30.6%	2868	46.5%	<.001
Avascular necrosis	405	5.7%	42	4.6%	363	5.9%	0.126
Iron overload	305	4.3%	37	4.1%	268	4.3%	0.704
Pulmonary hypertension	289	4.1%	41	4.5%	248	4.0%	0.486
Chronic renal disease	216	3.1%	53	5.8%	163	2.6%	<.001
Acute chest syndrome	248	3.5%	18	2.0%	230	3.7%	0.007
Pneumonia	727	10.3%	69	7.6%	658	10.7%	0.004
Stroke	133	1.9%	27	3.0%	106	1.7%	0.010
Splenomegaly	83	1.2%	7	0.8%	76	1.2%	0.227
Splenic sequestration	16	0.2%	4	0.4%	12	0.2%	0.146
Hypersplenism	8	0.1%	3	0.3%	5	0.1%	0.037
Leg ulcers	51	0.7%	6	0.7%	45	0.7%	0.817
Dactylitis	17	0.2%	2	0.2%	15	0.2%	0.894
Osteomyelitis	40	0.6%	8	0.9%	32	0.5%	0.175
Baseline Charlson Comorbidity Index
0	3589	50.8%	151	16.6%	3438	55.8%	<.001
1	1401	19.8%	134	14.8%	1267	20.6%
2+	2080	29.4%	623	68.6%	1457	23.6%
Clinical outcomes during observational period
Nephropathy	840	11.9%	254	28.0%	586	9.5%	<.001
Neuropathy	481	6.8%	161	17.7%	320	5.2%	<.001
Ophthalmic comorbidities	112	1.6%	80	8.8%	32	0.5%	<.001
Peripheral circulatory	440	6.2%	157	17.3%	283	4.6%	<.001
Foot Ulcer	319	4.5%	48	5.3%	271	4.4%	0.229
Amputation	6	0.1%	3	0.3%	3	0.0%	0.006
Myocardial Infarction	101	1.4%	34	3.7%	67	1.1%	<.001
Stroke	788	11.1%	219	24.1%	569	9.2%	<.001
Depression	998	14.1%	181	19.9%	817	13.3%	<.001
Coronary Artery Bypass Surgery	130	1.8%	40	4.4%	90	1.5%	<.001
Angioplasty	98	1.4%	35	3.9%	63	1.0%	<.001

SCD, sickle cell disease; T2DM, type 2 diabetes mellitus.

Clinical outcomes at observational period

Over the observational period, SCD patients with T2DM were observed to have significantly higher rates of chronic and acute complications. As shown in Table 2, patients with SCD and T2DM experienced higher rates of macroangiopathy and microangiopathy. Twenty-eight percent of SCD patients with T2DM had nephropathy, as compared to 9.5% among SCD patients without T2DM (p<0.001). Moreover, SCD patients with T2DM were more likely to have neuropathy (17.7% vs. 5.2%; p<0.001), ophthalmic comorbidities (8.8% vs. 0.5%; p<0.001), peripheral circulatory disorders (17.3% vs. 4.6%; p<0.001), myocardial infarction (3.7% vs. 1.1%; p<0.001) and stroke (24.1% vs. 9.2%; p<0.001).

Table 2 reports that SCD patients with T2DM had higher rates of coronary artery bypass surgery (CABG) (4.4% vs. 1.5%; p<0.001) and angioplasty (3.9% vs. 1.0%; p<0.001), when compared to SCD patients without T2DM. In subgroups of younger patients with SCD (20–44 years old at cohort entry) and older SCD patients (45 years old and above), T2DM was consistently associated with higher occurrence of comorbidities and procedures. (Supplemental Figure 5).

National-level prevalence rate of T2DM, 2009–2014

As shown in Table 3, annual incidence and prevalence rates of T2DM among patients with SCD demonstrated an increasing trend across the study period. The unadjusted prevalence rates of T2DM in the U.S. commercially insured SCD population had a trend increasing from 9.8% (95% CI 8.8%−10.9%) in 2009 to 11.8% (95% CI 10.6%−13.0%) in 2014, representing a 0.2%−0.5% year-to-year change (p = 0.066) (Figure 1B). The age-and sex-standardized prevalence increased from 15.7% in 2009 to 16.5% in 2014, indicating a statistically significant increasing trend in population-level prevalence of T2DM, (p trend = 0.026). (Table 3) There was an increasing trend towards greater prevalence rates of T2DM in patient groups with advanced age. (Figure 1A)

Table 3.

Trends in proportion of type 2 diabetes mellitus among patients with sickle cell disease, 2009–2014

	2009	2010	2011	2012	2013	2014	P trend*
Truven MarketScan® SCD cohort
Number of eligible SCD subjects in base cohort each year	3066	3328	3567	3430	3164	2856
Number of newly onset T2DM cases	49	73	92	77	60	49
Number of all T2DM cases (pre-existing and newly onset cases)	300	337	377	404	368	337
Crude prevalence rates (95% CI**)	9.8% (8.8%−10.9%)	10.1% (9.2%−11.2%)	10.6% (9.6%−11.6%)	11.8% (10.7%−12.9%)	11.6% (10.6%−12.8%)	11.8% (10.7%−13.0%)	0.066
Age-and sex-adjusted prevalence rates ***	15.7%	15.6%	14.9%	16.2%	16.6%	16.5%	0.026
NHANES cohort (Non-Hispanic black subjects)
Number at risk	1122		1455		1177
Prevalent cases	199		278		207
Crude prevalence rates (95% CI**)	17.7% (15.6%−20.1%)		19.1% (17.2%−21.2%)		17.6% (15.5%−19.9%)		0.957
Age-and sex-adjusted prevalence rates ***	14.9%		16.4%		15.1%		0.702
Crude rate differences	7.9%	7.6%	8.5%	7.3%	6.0%	5.8%
Adjusted rate differences	−0.8%	−0.7%	1.5%	0.2%	−1.5%	−1.4%
Concomitant dyslipidaemia and/or hypertension in the Truven MarketScan® SCD cohort
Diabetes with dyslipidaemia (N, %)	143 (4.7%)	155 (4.7%)	174 (4.9%)	188 (5.5%)	166 (5.2%)	172 (6.0%)	0.149
Diabetes with hypertension (N, %)	202 (6.6%)	221 (6.6%)	240 (6.7%)	255 (7.4%)	226 (7.1%)	218 (7.6%)	0.702
Diabetes with both dyslipidaemia and hypertension (N, %)	111 (3.6%)	123 (3.7%)	137 (3.8%)	149 (4.3%)	128 (4.0%)	140 (4.9%)	0.174
Diabetes with either hypertension or dyslipidaemia (N, %)	234 (7.6%)	253 (7.6%)	277 (7.8%)	294 (8.6%)	264 (8.3%)	250 (8.8%)	0.593

SCD, sickle cell disease; T2DM, type 2 diabetes mellitus.

^*P _trend was obtained with Poisson regression models, adjusted for age and sex

^**95% confidence interval calculated using the Agresti-Coull method assuming binomial distribution

^***Direct standardization using the 2010 U.S. Census data

Figure 1.

Proportion of type 2 diabetes mellitus at different age groups. (A) Trends in proportion of patients with type 2 diabetes mellitus among all eligible SCD subjects of the year, (B) Trends in proportion of SCD patients with type 2 diabetes mellitus compared with non-Hispanic Black subjects from the NHANES survey, (C) Proportion of type 2 diabetes mellitus at different age groups, stratified by cohort and sex.

SCD, sickle cell disease; NHANES, National Health and Nutrition Examination Survey.

NHANES subjects who identified as non-Hispanic Black had crude prevalence rates of T2DM ranging from 17.6% to 19.1% (p trend = 0.957). The overall prevalence rate of T2DM was 18.2% between 2009 and 2014. After age-and sex- standardization, the adjusted T2DM prevalence rates among adult non-Hispanic Black NHANES subjects were similar and ranged from 17.6% to 19.1% between 2009 and 2014 (p trend = 0.702). The patients with SCD had consistently similar rates of T2DM compare to the general black population enrolled in NHANES surveys across age groups. (Figure 1C)

Age-adjusted T2DM incidence

Over a total of 17,024 person-years (PYs) in the SCD patient population, 432 patients had new-onset T2DM such that the overall crude incidence rate was 25.4 per 1000 person-years (Supplemental Table 6). Wide variations in incidence rates were seen across age and sex groups. As shown in Figure 2A, incidence rates increased to a peak of 53.5 per 1000 PYs in men aged 45–49 years. The peak incidence of T2DM observed in women was for 60–64 year old at a rate of 60.1 per 1000 PYs. Five-year cumulative incidence rates of 11.4% and 12.4% were observed in male and female patients with SCD respectively. As illustrated in Figure 2B and 2C, risk of developing T2DM in male and female patients with SCD increased with age.

Figure 2.

Age and sex stratified incidence rates of type 2 diabetes mellitus among SCD patients, 2009–2014. (A) Incidence rates of type 2 diabetes mellitus at different age groups, stratified by sex, (B) Cumulative incidence of type 2 diabetes mellitus by age group in men with SCD, (C) Cumulative incidence of type 2 diabetes mellitus by age group in women with SCD.

SCD, sickle cell disease.

Co-occurrence of T2DM with hypertension and/or dyslipidaemia in patients with SCD

In multivariate Cox models, HRs and 95% CI were estimated for new-onset T2DM (Figure 3A). Hypertension (HR=1.45, 95%CI 1.14–1.83; p = 0.002) and dyslipidaemia (HR=1.43, 95%CI 1.04–1.96; p = 0.027) at baseline were positively associated with risk of T2DM. However, T2DM incidence was not associated with SCD-related complications, such as vaso-occlusive crisis, acute chest syndrome and stroke.

Figure 3.

Correlation between type 2 diabetes mellitus and other relevant risk factors. (A) Baseline clinical characteristics and SCD-related conditions associated with newly onset type 2 diabetes mellitus cases using Cox proportional hazards models, adjusted for age, sex and Charlson Comorbidity Index score, (B) Proportion of patients having type 2 diabetes mellitus among SCD patients with hypertension and/or dyslipidemia at different age groups.

As shown in Figure 3B, dyslipidaemia and hypertension strongly correlated with T2DM. A large proportion of SCD patients had concomitant hypertension, dyslipidaemia, or both, and rates were substantially higher in those with T2DM. Annual prevalence rates of comorbid hypertension and dyslipidaemia between 2009 and 2014 were 6.6% to 7.6% and 4.7% to 6.0% respectively in patients with SCD. Among patients with prevalent T2DM, over 72% had hypertension and/or dyslipidaemia and there was no significant change in comorbid hypertension and dyslipidaemia between 2009 and 2014 (Table 3).

Discussion

To our knowledge, this is the first epidemiologic study assessing the prevalence, incidence and risk factors for diabetes in a large, commercially insured U.S. population of patients with SCD. Contrary to prior studies reporting very low diabetes prevalence using historical institutional cohorts of patients with SCD, our results indicate that the prevalence of T2DM increased modestly between 2009 and 2014, approaching the prevalence observed in a general comparison population of African Americans in the U.S. Moreover, older age, hypertension and dyslipidaemia were positively associated and very frequently co-occurred with T2DM among patients with SCD.

Recent retrospective cohort studies suggest that lower rates of obesity in patients with SCD result in an apparent protective association to diabetes, although these trends differ by age group.(Mohamed, et al 2015, Zhang, et al 2015) One explanation is that high mortality rates among younger patients with SCD lead to a depletion of susceptible patients that survive until the age of typical clinical manifestation of diabetes.(Andersen, et al 2012, Fung, et al 2006, MORRISON, et al 1979) With advances in SCD care over the last two decades increasing the life expectancy in this population, it is expected that more patients with SCD will develop chronic metabolic conditions such as T2DM, hypertension and dyslipidaemia.(Lanzkron, et al 2013) Our analysis showed increasing annual prevalence of T2DM over the study period, a trend consistent with rates reported by NHANES.(Menke, et al 2015) The high co-occurrence of comorbid dyslipidaemia and hypertension in SCD patients with T2DM deserves further clinical investigation. These findings are consistent with other estimates in patients without SCD.(Menke, et al 2015) These metabolic conditions share many risk factors with T2DM including sedentary lifestyle, central obesity and poor diet. Multiple co-existing comorbid conditions complicate the treatment of T2DM through drug-drug interactions and could further increase the risks for diabetes-related complications like heart diseases and stroke. Therefore, addressing T2DM in aging patients with SCD also requires comprehensive prevention and management of chronic metabolic conditions.

Determining presence of chronic hyperglycaemia and diagnosis of T2DM in patients with SCD is difficult. Using elevated glycosylated haemoglobin (HbA_1c) as an indication of hyperglycaemia can result in false positives among patients with haemoglobin C (HbC) trait and clinically misinterpreted findings among patients with hemoglobinopathies such as sickle cell trait (SCT) thalassemia and SCD.(Reid, et al 1992, Rhea and Molinaro 2014, Schnedl, et al 1999) Due to a shortened lifespan of red blood cells in patients with SCD and SCT, red blood cells are removed at higher rates from circulation, resulting in lower levels of HbA_1c not representative of a patient’s glycaemic control.(Smiley, et al 2008) No consensus exists for diagnosing T2DM and monitoring long-term glycaemic control among patients with SCD. Fructosamine tests, which measure advanced glycosylated end products other than haemoglobin in the blood, are considered a reliable method due to this measure’s correlation with both plasma glucose and the HbA_1c in patients without SCD.₍Panzer, et al 1982₎ However, a short half-life, lack of standardization, and impracticality in outpatient settings are limitations to the wider adoption of fructosamine tests for screening of diabetes in patients with SCD.(Kosecki, et al 2005)

The understanding of body metabolism in older SCD patients remains unclear. SCD is associated with low lean body weight and fat mass in anthropomorphic studies.(Soliman, et al 1999, Zemel, et al 2007) The effects of chronic inflammation secondary to SCD in addition to oxidative stress result in a hypermetabolic state among patients with SCD.(Singhal, et al 1997) Therefore, it is possible that patients in our cohort who lived into adulthood represent mild or less severe forms of SCD, whereas the protective effects from mutant haemoglobin can be hampered. Given the increasing life expectancy among patients with SCD in recent decades, obesity is not uncommon in patients with SCD. In our study, we found that only 4% of patients had diagnosis codes indicating overweight or obesity. Local data for our own adult SCD patients at UIC indicate prevalence of overweight and obesity to be 6% and 25% in paediatric and adult SCD cases, respectively.(Akingbola, et al 2014) For all adults in the US, nearly 2 of 3 is either overweight or obese.(Disease 2017) Prior reports of a low prevalence of T2DM in SCD coupled with possible false negative screening with HbA_1c likely result in a general underestimation of the burden of T2DM. Current evidenced-based SCD management guidelines recommend diabetes screening only for patients with renal nephropathy and hypertension(NHLBI (National Heart 2014), our findings instead signal a need for T2DM screening, and patterns of comorbidity with hypertension and dyslipidaemia are important factors in the development of T2DM and its complications in patients with SCD above 45 years old.

Our research has important clinical and policy implications. Patients with T2DM can benefit from lifestyle changes, dietary counselling and interventions encouraging physical activity and exercise.(Umpierre, et al 2011) The apparent lower occurrence of T2DM with SCD diminishes as these patients age, and other metabolic risk factors including hypertension and dyslipidaemia are increasingly common. Racial differences in the burden of diabetes are due, to some considerable extent, to differences in socioeconomic status between racial and ethnic groups.(Dagogo-Jack 2003) While patients with SCD experience the burden of SCD management and its complications, many African Americans also face challenges from lower socioeconomic position and access to health care.(Farber, et al 1985) Clinicians treating patients with SCD should consider strategies to provide quality chronic condition care, including diabetes management, that also addresses the existing health disparities faced by African Americans.(Anderson, et al 2009, Telfair, et al 2003)

The present study should be interpreted in the context of limitations inherent to administrative claims data. First, we rely on claims information not collected for the purposes of research to identify diabetes and other conditions. To minimize potential misclassification, we employed validated algorithms in combination with pharmacy dispensing records for anti-diabetic medications. We also note that the most common oral medication, metformin, may also be used in patients without diabetes (e.g., polycystic ovarian syndrome). We further used NHANES as an external comparison for our estimates and identify trends in diabetes relative to the general U.S. African American population. Second, our study population includes those with continuous commercial health insurance coverage and may not be representative of SCD patients that are publicly insured or who lack any health coverage. The disease burden of diabetes in SCD patients reported here is likely a conservative estimate. Privately insured patients may have mild or less severe forms of SCD enabling them to remain in the workforce with employer-based health coverage. It is important to assess the prevalence and incidence of T2DM in patients with public insurance such as Medicaid and those without insurance to fully understand the epidemiology of diabetes and its impact on patients with SCD. Third, we lacked information on laboratory test results and other important clinical characteristics known to be associated with T2DM. These include body weight, adiposity, BMI and SCD-related genetic factors which could influence our estimates of the burden of diabetes in the general population with SCD. Similarly, lacking measurement of body weight and height, our measurement of the prevalence of obesity and overweight from diagnosis codes used for the purposes of provider reimbursement is likely underestimated.

In conclusion, we report here on the epidemiology of T2DM in patients with SCD with trends toward increasing prevalence over time. The high proportion of T2DM among older patients with comorbid conditions such as hypertension and/or dyslipidaemia signals for routine T2DM screening and opportunities to prevent T2DM and its complications in patients with SCD.