Optimal glycemic control is the aim of diabetes care. Clinical trials in both type 1 and type 2 diabetes have established the link between glycemic control and the risk for development of diabetes complications (1,2). The SEARCH for Diabetes in Youth (SEARCH) study first reported its cross-sectional analysis evaluating glycemic control in youth with diabetes in 2009, at a time when limited data were available on glycemic control in large populations in the U.S. (3). This initial work highlighted that a substantial percentage of youth with type 1 and type 2 diabetes had poor glycemic control (HbA1c ≥9.5%). SEARCH also helped shed light on disparities in glycemic outcomes in youth, as racial/ethnic minorities are more likely to have higher HbA1c levels compared with non-Hispanic White youth irrespective of the type of diabetes (4).

Since the 2009 SEARCH publication, the landscape of diabetes management has changed dramatically. Most notably, diabetes technology has rapidly evolved, with new technologies being developed and improved every year. The use of continuous subcutaneous insulin infusion and continuous glucose monitoring (CGM) systems in the U.S. has increased, especially among youth with type 1 diabetes (5,6). In addition, to reflect new evidence regarding the risks and benefits of tight glycemic control in children and adolescents with diabetes, recent national and international recommendations endorse lower HbA1c targets, resulting in providers prescribing more intensive diabetes management for all pediatric age groups (7,8).

The T1D Exchange Registry and the Pediatric Diabetes Consortium are two large registry studies in the U.S. that have reported on glycemic control among youth with type 1 diabetes and type 2 diabetes, respectively (9,10). While both have helped to describe the recent state of treatment of youth with diabetes in the U.S., the T1D Exchange Registry and the Pediatric Diabetes Consortium have only been able to comment on glycemic trends over the past decade. Beginning in 2002, SEARCH has recruited a series of incident racially/ethnically and socioeconomically diverse youth cohorts with both type 1 and type 2 diabetes who are well-characterized through a variety of surveys and physical and laboratory assessments soon after diagnosis and have been followed longitudinally. Given the availability of population-based SEARCH longitudinal data related to glycemic control to evaluate the impact of the changing landscape of diabetes management, the study objective was to describe temporal trends in glycemic control by age and diabetes duration in youth-onset diabetes, beginning 1 year after diagnosis. In addition, we sought to identify correlates of glycemic control among youth with type 1 and type 2 diabetes in the 2014–2019 SEARCH cohort of youth and young adults (YYA) to identify groups of patients who may benefit from targeted interventions to improve metabolic control.

SEARCH is a population-based registry network that includes five centers located in California, Colorado, Ohio, South Carolina, and Washington. Children and adolescents with diabetes diagnosed before 20 years of age were identified from ongoing surveillance of networks. Comprehensive details pertaining to the recruitment and study visit components of the SEARCH study have previously been published (11). In the first two phases of SEARCH (SEARCH 1 and 2), individuals newly diagnosed with diabetes in 2002–2006 and 2008 were contacted and recruited for a baseline research visit. Incident cases from 2002–2005 were also asked to return for visits at 12, 24, and 60 months after their baseline visit to measure risk factors for diabetes complications. Also, in the first phase of the study (SEARCH 1), prevalent cases of diabetes, diagnosed prior to 2002, were invited for a single visit. In the third phase (SEARCH 3), a subset of SEARCH participants with a duration of diabetes >5 years were recruited for an outcome visit between 2011 and 2015, for whom a single assessment of diabetes-related data collection was completed. In the fourth phase (SEARCH 4), all SEARCH participants aged >10 years were operationally split into a group invited to another study visit between 2015 and 2019 and those who were only invited to complete surveys. Those invited to the in-person research visit included all individuals with type 2 diabetes, all non-Whites, and a random sample of non-Hispanic Whites with type 1 diabetes. Since SEARCH is a population-based study, the study site that recruited the participant was often not the clinical location where participants received their diabetes care.

Research visits included questionnaire administration along with collection of anthropometric measurements and a blood sample. HbA1c levels were measured from blood samples obtained at a research study visit. Measurement of HbA1c was performed at the Northwest Lipid Metabolism and Diabetes Research Laboratories, University of Washington, which was the central laboratory for the study. Measurements were performed using an automated nonporous ion exchange high-performance liquid chromatography system (Tosoh Bioscience, Montgomeryville, PA). This method has demonstrated to be linear from a total area of 500 to >4,500, indicating that the results are accurate within a large range of number of red cells. If the total area is <500, results are not reported; if the total area is >4,500, the analysis is repeated after sample dilution. A set of quality control samples with low and high levels of HbA1c was analyzed several times per day to monitor the assay performance, and the between-assay coefficients of variation on the two controls were consistently <1.0% and <0.7%, respectively. Aliquots of six whole blood pools prepared in the laboratory with values of 5.0%, 6.0%, 7.0%, 7.5%, 8.0%, and 9.0% stored under liquid nitrogen and analyzed each month for several days to monitor the longitudinal stability of the assay.

SEARCH participants with a diabetes duration >1 year at a study visit were included in the study sample. Diabetes type was based on provider diagnosis within 6 months of diagnosis. Participants with a type 1 diabetes provider diagnosis who were not on insulin were excluded. For all participants, the parent, adolescent or young adult, or both provided consent or assent. The institutional review boards for all sites approved the study protocol.

SEARCH study visits were conducted from 2002 through 2019. All participants with an eligible visit are included in the analytic data set at least once. Participants could have had up to six visits over that time period. A sample of visits for this analysis was randomly selected from the SEARCH data in such a way as to prioritize inclusion in diabetes duration groups (1–4 years, 5–9 years, and ≥10 years) and time periods (2002–2007, 2008–2013, and 2014–2019; the 2002–2007 cohort represents the cohort reported in the initial 2009 SEARCH publication) while ensuring that no participant would have multiple visits within a duration group or time period (Supplementary Fig. 1). By doing this, independence assumptions of statistical methods and analysis stratified by diabetes duration or time period would not be violated due to multiple records per participant. A sensitivity analysis was conducted to verify that the analytic data set was consistent with other randomly drawn samples using the same criteria.

All analyses were stratified by diabetes type. Participant characteristics were summarized using frequencies and percentages for categorical variables or means and SDs for continuous measures. Unadjusted linear regression models stratified by duration group were used to evaluate differences in HbA1c across time periods. Multivariable linear regression models, stratified by duration group, were then used to test differences in HbA1c over time after adjustment for clinical site, age, sex, race/ethnicity, household income, health insurance status, insulin regimen, and disease duration. Repeated-measures linear models were not stratified by diabetes duration (“overall” in tables) to account for the multiple visits (over the three time periods) per participant. Plots were created to investigate the association of age group and time period with HbA1c using results of the fully adjusted stratified multivariable linear models. Multivariable linear models, stratified by diabetes type and adjusted for all covariates, were used to investigate the associations of HbA1c with participant characteristics during the last time period (2014–2019). All analyses were completed using SAS version 9.4 (SAS Institute, Cary, NC).

Longitudinal analysis included 6,369 (n = 5,482 type 1 and n = 887 type 2) SEARCH participants. Descriptive characteristics of the study sample are shown in Table 1. While the sex of participants across the three time periods was fairly similar irrespective of diabetes type, the 2014–2019 sample of YYA with type 1 diabetes is comprised of a higher percentage of non-Hispanic Black and Hispanic participants as compared with earlier cohorts due to the SEARCH 4 sampling strategy. Age and average duration of diabetes increased among participants with both type 1 and type 2 diabetes across successive SEARCH cohorts. There was also an increase in insulin pump use and CGM use in the more recent cohorts of participants with type 1 diabetes.

Table 1

Characteristics of YYA who participated in SEARCH study visits by study period and diabetes type

. Type 1 . Type 2 . 2002–2007 (N = 3,398) . 2008–2013 (N = 2,184) . 2014–2019 (N = 1,742) . P value . 2002–2007 (N = 379) . 2008–2013 (N = 327) . 2014–2019 (N = 519) . P value . HbA1c (%) 8.5 (1.5) 8.9 (1.8) 9.1 (2.0) <0.0001 8.4 (2.8) 8.3 (2.8) 8.9 (2.9) 0.0034 Diabetes duration, years (SD) 4.6 (3.6) 5.8 (2.8) 8.3 (4.4) <0.0001 2.9 (1.6) 5.3 (3.1) 7.0 (4.5) <0.0001 Insulin regimen (%)    <0.0001    0.0002  Not on insulin 0 0 0  222 (58.6) 176 (53.8) 275 (53.0)   Insulin pump 845 (24.9) 976 (44.7) 856 (49.1)  12 (3.2) 25 (7.6) 31 (6.0)   Basal-bolus injections 1,000 (29.4) 880 (40.3) 781 (44.8)  44 (11.6) 66 (20.2) 104 (20.0)   Other insulin regimens 1,544 (45.4) 261 (12.0) 77 (4.4)  95 (25.1) 51 (15.6) 92 (17.7)   Unknown 9 (0.3) 67 (3.1) 28 (1.6)  6 (1.6) 8 (2.4) 17 (3.3)  Medication regimen (%)    0.3678    <0.0001  Insulin only 3,310 (97.4) 2,120 (97.1) 1,685 (96.7)  63 (16.6) 61 (18.7) 119 (22.9)   Insulin plus oral agent 88 (2.6) 64 (2.9) 57 (3.3)  94 (24.8) 90 (27.5) 122 (23.5)   Metformin only 0 0 0  120 (31.7) 98 (30.0) 125 (24.1)   Other oral agent 0 0 0  58 (15.3) 27 (8.3) 26 (5.0)   None 0 0 0  44 (11.6) 51 (15.6) 117 (22.5)   Unknown 0 0 0  0 0 10 (1.9)  Blood glucose monitoring frequency (%)    <0.0001    <0.0001  Less than once a day 106 (3.1) 496 (22.7) 276 (15.8)  122 (32.2) 52 (15.9) 132 (25.4)   1–3 times/day 646 (19.0) 474 (21.7) 304 (17.5)  169 (44.6) 95 (29.1) 95 (18.3)   ≥4 times/day 2,506 (73.7) 674 (30.9) 341 (19.6)  64 (16.9) 63 (19.3) 55 (10.6)   CGM* 0 165 (7.6) 410 (23.5)  0 28 (8.6) 44 (8.5)   Unknown 140 (4.1) 375 (17.2) 411 (23.6)  24 (6.3) 89 (27.2) 193 (37.2)  Sex (%)    0.5429    0.5918  Female 1,680 (49.4) 1,073 (49.1) 885 (50.8)  235 (62.0) 205 (62.7) 338 (65.1)   Male 1,718 (50.6) 1,111 (50.9) 857 (49.2)  144 (38.0) 122 (37.3) 181 (34.9)  Race/ethnicity (%)    <0.0001    0.0058  Non-Hispanic White 2,622 (77.2) 1,649 (75.5) 1,028 (59.0)  70 (18.5) 60 (18.3) 95 (18.3)   Non-Hispanic Black 240 (7.1) 180 (8.2) 224 (12.9)  145 (38.3) 128 (39.1) 216 (41.6)   Hispanic 369 (10.9) 245 (11.2) 363 (20.8)  88 (23.2) 86 (26.3) 145 (27.9)   Native American 19 (0.6) 10 (0.5) 10 (0.6)  57 (15.0) 29 (8.9) 33 (6.4)   Other/unknown/multiple 148 (4.4) 100 (4.6) 117 (6.7)  19 (5.0) 24 (7.3) 30 (5.8)  Age, years (SD) 12.9 (4.4) 15.1 (5.0) 18.0 (5.7) <0.0001 17.1 (2.8) 19.5 (4.3) 21.4 (5.0) <0.0001 BMI z-score (SD) 0.6 (0.9) 0.6 (0.9) 0.6 (1.0) 0.0758 1.9 (0.8) 1.9 (0.7) 1.9 (0.8) 0.5024 Household income (%)    <0.0001    0.0002  <$25,000 423 (12.4) 312 (14.3) 278 (16.0)  145 (38.3) 132 (40.4) 175 (33.7)   $25,000–49,999 695 (20.5) 357 (16.3) 306 (17.6)  83 (21.9) 54 (16.5) 92 (17.7)   $50,000–74,999 695 (20.5) 362 (16.6) 231 (13.3)  42 (11.1) 28 (8.6) 33 (6.4)   >$75,000 1,335 (39.3) 863 (39.5) 555 (31.9)  31 (8.2) 30 (9.2) 38 (7.3)   Do not know/refused 250 (7.4) 290 (13.3) 372 (21.4)  78 (20.6) 83 (25.4) 181 (34.9)  Health insurance (%)    <0.0001    0.0002  None 43 (1.3) 60 (2.7) 49 (2.8)  18 (4.7) 47 (14.4) 53 (10.2)   Other 54 (1.6) 70 (3.2) 88 (5.1)  32 (8.4) 21 (6.4) 37 (7.1)   Medicaid/Medicare 568 (16.7) 431 (19.7) 425 (24.4)  144 (38.0) 131 (40.1) 229 (44.1)   Private 2,733 (80.4) 1,623 (74.3) 1,180 (67.7)  185 (48.8) 128 (39.1) 200 (38.5)   . Type 1 . Type 2 . 2002–2007 (N = 3,398) . 2008–2013 (N = 2,184) . 2014–2019 (N = 1,742) . P value . 2002–2007 (N = 379) . 2008–2013 (N = 327) . 2014–2019 (N = 519) . P value . HbA1c (%) 8.5 (1.5) 8.9 (1.8) 9.1 (2.0) <0.0001 8.4 (2.8) 8.3 (2.8) 8.9 (2.9) 0.0034 Diabetes duration, years (SD) 4.6 (3.6) 5.8 (2.8) 8.3 (4.4) <0.0001 2.9 (1.6) 5.3 (3.1) 7.0 (4.5) <0.0001 Insulin regimen (%)    <0.0001    0.0002  Not on insulin 0 0 0  222 (58.6) 176 (53.8) 275 (53.0)   Insulin pump 845 (24.9) 976 (44.7) 856 (49.1)  12 (3.2) 25 (7.6) 31 (6.0)   Basal-bolus injections 1,000 (29.4) 880 (40.3) 781 (44.8)  44 (11.6) 66 (20.2) 104 (20.0)   Other insulin regimens 1,544 (45.4) 261 (12.0) 77 (4.4)  95 (25.1) 51 (15.6) 92 (17.7)   Unknown 9 (0.3) 67 (3.1) 28 (1.6)  6 (1.6) 8 (2.4) 17 (3.3)  Medication regimen (%)    0.3678    <0.0001  Insulin only 3,310 (97.4) 2,120 (97.1) 1,685 (96.7)  63 (16.6) 61 (18.7) 119 (22.9)   Insulin plus oral agent 88 (2.6) 64 (2.9) 57 (3.3)  94 (24.8) 90 (27.5) 122 (23.5)   Metformin only 0 0 0  120 (31.7) 98 (30.0) 125 (24.1)   Other oral agent 0 0 0  58 (15.3) 27 (8.3) 26 (5.0)   None 0 0 0  44 (11.6) 51 (15.6) 117 (22.5)   Unknown 0 0 0  0 0 10 (1.9)  Blood glucose monitoring frequency (%)    <0.0001    <0.0001  Less than once a day 106 (3.1) 496 (22.7) 276 (15.8)  122 (32.2) 52 (15.9) 132 (25.4)   1–3 times/day 646 (19.0) 474 (21.7) 304 (17.5)  169 (44.6) 95 (29.1) 95 (18.3)   ≥4 times/day 2,506 (73.7) 674 (30.9) 341 (19.6)  64 (16.9) 63 (19.3) 55 (10.6)   CGM* 0 165 (7.6) 410 (23.5)  0 28 (8.6) 44 (8.5)   Unknown 140 (4.1) 375 (17.2) 411 (23.6)  24 (6.3) 89 (27.2) 193 (37.2)  Sex (%)    0.5429    0.5918  Female 1,680 (49.4) 1,073 (49.1) 885 (50.8)  235 (62.0) 205 (62.7) 338 (65.1)   Male 1,718 (50.6) 1,111 (50.9) 857 (49.2)  144 (38.0) 122 (37.3) 181 (34.9)  Race/ethnicity (%)    <0.0001    0.0058  Non-Hispanic White 2,622 (77.2) 1,649 (75.5) 1,028 (59.0)  70 (18.5) 60 (18.3) 95 (18.3)   Non-Hispanic Black 240 (7.1) 180 (8.2) 224 (12.9)  145 (38.3) 128 (39.1) 216 (41.6)   Hispanic 369 (10.9) 245 (11.2) 363 (20.8)  88 (23.2) 86 (26.3) 145 (27.9)   Native American 19 (0.6) 10 (0.5) 10 (0.6)  57 (15.0) 29 (8.9) 33 (6.4)   Other/unknown/multiple 148 (4.4) 100 (4.6) 117 (6.7)  19 (5.0) 24 (7.3) 30 (5.8)  Age, years (SD) 12.9 (4.4) 15.1 (5.0) 18.0 (5.7) <0.0001 17.1 (2.8) 19.5 (4.3) 21.4 (5.0) <0.0001 BMI z-score (SD) 0.6 (0.9) 0.6 (0.9) 0.6 (1.0) 0.0758 1.9 (0.8) 1.9 (0.7) 1.9 (0.8) 0.5024 Household income (%)    <0.0001    0.0002  <$25,000 423 (12.4) 312 (14.3) 278 (16.0)  145 (38.3) 132 (40.4) 175 (33.7)   $25,000–49,999 695 (20.5) 357 (16.3) 306 (17.6)  83 (21.9) 54 (16.5) 92 (17.7)   $50,000–74,999 695 (20.5) 362 (16.6) 231 (13.3)  42 (11.1) 28 (8.6) 33 (6.4)   >$75,000 1,335 (39.3) 863 (39.5) 555 (31.9)  31 (8.2) 30 (9.2) 38 (7.3)   Do not know/refused 250 (7.4) 290 (13.3) 372 (21.4)  78 (20.6) 83 (25.4) 181 (34.9)  Health insurance (%)    <0.0001    0.0002  None 43 (1.3) 60 (2.7) 49 (2.8)  18 (4.7) 47 (14.4) 53 (10.2)   Other 54 (1.6) 70 (3.2) 88 (5.1)  32 (8.4) 21 (6.4) 37 (7.1)   Medicaid/Medicare 568 (16.7) 431 (19.7) 425 (24.4)  144 (38.0) 131 (40.1) 229 (44.1)   Private 2,733 (80.4) 1,623 (74.3) 1,180 (67.7)  185 (48.8) 128 (39.1) 200 (38.5)  

The estimated average adjusted HbA1c for the 2014–2019 cohort of YYA with type 1 diabetes was 8.8 ± 0.04% (72 mmol/mol) (Table 2). There was a statistically significant difference in HbA1c among the three cohorts, with the average adjusted HbA1c for the 2014–2019 cohort with type 1 diabetes being 0.3% higher than the mean HbA1c for the 2002–2007 cohort (8.5 ± 0.03% [70 mmol/mol]). When examined by diabetes duration, YYA with type 1 diabetes and a diabetes duration of 5–9 years exhibited a temporal trend of worse glycemic control in recent years (2002–2007: 8.6% [70 mmol/mol] vs. 2008–2013: 9.1% [76 mmol/mol] vs. 2014–2019: 9.2% [77 mmol/mol]). There was also a statistically significant increase in HbA1c among the 10–14-, 15–19-, and 20–24-year-old age groups of YYA with type 1 diabetes when comparing mean HbA1c in 2014–2019 to 2002–2007 (Fig. 1A and Supplementary Table 1). In 2014–2019, SEARCH participants had relatively comparable glycemic control compared with similarly aged YYA with type 1 diabetes in 2008–2013, except for the 20–24-year-old age group, which had a statistically significant lower adjusted HbA1c (8.7% [72 mmol/mol]) in 2014–2019 compared with 2008–2013 (8.9% [74 mmol/mol]).

Figure 1

Model adjusted for age, clinical site, disease duration, health insurance status, household income, insulin regimen, race/ethnicity, and sex. Mean HbA1c by age group across study periods.

Figure 1

Model adjusted for age, clinical site, disease duration, health insurance status, household income, insulin regimen, race/ethnicity, and sex. Mean HbA1c by age group across study periods.

Table 2

Model adjusted* mean HbA1c stratified by diabetes duration

. HbA1c (%) . P value . 2002–2007 . 2008–2013 . 2014–2019 . Type 1      Overall 3,398: 8.5 (0.03) 2,184: 8.9 (0.03) 1,742: 8.8 (0.04) <0.0001  1–4 years 2,288: 8.5 (0.04) 684: 8.6 (0.06) 380: 8.7 (0.09) 0.0444  5–9 years 765: 8.6 (0.07) 1,320: 9.1 (0.05) 670: 9.2 (0.07) <0.0001  ≥10 years 345: 8.6 (0.13) 180: 9.3 (0.13) 692: 9.0 (0.08) 0.0005 Type 2      Overall 379: 8.7 (0.14) 327: 8.3 (0.13) 519: 8.6 (0.12) 0.0330  1–4 years 336: 8.4 (0.14) 148: 7.9 (0.19) 190: 8.2 (0.19) 0.1349  5–9 years 43: 9.6 (0.44) 154: 8.8 (0.21) 167: 9.2 (0.20) 0.2600  ≥10 years — 25: 8.4 (0.60) 162: 10.1 (0.35) 0.0140  . HbA1c (%) . P value . 2002–2007 . 2008–2013 . 2014–2019 . Type 1      Overall 3,398: 8.5 (0.03) 2,184: 8.9 (0.03) 1,742: 8.8 (0.04) <0.0001  1–4 years 2,288: 8.5 (0.04) 684: 8.6 (0.06) 380: 8.7 (0.09) 0.0444  5–9 years 765: 8.6 (0.07) 1,320: 9.1 (0.05) 670: 9.2 (0.07) <0.0001  ≥10 years 345: 8.6 (0.13) 180: 9.3 (0.13) 692: 9.0 (0.08) 0.0005 Type 2      Overall 379: 8.7 (0.14) 327: 8.3 (0.13) 519: 8.6 (0.12) 0.0330  1–4 years 336: 8.4 (0.14) 148: 7.9 (0.19) 190: 8.2 (0.19) 0.1349  5–9 years 43: 9.6 (0.44) 154: 8.8 (0.21) 167: 9.2 (0.20) 0.2600  ≥10 years — 25: 8.4 (0.60) 162: 10.1 (0.35) 0.0140 

In the multivariate analysis of participants with type 1 diabetes from the 2014–2019 cohort, glycemic control (i.e., HbA1c) was significantly associated with race/ethnicity, age, BMI, insulin regimen, blood glucose monitoring frequency, and household income (Table 3). Non-Hispanic Black and Native American YYA with type 1 diabetes had higher HbA1c levels than non-Hispanic White participants (Supplementary Fig. 2). Other statistically significant correlates of poorer glycemic control in the multivariate model for T1D included younger age, lower BMI z-score, not using an insulin pump, infrequent self-monitoring of blood glucose, and lower household income.

Table 3

Associations of HbA1c with participant characteristics: SEARCH 2014–2019 cohort*

. Type 1 (n = 1,805 observations used) . Type 2 (n = 478 observations used) . Estimate . 95% CI . P value . Estimate . 95% CI . P value . Sex   0.07   0.44  Female Reference  Reference   Male −0.16 −0.33, 0.01  −0.18 −0.62, 0.27  Race/ethnicity   <0.0001   0.34  Non-Hispanic White Reference  Reference   Non-Hispanic Black 1.22 0.94, 1.51  0.4 −0.23, 1.03   Hispanic 0.19 −0.06, 0.44  0.45 −0.32, 1.22   Native American 1.21 0.09, 2.33  0.66 −0.50, 1.83   Other/unknown/multiple 0.08 −0.27, 0.43  1.06 −0.01, 2.14  Age (years) −0.06 −0.08, −0.04 <0.0001 0.02 −0.06, 0.10 0.58 Diabetes duration (years) 0.01 −0.03, 0.04 0.7545 0 −0.11, 0.11 0.94 BMI z-score −0.23 −0.32, −0.14 <0.0001 −0.43 −0.72, −0.14 0.004 Insulin regimen   <0.0001     Insulin pump Reference      Basal-bolus injections 0.42 0.23, 0.60      Other insulin regimens 0.83 0.39, 1.26      Unknown 0.5 −0.19, 1.19     Medication regimen      <0.0001  Insulin only    Reference   Insulin plus oral agent    −0.26 −0.89, 0.37   Metformin only    −3.09 −3.77, −2.42   Other oral agent    −1.91 −2.98, −0.83   None    −2.9 −3.57, −2.24   Unknown    −1.77 −3.35, −0.19  Blood glucose monitoring frequency   <0.0001   0.49  Less than once a day Reference  Reference   1–3 times/day 0.43 0.13, 0.73  −0.16 −0.83, 0.52   4 or more times/day −0.31 −0.60, −0.02  −0.47 −1.26, 0.31   CGM −0.48 −0.76, −0.20  −0.21 −1.07, 0.65   Unknown −1.13 −1.47, −0.80  −0.72 −1.57, 0.12  Household income   <0.0001   0.33  <$25,000 Reference  Reference   $25,000–49,999 −0.12 −0.41, 0.18  −0.19 −0.83, 0.46   $50,000–74,999 −0.55 −0.88, −0.22  −0.56 −1.52, 0.39   >$75,000 −0.72 −1.02, −0.42  −0.81 −1.76, 0.14   Do not know/refused −0.24 −0.53, 0.05  0.08 −0.44, 0.60  Health insurance   0.27   0.33  Private Reference  Reference   Medicaid/Medicare 0.22 −0.01, 0.45  −0.43 −0.96, 0.10   Other 0.12 −0.27, 0.52  0.16 −0.73, 1.06   None 0.28 −0.26, 0.82  −0.19 −0.96, 0.10   . Type 1 (n = 1,805 observations used) . Type 2 (n = 478 observations used) . Estimate . 95% CI . P value . Estimate . 95% CI . P value . Sex   0.07   0.44  Female Reference  Reference   Male −0.16 −0.33, 0.01  −0.18 −0.62, 0.27  Race/ethnicity   <0.0001   0.34  Non-Hispanic White Reference  Reference   Non-Hispanic Black 1.22 0.94, 1.51  0.4 −0.23, 1.03   Hispanic 0.19 −0.06, 0.44  0.45 −0.32, 1.22   Native American 1.21 0.09, 2.33  0.66 −0.50, 1.83   Other/unknown/multiple 0.08 −0.27, 0.43  1.06 −0.01, 2.14  Age (years) −0.06 −0.08, −0.04 <0.0001 0.02 −0.06, 0.10 0.58 Diabetes duration (years) 0.01 −0.03, 0.04 0.7545 0 −0.11, 0.11 0.94 BMI z-score −0.23 −0.32, −0.14 <0.0001 −0.43 −0.72, −0.14 0.004 Insulin regimen   <0.0001     Insulin pump Reference      Basal-bolus injections 0.42 0.23, 0.60      Other insulin regimens 0.83 0.39, 1.26      Unknown 0.5 −0.19, 1.19     Medication regimen      <0.0001  Insulin only    Reference   Insulin plus oral agent    −0.26 −0.89, 0.37   Metformin only    −3.09 −3.77, −2.42   Other oral agent    −1.91 −2.98, −0.83   None    −2.9 −3.57, −2.24   Unknown    −1.77 −3.35, −0.19  Blood glucose monitoring frequency   <0.0001   0.49  Less than once a day Reference  Reference   1–3 times/day 0.43 0.13, 0.73  −0.16 −0.83, 0.52   4 or more times/day −0.31 −0.60, −0.02  −0.47 −1.26, 0.31   CGM −0.48 −0.76, −0.20  −0.21 −1.07, 0.65   Unknown −1.13 −1.47, −0.80  −0.72 −1.57, 0.12  Household income   <0.0001   0.33  <$25,000 Reference  Reference   $25,000–49,999 −0.12 −0.41, 0.18  −0.19 −0.83, 0.46   $50,000–74,999 −0.55 −0.88, −0.22  −0.56 −1.52, 0.39   >$75,000 −0.72 −1.02, −0.42  −0.81 −1.76, 0.14   Do not know/refused −0.24 −0.53, 0.05  0.08 −0.44, 0.60  Health insurance   0.27   0.33  Private Reference  Reference   Medicaid/Medicare 0.22 −0.01, 0.45  −0.43 −0.96, 0.10   Other 0.12 −0.27, 0.52  0.16 −0.73, 1.06   None 0.28 −0.26, 0.82  −0.19 −0.96, 0.10  

The 2014–2019 cohort with type 2 diabetes had an adjusted HbA1c of 8.6 ± 0.12% (70 mmol/mol). There was a statistically significant difference in HbA1c level across the three time periods for participants with type 2 diabetes. The adjusted HbA1c level for the 2014–2019 cohort (8.6% [70 mmol/mol]) was relatively comparable to the 2002–2007 cohort (8.7% [72 mmol/mol] but was higher than the 2008–2013 cohort (8.3% [67 mmol/mol]). When examined by diabetes duration, YYA with type 2 diabetes with a diabetes duration of ≥10 years exhibited a temporal trend of worse glycemic control (2008–2013: 8.4% [68 mmol/mol] vs. 2014–2019: 10.1% [87 mmol/mol]). There was no statistically significant difference in the 2014–2019 cohort when compared with similarly aged YYA groups in the other two cohorts (Fig. 1B and Supplementary Table 1). Among participants with type 2 diabetes in the 2014–2019 cohort, the multivariate results revealed that HbA1c was associated with BMI and medication regimen, with those on metformin having a lower HbA1c as compared with those on insulin.

Many YYA with diabetes in the U.S. are not meeting desired glycemic targets despite increased availability of advanced diabetes technologies, newer therapies, and more aggressive glycemic targets over time. Overall, we found that adjusted mean HbA1c has increased for YYA with type 1 diabetes since the start of the SEARCH study, while the adjusted mean HbA1c in YYA with type 2 diabetes is relatively unchanged when comparing the SEARCH 2002–2007 to the most recent 2014–2019 SEARCH cohort. This contrasts with other countries where improved glycemic control and outcomes have been observed in YYA with diabetes. The SWEET project, which includes 22 centers from Europe, Australia, Canada, and India, demonstrated an improvement in glycemic control; individuals <25 years of age with type 1 diabetes had a mean adjusted HbA1c that declined from 8.4% [68 mmol/mol] to 7.9% [63 mmol/mol] between 2008–2010 and 2016–2018 (12). Similarly, the National Paediatric Diabetes Audit in England and Wales also reported a decline in median HbA1c of 8.6% [71 mmol/mol] in 2011/2012 to 7.8% [61.5 mmol/mol] in 2019/2020 in the pediatric population with diabetes (13).

In addition to differences in adjusted HbA1c over time (Table 2), examination by age group demonstrated significant differences in glycemic control (Fig. 1). We found that adjusted mean HbA1c levels for YYA with type 1 diabetes have increased over time across many age groups. Similar to the 2002–2007 SEARCH cohort, adjusted mean HbA1c was highest among 15–19-year-old participants with type 1 diabetes at 9.3% (78 mmol/mol) for the 2014–2019 cohort. However, it is notable that the adjusted HbA1c for this age group in the most recent SEARCH cohort was 0.5% (5 mmol/mol) higher when compared with 2002–2007 (8.8% [73 mmol/mol]). A comparable increase in adjusted mean HbA1c of 0.4% (4 mmol/mol) was observed in emerging adults, as the adjusted HbA1c for 20–24-year-old participants increased from 8.3% (67 mmol/mol) in the 2002–2007 cohort to 8.7% (72 mmol/mol) in the most recent cohort. The mean glycemic control for these age groups is comparable to what has recently been reported by the U.S. T1D Exchange Registry (15–18 years old: 9.3% [78 mmol/mol]; 18–25 years old: 8.9% [74 mmol/mol]), which reported a worsening in glycemic control in U.S. YYA with diabetes between 2010–2012 and 2016–2018 (9). While there have been significant advances in type 1 diabetes management (14), the burden of diabetes self-care remains demanding. Our findings highlight that the need to monitor blood glucose levels frequently or continuously and administer insulin reliably multiple times per day while balancing diet, physical activity, and other life activities continues to be challenging for adolescents and emerging adults with type 1 diabetes.

SEARCH has provided important information about YYA with type 2 diabetes in the U.S over the past two decades (3,15,16). Results from this study demonstrate that despite a growing recognition of the more rapidly progressive decline in β-cell function (17) and accelerated development of diabetes complications in youth-onset type 2 diabetes compared with adult-onset type 2 diabetes (18), glycemic control in U.S. YYA with type 2 diabetes has not improved over time. The adjusted mean HbA1c of YYA with type 2 diabetes with an average diabetes duration of 7.0 years in our sample was 8.6% (70 mmol/mol), which is comparable to the 8.4% (68 mmol/mol) reported for the Pediatric Diabetes Consortium Type 2 Diabetes Registry for youth with type 2 diabetes with a diabetes duration of ≥4 years (10).

The observed pattern of worsening glycemic control with increasing duration of type 2 diabetes, independent of many other potential correlates, was likely driven in part by the progressive loss of β-cell function in these participants. The adjusted HbA1c of 10.1% (87 mmol/mol) for type 2 diabetes participants with a disease duration of ≥10 years in the most recent SEARCH cohort is very concerning and provides further confirmation of the high degree of treatment failure in YYA with type 2 diabetes, as well as the need for aggressive intervention (19). Given the evidence that lifestyle intervention alone to achieve or maintain normal blood glucose levels in type 2 diabetes is often unsuccessful in YYA, efforts are needed to support medical providers providing care for YYA with type 2 diabetes to become more comfortable with recommended care and the use of pharmacologic agents beyond metformin and insulin, such as liraglutide, in this population (7).

Previous research, including data from meta-analyses and large international registries, has demonstrated that the use of insulin pumps in youth with type 1 diabetes is associated with lower HbA1c as compared with multiple daily injections (20–22). Separately, CGM alone has been associated with improved glycemic control, with the benefits increasing as CGM use increases (23). Similarly, we found that 2014–2019 SEARCH participants using an insulin pump had an HbA1c level that was 0.4% (4 mmol/mol) lower than those managing their type 1 diabetes with basal-bolus injections. CGM users had an ∼0.2% lower HbA1c level even when compared with those who were monitoring their glucose four or more times per day with fingerstick blood glucose checks. It is worth noting, however, that despite the literature supporting the benefits of pump therapy and CGM systems in the YYA population, universal adoption has not been achieved (24,25). For example, while insulin pump use nearly doubled from 2002–2007 (24.9%) to 2014–2019 (49.1%), there was a minimal increase in the 2014–2019 cohort from the 44.7% found to be using insulin pumps in 2008–2013. Previous studies have shown that household income and parental education are predictive of insulin pump use (24,26). Given the challenges that YYA with type 1 diabetes face in achieving the goals of therapy, addressing differences in access and use of diabetes technologies offers a potential avenue to help the YYA population meet glycemic targets.