446 Care Volume 41, March 2018

Paola Gilsanz,1,2 Andrew J. Karter,1 The Bidirectional Association Michal Schnaider Beeri,3,4 Charles P. Quesenberry Jr.,1 and Between Depression and Severe Rachel A. Whitmer 1,2 Hypoglycemic and Hyperglycemic Events in Type 1 Diabetes Diabetes Care 2018;41:446–452 | https://doi.org/10.2337/dc17-1566

OBJECTIVE Severe and (“severe dysglycemia”) are serious compli- cations of type 1 diabetes (T1D). Depression has been associated with severe dys- glycemia in type 2 diabetes but has not been thoroughly examined specifically in T1D. We evaluated bidirectional associations between depression and severe dysglycemia among older people with T1D.

EPIDEMIOLOGY/HEALTH SERVICES RESEARCH RESEARCH DESIGN AND METHODS We abstracted depression and severe dysglycemia requiring emergency room visit or hospitalization from medical health records in 3,742 patients with T1D during the study period (1996–2015). Cox proportional hazards models estimated the associa- tions between depression and severe dysglycemia in both directions, adjusting for

demographics, micro- and macrovascular complications, and HbA1c.

RESULTS During the study period, 41% had depression and 376 (11%) and 641 (20%) had hyperglycemia and hypoglycemia, respectively. Depression was strongly associated with a 2.5-fold increased risk of severe hyperglycemic events (hazard ratio [HR] 2.47 [95% CI 2.00, 3.05]) and 89% increased risk of severe hypoglycemic events (HR 1.89 [95% 1Division of Research, Kaiser Permanente, Oak- CI 1.61, 2.22]). The association was strongest within the first 6 months (HRhyperglycemia land, CA 2Department of Epidemiology and Biostatistics, 7.14 [95% CI 5.29, 9.63]; HRhypoglycemia 5.58 [95% CI 4.46, 6.99]) to 1 year (HRhyperglycemia 5.16 [95% CI 3.88, 6.88]; HR 4.05 [95% CI 3.26, 5.04]) after depression University of California, San Francisco, San Fran- hypoglycemia cisco, CA diagnosis. In models specifying severe dysglycemia as the exposure, hyperglycemic 3Icahn School of Medicine at Mount Sinai, New and hypoglycemic events were associated with 143% (HR 2.43 [95% CI 2.03, 2.91]) York, NY and 74% (HR 1.75 [95% CI 1.49, 2.05]) increased risk of depression, respectively. 4The Joseph Sagol Neuroscience Center, Sheba Medical Center, Ramat Gan, Israel CONCLUSIONS Corresponding author: Paola Gilsanz, paola Depression and severe dysglycemia are associated bidirectionally among patients [email protected]. with T1D. Depression greatly increases the risk of severe hypoglycemic and hyper- Received 28 July 2017 and accepted 20 Novem- ber 2017. glycemic events, particularly in the first 6 months to 1 year after diagnosis, and This article contains Supplementary Data online depression risk increases after severe dysglycemia episodes. at http://care.diabetesjournals.org/lookup/ suppl/doi:10.2337/dc17-1566/-/DC1. Maintaining healthy blood levels is critical for people with type 1 diabetes (T1D) © 2017 by the American Diabetes Association. to reduce risk of diabetic microvascular (e.g., retinopathy) and macrovascular (e.g., Readers may use this article as long as the work is properly cited, the use is educational and not cardiovascular disease) complications. Hypoglycemia (e.g., blood glucose levels for profit, and the work is not altered. More infor- ,54 mg/dL) (1) can interfere with brain glucose supply, balance, and coordination mation is available at http://www.diabetesjournals (2). Hyperglycemia occurs at the other end of the spectrum (e.g., blood glucose .org/content/license. care.diabetesjournals.org Gilsanz and Associates 447

levels .250 mg/dL) (3) and is associated and depression (24), and it is critical to anxiety and depressed mood (309.28), with endothelial dysfunction, arterial stiff- identify if severe hypo- and hyperglyce- and depressive disorder not otherwise ness, and neuropathy (4). Severe cases of mic events could trigger a similar harmful classified (311). In sensitivity analyses, a hypoglycemia and hyperglycemia are de- cycle between depression and future re- more narrow definition of depression was fined by the need for intervention by an- peat severe dysglycemic events. used, including only diagnosis of major other person (5), and both types of severe In this study, we examine the longitu- depressive disorder, depressive type psy- dysglycemia can result in coma, ventricu- dinal associations between depression chosis, and dysthymic disorder. An indi- lar premature beats, or death (6). More- and severe hypo- and hyperglycemic vidual was considered depressed from over, both types are strongly associated events that required hospitalization or the date of diagnosis onward. with rates of hospital readmission (7). an emergency room visit in a cohort of Severe hypoglycemia is associated with older individuals with T1D. Additionally, Severe Hyper- or Hypoglycemic Events fi triple the risk of mortality (8) and strongly we examine if this relationship is bidirec- Severe dysglycemic events were de ned linked to poorer quality of life in older tional by determining if severe dysglyce- as episodes of hyperglycemia or hypogly- patients with diabetes (9). Diabetic mic events elevate the risk of depression. cemia that resulted in hospital admission ketoacidosis (DKA) is the most common or emergency room care. Hyperglycemic form of severe hyperglycemia among RESEARCH DESIGN AND METHODS episodes between 1996 and 2015 were fi people with T1D and is triggered by re- identi ed as the following ICD-9, Clini- Study Population cal Modification (ICD-9-CM) codes: duced effective and increased This study examines a dynamic cohort of 249.2 (secondary diabetes mellitus with counterregulatory hormones. DKA is the Kaiser Permanente Northern California hyperosmolarity), 250.1 (diabetes with leading cause of mortality among children (KPNC) members with T1D who were at ketoacidosis), and 250.2 (diabetes and young adults with T1D (6). Unfortu- least 50 years old during the study period with hyperosmolarity). The following set nately, both types of severe dysglycemia (1 January 1996 to 30 September 2015) as of ICD-9-CM codes was used to identify are common; severe hypoglycemia occurs part of an aging and diabetes study. KPNC hypoglycemic episodes between 1996 110–320 times per 100 patients years is an integrated health care delivery sys- and 2015 (these codes were validated (10), and DKA is estimated to occur up tem serving over 3 million members, who in a prior study compared with medical to 12 times per 100 patients years in chil- are representative of the population of the record review) (31): 251.0 (hypoglycemic dren and adolescents with T1D (3). geographic region with the exceptions of coma), 251.1 (other specified hypoglyce- Depression is approximately two to the extremes of income (25–27). KPNC mia), and 251.2 (hypoglycemia, unspec- three times more common among adults maintains the Diabetes Registry that en- ified). The following ICD-9-CM codes with T1D than those without (11,12). Prev- compasses all members with diabetes were not used to identify hypoglycemic alence estimates vary from 5 to 32% de- identified through laboratory data, phar- episodes: 250.3 diabetes mellitus with pending on the study design and how macy records, and inpatient and outpa- coma (because it does not distinguish be- depression was defined (12,13). Studies tient diagnoses of diabetes (28,29). A tween hypoglycemia and DKA), 250.8 di- show that depression is associated with review of the Diabetes Registry identified abetes mellitus with other specified nonadherence to diabetes treatment individuals with T1D using the following manifestations (because it does not spec- (14,15) and worse glycemic control (16). three criteria, all of which had to be met: ify hypoglycemia), 270.3 leucine-induced The vast majority of research on depres- 1) having at least two T1D ICD-9 codes hypoglycemia, 775.6 neonatal hypoglyce- sion and severe dysglycemic events has without a T2D code or 75% or more of mia, and 775.0 hypoglycemia in an infant had samples comprised of people exclu- their diagnostic codes indicated T1D, 2) born to a mother with diabetes. sively (17–19) or predominantly with filled an insulin prescription at any point type 2 diabetes (T2D) (20,21) even though fi between 1996 and 2015, and 3) did not ll Death these events are much more common prescriptions of any hypoglycemic agents Dates of death were extracted from med- among people with T1D. People with other than insulin (30). Individuals en- ical records, Social Security Administration T1D are up to four times more likely to tered the cohort (i.e., their baseline date) data sets, and California death certificates experience severe hypoglycemia (6) and at the first date between 1 January 1996 between 1996 and 2015. about three times more likely to experi- and 30 September 2015 that an individual ence DKA than those with T2D (3). To our was at least 50 years old and met the Covariates knowledge, no previous study has exam- above criteria for T1D. The final analytic All covariates were from a participant’s ined the longitudinal associations between sample included 3,742 individuals with baseline (i.e., the participant’s time of depression and severe dysglycemic events T1D. entry into the cohort). Demographic in- in a large sample of well-characterized formation, including age, sex, and race, people with T1D. Depression was obtained from the KPNC membership The association between depression The following ICD-9 codes were used to database. Glycosylated hemoglobin and glycemic control is posited to be bi- identify time-updated depression through- (HbA1c) measurements and the following directional (16,22,23) and can result in a out the study period in the electronic med- microvascular and macrovascular compli- harmful cycle in which poor glycemic con- ical records: major depressive disorder cations were abstracted from the start of trol leads to a depressed mood, further (296.2x and 296.3x), depressive type psy- the electronic medical records (1996) to exacerbating poor self-care (16). A cross- chosis (298.0x), dysthymic disorder participant entry date (ICD-9 code defini- sectional study also found an association (300.4x), adjustment disorder with de- tion in Supplementary Table 1): neuropa- between severe hypoglycemic events pressed mood (309.0x) or with mixed thy, peripheral artery disease, myocardial 448 Depression and Dysglycemia in Type 1 Diabetes Diabetes Care Volume 41, March 2018

infarction, and stroke. In sensitivity anal- quintiles of baseline HbA1c in fully ad- RESULTS yses allowing for a nonlinear association justed models with the third quintile as Overall, the sample had a mean age of between HbA1c and severe dysglycemic the reference group. Third, we further 56 years and was 79% white, 4% Asian, events, baseline HbA1c was grouped into adjusted for baseline values of the dysgly- and 15% black/Hispanic/other (Table 1). quintiles, with the third quintile serving cemic event of interest when examining Depression was present among 20% of as the reference group. The 1,068 individ- the association between our original def- the sample at baseline and an additional uals missing baseline values of HbA1c inition of depression and severe dysglyce- 21% was diagnosed during the study. A fl were agged by a missing indicator and mic events during the four time periods. narrow definition of depression (i.e., ma- assigned to the reference group of the This was the only analysis for which indi- jor depressive disorder, depressive type quintile variable and the mean for the viduals with baseline dysglycemic events psychosis, and dysthymic disorder) was continuous variable. were included in the sample. Last, we ex- present in 11% of the sample and 14% amined a possible effect modification of received a diagnosis during follow-up. Statistical Analysis the relationship between depression and There were 376 cases (11%) of severe hy- We examined the baseline characteristics time to a severe dysglycemic event by perglycemia and 641 cases (20%) of severe of members with and without depression race,sex,baselineHbA status, and micro- ; during follow-up. Cox proportional haz- 1c hypoglycemia during follow-up; 5% expe- ards models were specified to evaluate or macrovascular complications using in- rienced both. The mean follow-up time fi – if depression during follow-up (i.e., time- teraction terms and strati ed models. was 5.9 years (SD 5.25; range 0.003 19.7) varying exposure) was associated with Cox proportional hazards models were for analyses examining severe hypergly- elevated risk of severe dysglycemia, also implemented to examine if severe cemia as the outcome and 5.1 years examining severe hyperglycemic and hy- dysglycemic events during follow-up (SD 4.80; range 0.003–19.7) for analyses poglycemic events in separate models. were associated with risk of depression examining severe hypoglycemia as the Individuals with baseline values (i.e., be- during follow-up, excluding those with outcome. At the end of follow-up for the tween 1996 and cohort entry) of the baseline depression. Hyperglycemia and hyperglycemia analyses, 16% died with- outcome were excluded. For example, in- hypoglycemia were separately examined out a severe hyperglycemic event during dividuals with baseline hyperglycemia as exposures and covariates were added follow-up, 25% were censored due to a were excluded from analyses examining to the models in the same order as before lapse in membership, and 48% were live the association between depression and (demographics, microvascular complica- members of KPNC without a severe hy- hyperglycemia during follow-up. Individ- tions, macrovascular complications, and perglycemic event during follow-up. At the uals were censored on the date of the HbA1c). Last, for each set of dysglycemia end of follow-up for hypoglycemia analyses, first severe hyper- or hypoglycemic event models, we further adjusted for the base- 15% died without a severe hypoglycemic (depending on the model’s outcome of in- line values of the other form of dysglyce- event during follow-up, 24% were censored terest), the beginning of a gap in member- mia. All analyses were conducted on SAS due to a lapse in membership, and 40% ship .90 days, or death, or at the end of statistical software version 9.3 (SAS Insti- were members of KPNC without a severe the study period on 30 September 2015. tute Inc., Cary. NC). hypoglycemic event during follow-up. To assess if the association between de- pression and severe dysglycemic events Table 1—Baseline characteristics by time-updated depression varied by time since depression, we com- No depression during Depression during Total pared the risk of dysglycemic events oc- study study sample P value curring within 6 months, within 1 year, and .1 year after depression with those Number of people 2,200 (58.8)* 1,542 (41.2)* 3,742 without depression. Covariates were added Demographics Age at cohort entry, to the model in groups: demographics mean (SD) 57.0 (9.1) 54.7 (7.3) 56.1 (8.5) ,0.0001 (baseline age, race, and sex), microvascular Male 1,304 (59.3) 666 (43.2) 1,970 ,0.0001 complications (neuropathy and peripheral (52.7) artery disease), and macrovascular compli- White 1,702 (77.4) 1,254 (81.3) 2,956 0.23 cations (myocardial infarction and stroke). (79.0) In fully adjusted models, we also controlled Asian 99 (4.5) 47 (3.1) 146 (3.9) 0.05 Other 317 (14.4) 224 (14.5) 541 (14.5) 0.10 for HbA , which likely serves as a media- 1c Missing 82 (3.7) ,20 ,102 ,0.0001 tor of the depression-dysglycemia rela- Baseline health conditions tionship in both directions. Neuropathy 214 (9.7) 301 (19.5) 515 (13.8) ,0.0001 We conducted four sets of sensitivity PAD 115 (5.2) 149 (9.7) 264 (7.1) ,0.0001 analyses examining depression as a risk MI 38 (1.7) 57 (3.7) 95 (2.5) ,0.0001 factor for severe dysglycemic events. Stroke 64 (2.9) 86 (5.6) 150 (4.0) ,0.0001 First, we used the narrower definition of Hyperglycemia 118 (5.4) 198 (12.8) 316 (8.4) ,0.0001 depression and examined its association Hypoglycemia 263 (12.0) 279 (18.1) 542 (14.5) ,0.0001 with severe dysglycemic events any time HbA1c, mean (SD) 8.3 (2.0) 8.3 (1.9) 8.3 (1.9) 0.88 Depression 0 (0) 752 (48.8) 752 (20.1) ,0.0001 after depression. Second, we allowed for a nonlinear association between HbA Data are n (%) unless stated otherwise. MI, myocardial infarction; PAD, peripheral artery disease. 1c *Row percent. and severe dysglycemic events by using care.diabetesjournals.org Gilsanz and Associates 449

Depression and Severe Glycemic risk of a severe hyperglycemic event (HR hypoglycemic event any time after de- Event Onset 5.16 [95% CI 3.88, 6.88]) and over four pression (Supplementary Table 2). These After controlling for demographics, times the risk of a severe hypoglycemic associations persist after adjusting for HbA1c, and microvascular and macrovas- event (HR 4.05 [95% CI 3.26, 5.04]) in the micro- and macrovascular complica- fi cular complication, depression was asso- rst year. In fully adjusted models, the risk tion (HRhyperglycemic event 2.49 [95% fi ciated with more than double the risk of of hyperglycemia after the rst year was CI 1.98, 3.12]; HRhypoglycemic event 1.95 severe hyperglycemic events (hazard ra- less elevated but significantly higher for [95% CI 1.63, 2.32]). Effect estimates tio [HR] 2.47 [95% CI 2.00, 3.05]) and those with depression compared with those of the relationship between depres- almost a doubling of risk of severe hypo- without (HR 1.44 [95% CI 1.14, 1.83]). De- sion and dysglycemic events remained glycemic events (HR 1.89 [95% CI 1.61, pression was not significantly associated similar when adjusting for quintiles with an elevated risk of a severe hypoglyce- 2.22]) at any point after depression diag- of HbA1c (versus a linear form) (Sup- nosis (Table 2). The association between mic event occurring more than 1 year or plementary Table 3) and after further depression and risk of severe dysglycemic more later (HR 1.10 [95% CI 0.91, 1.32]). adjustment for baseline dysglycemia events was substantially stronger within Sensitivity Analyses of Depression (Supplementary Table 4). In models ad- fi fi the rst 6 months and within the rst year as a Risk Factor for Severe Dysglycemic justing for baseline severe dysglycemic than in the later time period. Compared Events events, risk for either dysglycemic event with people without depression, those In models adjusting for demographics, was greatest during the first 6 months with depression had over sevenfold and depression defined more narrowly (i.e., after depression diagnosis (fully adjusted fi vefold risk of severe hyper- and hypogly- diagnosis of major depressive disorder, HRsevere hyperglycemic event 7.82 [95% CI fi cemic events within the rst 6 months, depressive type psychosis, or dysthymic 6.15, 9.94]; HRsevere hypoglycemic event 5.68 respectively (HRhyperglycemic event 7.14 disorder) was associated with more [95% CI 4.77, 6.76]). Effect estimates [95% CI 5.29, 9.63]; HRhypoglycemic event than double the risk (HR 2.46 [95% CI were lowest, but still significant, for dys- 5.58 [95% CI 4.46, 6.99]). Compared 1.97, 3.08]) of a severe hyperglycemic glycemic events occurring more than with people without depression, people event and almost double the risk (HR 1 year after depression diagnosis (fully with depression had over five times the 1.98 [95% CI 1.66. 2.37]) of a severe adjusted HRsevere hyperglycemic event 1.38

Table 2—HR (95% CI) of depression predicting time to dysglycemic events adjusting for baseline covariates Severe hyperglycemic event Severe hypoglycemic event Adjusted for severe dysglycemic events any time after depression Number of eligible severe glycemic events 376 641 Demographics 2.38 (1.93, 2.94) 1.89 (1.61, 2.22) Demographics and microvascular complications 2.40 (1.95, 2.94) 1.88 (1.60, 2.20) Demographics and macrovascular complications 2.39 (1.94, 2.95) 1.87 (1.60, 2.20) Demographics and micro- and macrovascular complications 2.40 (1.94, 2.97) 1.86 (1.58, 2.19)

Demographics, micro- and macrovascular complications, and HbA1c 2.47 (2.00, 3.05) 1.89 (1.61, 2.22) Adjusted for severe dysglycemic events within 6 months of depression Number of eligible severe glycemic events 244 447 Demographics 7.13 (5.29, 9.62) 5.72 (4.57, 7.15) Demographics and microvascular complications 7.01 (5.19, 9.45) 5.66 (4.52, 7.09) Demographics and macrovascular complications 7.13 (5.29, 9.62) 5.69 (4.55, 7.12) Demographics and micro- and macrovascular complications 7.01 (5.19, 9.45) 5.36 (4.49, 7.05)

Demographics, micro- and macrovascular complications, and HbA1c 7.14 (5.29, 9.63) 5.58 (4.46, 6.99) Adjusted for severe dysglycemic events within first year of depression Number of eligible severe glycemic events 251 456 Demographics 5.18 (3.90, 6.90) 4.18 (3.36, 5.18) Demographics and microvascular complications 5.11 (3.84, 6.81) 4.14 (3.33, 5.14) Demographics and macrovascular complications 5.18 (3.89, 6.89) 4.15 (3.34, 5.15) Demographics and micro- and macrovascular complications 5.11 (3.84, 6.80) 4.10 (3.30, 5.10)

Demographics, micro- and macrovascular complications, and HbA1c 5.16 (3.88, 6.88) 4.05 (3.26, 5.04) Severe dysglycemic events 1+ years after depression Number of eligible severe glycemic events 306 524 Demographics 1.40 (1.11, 1.77) 1.11 (0.93, 1.34) Demographics and microvascular complications 1.39 (1.10, 1.76) 1.09 (0.91, 1.32) Demographics and macrovascular complications 1.42 (1.12, 1.80) 1.10 (0.92, 1.33) Demographics and micro- and macrovascular complications 1.41 (1.11, 1.79) 1.09 (0.90, 1.31)

Demographics, micro- and macrovascular complications, and HbA1c 1.44 (1.14, 1.83) 1.10 (0.91, 1.32) Demographics included age, sex, and race. Microvascular complications included neuropathy and peripheral artery disease. Macrovascular complications included myocardial infarction and stroke. 450 Depression and Dysglycemia in Type 1 Diabetes Diabetes Care Volume 41, March 2018

[95% CI 1.13, 1.69]; HRsevere hypoglycemic event with depression are considerably more 75% elevated risk of depression. The bidi- 1.17 [95% CI 1.01, 1.37]). There was no likely to experience severe dysglycemic rectional relationship likely enables a harm- evidence of effect modification of the events, particularly within the first ful feed-forward cycle, with depression and relationship between depression and 6 months to 1 year after the diagnosis severe glycemic events increasing the risk severe glycemic events by race, sex, of depression. Depression was associated of the other. baseline macrovascular or microvas- with more than double the risk of severe A meta-analysis has shown an associa- cular complications, or baseline HbA1c hyperglycemic events and 86% increased tion between depression and glycemic (Supplementary Table 5). risk of severe hypoglycemic events any control (16), although some more recent time after depression diagnosis, regard- studies have shown conflicting results Severe Dysglycemic Events and Risk less of microvascular and macrovascular (32,33). Severe hyper- and hypoglycemic of Depression complications. The increased risk was par- events are common among people with Among those who did not have depression ticularly robust in the first 6 months when T1D and increase the risk of neurological at the start of the study, individuals who people with depression had more than damage (34), coma (6), dementia (31), experienced a severe hyperglycemic event five to seven times the risk of such a severe and mortality (6,8). Yet little research during the study had more than double dysglycemic event even after accounting has been conducted examining the asso- (HR 2.39 [95% CI 2.00, 2.84]) the risk of for microvascular and macrovascular com- ciation between depression and severe dysglycemic events. To our knowledge, future depression than their counterparts plications. Depression was associated with no study has previously examined the as- who did not experience a hyperglycemic increased risk of severe dysglycemic events sociation between depression and severe event, adjusting for demographics (Table at all time periods even after adjusting for hyperglycemic events, and only two prior 3). This association was similar after ad- baseline dysglycemic events, and the first studies have examined severe hypoglyce- justing for baseline HbA1c, microvascular 6 months continued to be associated with and macrovascular complications, and mic events (20,21). Both studies predom- five to seven times the risk of such events. baseline hypoglycemia (HR 2.43 [95% CI inantly included people with T2D and one This suggests that there is a critical period 2.00, 2.99]). Among those without depres- was cross-sectional (20). In a 5-year study shortly after depression diagnosis during sion at baseline, those with a severe hypo- including 4,117 health care plan members which patients are at extremely high risk, glycemic episode had a 75% greater risk of with T1D and T2D (95.6% of the sample), and treatment of depression and glucose depression, adjusting for demographics (HR depression was associated with a 78% in- monitoring may be particularly important 1.75 [95% CI 1.50, 2.05]). This association creased risk of a severe hypoglycemic event for preventing these dangerous episodes. persisted in models further adjusting for (i.e., hypoglycemic episode requiring emer- The relationship between depression microvascular and macrovascular compli- gency room visits or hospitalizations) (21). and severe dysglycemic events was bidi- cations, HbA1c, and baseline hyperglycemia Thereisalargebodyofliteratureshow- (HR 1.74 [95% CI 1.48, 2.05]). rectional. Compared with individuals who ing that depression in diabetes impacts did not experience severe glycemic events, self-care.Depressionmayleadtoanin- individuals who experienced severe hyper- creased risk of severe dysglycemic events CONCLUSIONS glycemic events had more than twice the through altered behaviors, e.g., worsening In this large sample of older people with risk of depression and those who experi- adherence to medication regimen, diet, T1D, we found evidence that individuals enced severe hypoglycemic events had a physical activity, and self-monitoring of blood glucose (14,15,21,35,36). Poorer self-care may worsen glycemic control Table 3—HR (95% CI) of dysglycemic event predicting time to depression adjusting and put individuals at risk for severe dys- for baseline covariates glycemic events. Physiological changes Depression due to depression may also mediate the depression–severe dysglycemic event Severe hyperglycemic events Demographics 2.39 (2.00, 2.84) relationship. Depression is associated Demographics and microvascular complications 2.33 (1.95, 2.78) with dysregulation of the hypothalamic- Demographics and macrovascular complications 2.39 (2.00, 2.85) pituitary-adrenal axis (37). Increased levels Demographics and micro- and macrovascular complications 2.35 (1.97, 2.80) of catabolic counterregulatory hormones Demographics, micro- and macrovascular complications, and HbA1c 2.43 (2.03, 2.91) increase blood glucose levels (37) and Demographics, micro- and macrovascular complications, HbA1c,and may make it harder to have appropriate severe hypoglycemia 2.43 (2.00, 2.99) glycemic control. Severe hypoglycemic event By definition, all cases of severe dysgly- Demographics 1.75 (1.50, 2.05) Demographics and microvascular complications 1.73 (1.47, 2.02) cemic events captured in this sample had Demographics and macrovascular complications 1.73 (1.48, 2.03) resulted in an emergency room visit or Demographics and micro- and macrovascular complications 1.71 (1.46, 2.01) hospitalization and very likely were con- Demographics, micro- and macrovascular complications, and HbA1c 1.75 (1.49, 2.05) sidered a stressful life event. Our finding Demographics, micro- and macrovascular complications, HbA1c,and that severe dysglycemic events predicted severe hyperglycemia 1.74 (1.48, 2.05) depression diagnosis is consistent with Analyses exclude individuals with depression at baseline. Covariates from baseline included a body of literature linking stressful life the following: demographics (age, sex, and race), microvascular complications (neuropathy and events to depression onset (38,39). Our peripheral artery disease), and macrovascular complications (myocardial infarction and stroke). findings are also consistent with a body of care.diabetesjournals.org Gilsanz and Associates 451

literature providing evidence that poor an approximation. Additionally, it is possible self-care, including depression. The rela- glycemic control may worsen mood that there are cases of undiagnosed depres- tionship between depression and severe (22). Hyperglycemia has been shown to sion in our sample. If individuals with un- dysglycemic events is also important to incite anxiety and, in a cross-sectional diagnosed depression are also more likely study among younger populations, espe- study, was associated with anger, tension, to experience dysglycemic events than in- cially since some research has found de- and sadness among people with T1D dividuals without depression, we under- pression to be most prevalent in the first (22,40). A study including people with estimate the effect of depression on years after diagnosis (43). Further re- T1D and T2D found an association be- dysglycemic events. Our findings show sig- search is needed to develop multifaceted tween severe hypoglycemic events and nificant differences in severe dysglycemic interventions that address the intersec- depression during a 1-year period but events by depression status but no differ- tion of physical and mental health of pa- also did not establish temporal order (24). ence in baseline HbA1c levels, suggesting tients with T1D to reduce the morbidity To our knowledge, our study is the first to that baseline HbA1c levels may not ade- and mortality attributed to either sphere examine severe dysglycemic events as a quately capture glycemic control. We did of health. Given the danger of these se- risk factor for depression and to examine not have access to information regarding vere dysglycemic events, patients with the temporal ordering between severe antidepressant use, which some studies T1D with comorbid depression represent dysglycemic events and depression dur- have shown to be associated with in- an extremely vulnerable population that ing the follow-up period. creased risk of dysglycemic events (41, requires increased vigilance in clinical care. In general, people with diabetic com- 42); thus, we do not know how antide- plications are more likely to experience pressant usage may impact associations depression (37); thus, diabetic complica- between depression and risk of dysglyce- Funding. This work was supported by the Na- tions may mediate the association be- mic events in the current study. Our data tional Institute on Aging (grant R01 AG047500 to tween hyper- and hypoglycemia and also do not include brain imaging, so we R.A.W.). P.G. is supported by the University of California, San Francisco, Training for Research on depression. However, in our study, the are unable to examine the direct effect Aging and Chronic Disease (National Institute on association persisted even when taking of depression and severe dysglycemic Aging grant T32 AG049663). A.J.K. is supported by complications into account, suggesting events on brain structure or integrity. the National Institute of Diabetes and Digestive that other unidentified pathways are We do not know if individuals who were and Kidney Diseases (grants R01 DK103721 and also at play. Understanding the associa- censored due to lapse in membership or R01 DK081796) and the Health Delivery Systems Center for Diabetes Translational Research (Na- tion between severe dysglycemic events death would have experienced severe tional Institute of Diabetes and Digestive and and depression is critical to uncovering dysglycemic events had they remained Kidney Diseases grant P30 DK092924). potential targets of intervention to break alive and in the sample. Last, these analy- Thefundershadnoroleinthedesignandconduct the cyclical association between them. ses were restricted to individuals 50 years of the study; collection, management, analysis, and This study includes a well-characterized old and above as part of a study examining interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit sample of people with T1D with prospec- aging among people with T1D, which may the manuscript for publication. tively identified depression diagnoses and limit the generalizability of these results Duality of Interest. No potential conflicts of in- severe dysglycemic events. As an inte- to younger populations. terest relevant to this article were reported. grated health care delivery system, the This is the first longitudinal study of the Author Contributions. P.G. performed the liter- vast majority of care to members is pro- bidirectional relationship between severe ature search, analyzed and interpreted data, and wrote the manuscript. A.J.K. and M.S.B. interpreted vided at KPNC facilities and all diagnoses dysglycemic events and depression in a data and revised the manuscript. C.P.Q. analyzed and are captured by comprehensive electronic large sample of older adults with T1D. interpreted data and revised the manuscript. R.A.W. medical records. These records provide in- Our study shows a substantially increased designed the study, analyzed and interpreted data, formation on a wide range of complications risk of severe hyperglycemic and hypogly- and revised the manuscript. P.G. is the guarantor of this work and, as such, had full access to all the data in relatedtoT1D.Detailedinformationre- cemic events overall and particularly in the study and takes responsibility for the integrity of garding date of depression diagnosis and the first year after diagnosis of depres- the data and the accuracy of the data analysis. severe dysglycemic events enabled us to sion, highlighting the need for increased implement time-dependent analyses. Un- clinical vigilance for these patients during References fortunately, we do not have measures of the vulnerable period after a depression 1. American Diabetes Association. Glycemic the severity of depressive symptoms and diagnosis. This study also demonstrated targets. Sec. 6. In Standards of Medical Care in d could not test for a dose-response associa- an elevated risk of depression after se- Diabetes 2017. Diabetes Care 2017;40(Suppl. 1): S48–S56 tion. However, we do use two different vere hyperglycemic and hypoglycemic 2. Frier BM. Hypoglycaemia in diabetes mellitus: definitions of depression, the broader one events. Together, these results suggest epidemiology and clinical implications. Nat Rev of which includes less severe symptom lev- that people with T1D who are recently Endocrinol 2014;10:711–722 els, such as in dysthymic disorder. The tim- diagnosed with comorbid depression are 3. Rewers A. Chapter 17. Acute metabolic com- ing of a depression diagnosis may lag at a very high risk of severe dysglycemic plications in diabetes. In Diabetes in America.3rd ed. Cowie CC, Casagrande SS, Menke A, Cissell considerably from the onset of depressive events and that any resulting severe dys- MA, Eberhardt MS, Meigs JB, Gregg EW, Knowler symptoms, and a severe dysglycemic event glycemic events may be associated with WC, Barrett-Connor E, Becker DJ, Brancati FL, resulting in emergency department or hos- greater risk of depression. Glycemic con- Boyko EJ, Herman WH, Howard BV, Narayan pital utilization could have been preceded trol is the cornerstone of prevention of KMV, Rewers M, Fradkin JE, Eds. Bethesda, MD, National Institutes of Health, 2017 by a longer history of less severe hypoglyce- acute and chronic diabetic complications, 4. Retnakaran R, Zinman B. Type 1 diabetes, hy- mia or hyperglycemia. Thus, the time order- and it is critical that clinicians treating pa- perglycaemia, and the heart. Lancet 2008;371: ing of our analyses should be considered tients with diabetes mitigate obstacles to 1790–1799 452 Depression and Dysglycemia in Type 1 Diabetes Diabetes Care Volume 41, March 2018

5. Workgroup on Hypoglycemia, American Dia- 18. Green AJ, Fox KM, Grandy S; SHIELD Study electronic health record data. Diabetes Care betes Association. Defining and reporting hypo- Group. Self-reported hypoglycemia and impact 2013;36:914–921 glycemia in diabetes: a report from the American on quality of life and depression among adults 31. Whitmer RA, Karter AJ, Yaffe K, Quesenberry Diabetes Association Workgroup on Hypoglyce- with type 2 diabetes mellitus. Diabetes Res Clin CP Jr, Selby JV. Hypoglycemic episodes and risk of mia. Diabetes Care 2005;28:1245–1249 Pract 2012;96:313–318 dementia in older patients with type 2 diabetes 6. Umpierrez G, Korytkowski M. Diabetic 19. Zhang Y, Ting RZ, Yang W, et al.; China De- mellitus. JAMA 2009;301:1565–1572 emergencies - ketoacidosis, hyperglycaemic hy- pression in Chinese Patients with Type 2 Diabetes 32. Fisher L, Glasgow RE, Strycker LA. The rela- perosmolar state and hypoglycaemia. Nat Rev En- (DD2) Study Group. Depression in Chinese pa- tionship between diabetes distress and clinical docrinol 2016;12:222–232 tients with type 2 diabetes: associations with hy- depression with glycemic control among patients 7. McCoy RG, Lipska KJ, Herrin J, Jeffery MM, perglycemia, hypoglycemia, and poor treatment with type 2 diabetes. Diabetes Care 2010;33: Krumholz HM, Shah ND. Hospital readmissions adherence. J Diabetes 2015;7:800–808 1034–1036 among commercially insured and Medicare Ad- 20. Honkasalo MT, Elonheimo OM, Sane T. Se- 33. Fisher L, Mullan JT, Arean P, Glasgow RE, vantage beneficiaries with diabetes and the im- vere hypoglycaemia in drug-treated diabetic pa- Hessler D, Masharani U. Diabetes distress but pact of severe hypoglycemic and hyperglycemic tients needs attention: a population-based study. not clinical depression or depressive symptoms events. J Gen Intern Med 2017;32:1097–1105 Scand J Prim Health Care 2011;29:165–170 is associated with glycemic control in both cross- 8. McCoy RG, Van Houten HK, Ziegenfuss JY, Shah 21. Katon WJ, Young BA, Russo J, et al. Associa- sectional and longitudinal analyses. Diabetes Care ND, Wermers RA, Smith SA. Increased mortality of tion of depression with increased risk of severe 2010;33:23–28 patients with diabetes reporting severe hypogly- hypoglycemic episodes in patients with diabetes. 34. Suh SW, Hamby AM, Swanson RA. Hypogly- cemia. Diabetes Care 2012;35:1897–1901 Ann Fam Med 2013;11:245–250 cemia, brain energetics, and hypoglycemic neuro- 9. Laiteerapong N, Karter AJ, Liu JY, et al. Corre- 22. Lustman PJ, Clouse RE. Depression in diabetic nal death. Glia 2007;55:1280–1286 lates of quality of life in older adults with diabetes: patients: the relationship between mood and gly- 35. Lustman PJ, Clouse RE, Ciechanowski PS, the diabetes & aging study. Diabetes Care 2011; cemic control. J Diabetes Complications 2005;19: Hirsch IB, Freedland KE. Depression-related hy- 34:1749–1753 113–122 perglycemia in type 1 diabetes: a mediational ap- 10. Little SA, Leelarathna L, Barendse SM, et al. 23. Renn BN, Feliciano L, Segal DL. The bidirec- proach. Psychosom Med 2005;67:195–199 Severe hypoglycaemia in type 1 diabetes mellitus: tional relationship of depression and diabetes: 36. Schmitt A, Reimer A, Hermanns N, et al. De- underlying drivers and potential strategies for a systematic review. Clin Psychol Rev 2011;31: pression is linked to hyperglycaemia via subopti- successful prevention. Diabetes Metab Res Rev 1239–1246 mal diabetes self-management: a cross-sectional 2014;30:175–190 24. Shao W, Ahmad R, Khutoryansky N, Aagren mediation analysis. J Psychosom Res 2017;94: 11. Barnard KD, Skinner TC, Peveler R. The prev- M, Bouchard J. Evidence supporting an associa- 17–23 alence of co-morbid depression in adults with tion between hypoglycemic events and depres- 37. Musselman DL, Betan E, Larsen H, Phillips LS. type 1 diabetes: systematic literature review. Dia- sion. Curr Med Res Opin 2013;29:1609–1615 Relationship of depression to diabetes types 1 and bet Med 2006;23:445–448 25. Gordon NP, Kaplan GA. Some evidence refut- 2: epidemiology, biology, and treatment. Biol Psy- 12. Gendelman N, Snell-Bergeon JK, McFann K, ing the HMO “favorable selection” hypothesis: chiatry 2003;54:317–329 et al. Prevalence and correlates of depression in the case of Kaiser Permanente. Adv Health Econ 38. Kendler KS, Gardner CO. Depressive vulnera- individuals with and without type 1 diabetes. Di- Health Serv Res 1991;12:19–39 bility, stressful life events and episode onset of abetes Care 2009;32:575–579 26. Krieger N. Overcoming the absence of socio- major depression: a longitudinal model. Psychol 13. Trief PM, Xing D, Foster NC, et al.; T1D Ex- economic data in medical records: validation and Med 2016;46:1865–1874 change Clinic Network. Depression in adults in the application of a census-based methodology. Am J 39. Paykel ES. Life events and affective disorders. T1D Exchange Clinic Registry. Diabetes Care 2014; Public Health 1992;82:703–710 Acta Psychiatr Scand Suppl 2003;61–66 37:1563–1572 27. Gordon, NP. Similarity of the Kaiser Perma- 40. Gonder-Frederick LA, Cox DJ, Bobbitt SA, 14. Gonzalez JS, Peyrot M, McCarl LA, et al. De- nente senior member population in northern Pennebaker JW. Mood changes associated with pression and diabetes treatment nonadherence: California to the non-Kaiser Permanente covered blood glucose fluctuations in insulin-dependent a meta-analysis. Diabetes Care 2008;31:2398– and general population of seniors in northern Cal- diabetes mellitus. Health Psychol 1989;8:45–59 2403 ifornia: statistics from the 2009 California Health 41. Derijks HJ, Heerdink ER, De Koning FH, 15. Bauer AM, Parker MM, Moffet HH, et al. De- Interview Survey. Oakland, CA, Kaiser Perma- Janknegt R, Klungel OH, Egberts AC. The associa- pressive symptoms and adherence to cardiome- nente Northern California Division of Research, tion between antidepressant use and hypoglycae- tabolic therapies across phases of treatment 2012 mia in diabetic patients: a nested case-control among adults with diabetes: the Diabetes Study 28. Karter AJ, Schillinger D, Adams AS, et al. Ele- study. Pharmacoepidemiol Drug Saf 2008;17: of Northern California (DISTANCE). Patient Prefer vated rates of diabetes in Pacific Islanders and 336–344 Adherence 2017;11:643–652 Asian subgroups: the Diabetes Study of Northern 42. Derijks HJ, Meyboom RHB, Heerdink ER, et al. 16. Lustman PJ, Anderson RJ, Freedland KE, de California (DISTANCE). Diabetes Care 2013;36: The association between antidepressant use and Groot M, Carney RM, Clouse RE. Depression and 574–579 disturbances in glucose homeostasis: evidence poor glycemic control: a meta-analytic review of 29. Karter AJ, Ferrara A, Liu JY, Moffet HH, from spontaneous reports. Eur J Clin Pharmacol the literature. Diabetes Care 2000;23:934–942 Ackerson LM, Selby JV. Ethnic disparities in dia- 2008;64:531–538 17. Kikuchi Y, Iwase M, Fujii H, et al. Association betic complications in an insured population. 43.KamplingH,PetrakF,FarinE,KulzerB, of severe hypoglycemia with depressive symp- JAMA 2002;287:2519–2527 Herpertz S, Mittag O. Trajectories of depression toms in patients with type 2 diabetes: the Fu- 30. Klompas M, Eggleston E, McVetta J, Lazarus R, in adults with newly diagnosed type 1 diabetes: kuoka Diabetes Registry. BMJ Open Diabetes Li L, Platt R. Automated detection and classifi- results from the German Multicenter Diabetes Res Care 2015;3:e000063 cation of type 1 versus type 2 diabetes using Cohort Study. Diabetologia 2017;60:60–68