22 Care Volume 38, January 2015

Predictors of Nonsevere and Severe The ORIGIN Trial Investigators* During - Lowering Treatment With Glargine or Standard Drugs in the ORIGIN Trial Diabetes Care 2015;38:22–28 | DOI: 10.2337/dc14-1329 CLIN CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL

OBJECTIVE Hypoglycemia is a leading risk of glucose-lowering therapy. Treatment with insulin glargine compared with standard care early in the course of dysglycemia in the Outcome Reduction with an Initial Glargine Intervention (ORIGIN) trial provides information on the frequency and predictors of hypoglycemia in this setting.

RESEARCH DESIGN AND METHODS A total of 12,537 people with high cardiovascular risk and dysglycemia treated with one or no oral glucose-lowering agents were randomized to add glargine titrated to a fasting glucose level of £5.3 mmol/L (£95 mg/dL) or to use standard therapies. Independent associations of both nonsevere hypoglycemia (symptom- atic and confirmed with a glucose level of £3 mmol/L [£54 mg/dL]) and severe hypoglycemia with characteristics at baseline, treatment allocation, and average

HbA1c were assessed by Cox proportional hazards models.

RESULTS During a median follow-up period of 6.2 years, 28% of participants reported non- severe hypoglycemia, and 3.8% reported severe hypoglycemia. Prior use of a and allocation to glargine independently predicted a higher risk for Corresponding author: Matthew C. Riddle, riddle@ both categories of participants. Nonsevere events were independently associated ohsu.edu. with younger age, lower BMI, the presence of diabetes, and higher baseline HbA1c Received 26 May 2014 and accepted 1 October level. Severe events were associated with older age, hypertension, higher serum 2014. creatinine level, and lower cognitive function, but not baseline glycemic status. Clinical trial reg. no. NCT 000069784, clinicaltrials Progressively higher on-treatment HbA1c level was associated with a lower risk of .gov. nonsevere events in both treatment groups; a lower risk of severe events in the This article contains Supplementary Data online glargine group, and a higher risk of severe events with standard care. at http://care.diabetesjournals.org/lookup/ suppl/doi:10.2337/dc14-1329/-/DC1. CONCLUSIONS *The members of the Writing Committee for the Hypoglycemia was relatively uncommon in the ORIGIN trial, but was more fre- ORIGIN Trial Investigators are listed in the APPENDIX. quent with sulfonylurea use at baseline and allocation to glargine. Nonsevere and © 2015 by the American Diabetes Association. Readers may use this article as long as the work severe events were associated with different clinical characteristics, awareness of is properly cited, the use is educational and not which may guide individualized therapy. for profit, and the work is not altered. care.diabetesjournals.org The ORIGIN Trial Investigators 23

Hypoglycemia is an important adverse who were $50 years of age and had one episode of hypoglycemia was esti- effect of treatment with glucose-lowering levels, im- mated as the number of people having agents such as insulin and . paired glucose tolerance, or previously experienced one or more episodes Although only a subset of patients to known or recently detected type 2 dia- divided by the total number of person- whom these agents are prescribed expe- betes were enrolled in the study be- years of observation. The annualized in- rience hypoglycemia, these episodes tween 2003 and 2005. Those with a cidence of all hypoglycemic events (e.g., limit the level of glycemic control that prior diagnosis of diabetes could be with an individual counted twice if they can be achieved, affect quality of life, taking no more than one oral glucose- had experienced two events, three and may lead to loss of consciousness lowering agent together with a lifestyle times if they had experienced three and hospitalization. Moreover, epidemi- intervention. Participants were random- events, etc.) was estimated as the total ologic studies have shown that people ized to either add insulin glargine to number of events divided by the num- who experience hypoglycemic events prior treatment, with titration of the ber of person-years of follow-up. For are also at increased risk of a wide range dosage seeking a fasting plasma glucose each individual who had experienced of adverse health consequences includ- level of #5.3 mmol/L (#95 mg/dL), or to at least one hypoglycemic event, the an- ing cardiovascular (CV) events, cancers, use standard glycemic care beginning nualized incidence of all hypoglycemic and death (1–5). Whether nonsevere with oral therapies according to local events during the study divided by the and severe hypoglycemia occur in similar guidelines. Prior glucose-lowering thera- number of person-years of follow-up settings and require similar preventive pies were usually continued after the initi- contributed by that individual was esti- tactics is unclear. These observations ation of randomized treatment. Usage of mated and the mean (SD) and median highlight the importance of identifying specific glucose-lowering agents at study (interquartile range) was calculated. Con- clinical characteristics that are associ- enrollment and after randomization has tinuous variables were expressed as ated with a higher risk of either nonse- been reported previously (9). Participants means and SDs or medians and interquar- vere or severe hypoglycemia, so that were also randomized to take either 1 g of tile ranges, and were compared using t susceptible individuals can be identified n-3 fatty acids daily or a placebo. The main tests or Kruskal-Wallis tests as appropri- and provided with strategies to reduce end point of the comparison of glargine ate. Baseline categorical variables were the risks of both hypoglycemic events with standard care was a composite of expressed as numbers and percentages, and their potential consequences. nonfatal myocardial infarction, stroke, or and were compared using x2 tests. The Outcome Reduction with an Ini- CV death. Cox proportional hazards models were tial Glargine Intervention (ORIGIN) trial used to estimate the unadjusted hazard compared treatment with titrated basal Hypoglycemic Events of confirmed nonsevere and severe hy- During a median on-treatment follow-up insulin glargine targeting normal fasting poglycemic events for 33 baseline period of 6.2 years (interquartile range glucose levels with standard oral agent– demographic, medical history, examina- 5.8–7.2 years), participants kept diaries based therapy for a median period of 6.2 tion, medication use, and laboratory recording symptoms suggesting hypo- years, in a population of 12,537 people characteristics (which have been linked glycemia and self-measured plasma- with dysglycemia and additional CV risk to hypoglycemia in other epidemiologic referenced glucose values at the time of factors (6,7). As previously reported, the studies), as well as baseline treatment symptoms. These were transcribed at glargine intervention had a neutral ef- allocation and the updated HbA level clinic visits at 0.5, 1, 2, and 4 months, 1c fect on CV and cancer outcomes (7). Al- (estimated using the mean of all of the and every 4 months thereafter. Nonse- though absolute rates were low in both available HbA levels from the time of vere confirmedhypoglycemiawas asymp- 1c treatment groups, hypoglycemia was randomization until the end of the tomatic event with concurrent glucose more common with glargine treatment study). All hazard ratios (HRs) were cal- measurement of #3.0 mmol/L (#54 than with standard care (7,8). Because culatedusingafrailty model that ac- mg/dL), but not requiring assistance by both nonsevere and severe hypoglyce- counts for the multiplicity of events another person. Severe hypoglycemia mic events were carefully ascertained and the propensity of experiencing a was defined as requiring assistance, with during long-term treatment, the ORIGIN second event in a person who has al- either prompt recovery after oral admin- trial provides a unique opportunity to ready experienced one event (10). Vari- istration of a carbohydrate, intravenous identify clinically important baseline ables for which the univariate HR was glucose, or glucagon, or documentation risk factors for both categories of hypo- associated with a P value of ,0.1 were by self-measurement or laboratory mea- glycemia, and to assess the independent then included in multivariable analyses surements of plasma glucose levels of effects of allocation to glargine and on- using the frailty model. The effect of a #2.0 mmol/L (#36 mg/dL). treatment HbA1c levels on such events. diagnosis of diabetes prior to participa- Statistical Analysis tion in the ORIGIN trial and of the up- RESEARCH DESIGN AND METHODS All incidence rates were expressed as dated HbA1c level was assessed by Study Design and Participants the number of incidents per 100 per- estimating multivariable frailty models As previously reported, the ORIGIN trial son-years of follow-up. To determine with and without these variables. was a large, randomized, 2 3 2factorial the yearly incidence of hypoglycemia, Whether the relationship between up- design trial that tested the effects of two the number of person-years of exposure dated HbA1c levels and the risk of hypo- pairs of treatments on CV outcomes for each participant was calculated as glycemiadifferedinparticipants in people with high CV risk and dysgly- the number of years until the first hypo- allocated to receive insulin glargine ver- cemia (6,7). People from 40 countries glycemic event. The incidence of at least sus standard care was assessed by 24 Nonsevere and Severe Hypoglycemia in ORIGIN Diabetes Care Volume 38, January 2015

adding an interaction term (updated did not require third-party assistance (P , 0.008). A similar analysis for severe HbA1c*allocation) to the model. The re- (nonsevere event) was reported for hypoglycemia (Table 1) revealed that lationship between the independent 3,518 participants (28%). Larger propor- the 472 participants affected reported variables and the risk of a hypoglycemic tions of participants randomized to re- 591 events. While 359 participants allo- event was expressed as an HR and its ceive glargine than standard care cated to glargine reported at least one 95% CI. experienced nonsevere events (42% severe event, 113 of those receiving Kaplan-Meier plots illustrating the cu- [2,614/6,264] vs. 14% [904/6,273]; standard care did so. Of participants mulative incidence of the first nonse- mean yearly proportions 9.8% vs. with and without diabetes at baseline, vere hypoglycemic event in people 2.7%). Also, more participants with dia- 436 and 36, respectively, had experi- allocatedtoreceiveglargineversus betes at baseline experienced nonsevere enced severe events. Among those par- standard care, and in people with a his- events compared with those without di- ticipants who had experienced at least tory of diabetes versus no diabetes were abetes (30% [3,296/11,081] vs. 15% one severe event, the yearly rates were plotted. Plots illustrating the relation- [222/1,456]; mean yearly proportions 0.3 and 0.2 (P = 0.8) for glargine versus ship between the incidence of hypogly- 6.3% vs. 2.8%). Patterns were similar standard treatment, and also 0.3 vs. 0.2 cemic events and the updated HbA1c for severe events. A total of 472 partic- (P = 0.3) for diabetes versus no diabetes. level were estimated from a multivari- ipants (3.8%) had experienced at least able model based on just the glargine one severe hypoglycemic episode, with Unadjusted Associations of participants and another model based the proportions of participants affected Hypoglycemia With Baseline on just the standard care participants. yearly being 1.00% vs. 0.31% for glargine Characteristics People with both nonsevere and severe For each group, the HRs per HbA1c in- versus standard care, and 0.68% vs. events were more likely to be using a crement from the model were used to 0.42% for individuals with or without di- sulfonylurea and to have higher HbA generate an estimated incidence of hy- abetes at baseline. The time course of 1c levels at baseline. Those who had expe- poglycemia at each updated HbA1c level the appearance of the first nonsevere rienced nonsevere events generally with reference to the observed rate of and first severe events during treatment hypoglycemia at the median updated were younger, whereas those who had is shown in Fig. 1. The frequency of new experienced severe events were older. A HbA1c level, and expressed as the per- events was consistently higher with glar- centage per year. All statistical analyses complete listing of unadjusted associa- gine treatment and for people with di- tions with nonsevere and severe events were performed using SAS version 9.2 abetes versus those without diabetes (SAS Institute), and figures were gener- is shown in Supplementary Table 1. during the whole period of follow-up. ated using SPlus version 8.1.1 (TIBCO). The 3,518 participants with at least Multivariable Models one confirmed nonsevere event had a The results of Cox proportional hazards RESULTS total of 24,680 events during the study, frailty models (which account for multi- Frequency and Time Course of with a mean (SD) of 1.2 (2.1) events per ple events in a given individual) describ- Hypoglycemic Events person-year (Table 1). For participants ing independent associations of risk As previously reported (7), 5,155 of allocated to treatment with glargine with various clinical factors for each cat- 12,537 randomized participants (41%) vs. standard care, the annualized fre- egory of hypoglycemia are shown in experienced at least one symptomatic quency of events was 1.3 vs. 0.8 (P , Table 2. The models included baseline hypoglycemic event. As noted in Table 1, 0.001) yearly, respectively. Similarly cal- characteristics selected for having uni- at least one hypoglycemic event that was culated yearly rates for diabetes versus variate associations with events with a confirmed by glucose measurement but no diabetes at baseline were 1.2 vs. 0.8 P value of ,0.1, and also allocation to

Table 1—Nonsevere and severe hypoglycemic episodes All Glargine Standard care Diabetes No diabetes (N = 12,537) (N =6,264) (N =6,273) P (N = 11,081) (N =1,456) P Nonsevere (confirmed) Total person-years 60,285 26,588 33,697 52,372 7,913 People with $1 incidents, N/100 py 35,18 (5.8) 2,614 (9.8) 904 (2.7) ,0.001 3,296 (6.3) 222 (2.8) ,0.001 Events, N/py 24,680 (0.41) 20,639 (0.78) 4,041 (0.12) ,0.001 23,537 (0.45) 1,143 (0.14) ,0.001 Mean, N/person/py (SD)* 1.2 (2.1) 1.3 (2.3) 0.8 (1.2) ,0.001 1.2 (2.2) 0.8 (1.1) 0.008 Median, N/person/py (IQR)* 0.47 (0.17–1.19) 0.52 (0.18–1.38) 0.34 (0.17–0.81) ,0.001 0.47 (0.17–1.20) 0.44 (0.17–0.93) 0.032 Severe Total person-years 72,580 35,809 36,770 63,969 8,611 People with $1 incidents, N/100 py 472 (0.7) 359 (1.0) 113 (0.3) ,0.001 436 (0.7) 36 (0.4) 0.004 Events, N/py 591 (0.008) 457 (0.013) 134 (0.004) ,0.001 549 (0.009) 42 (0.005) ,0.001 Mean, N/person/py (SD)* 0.2 (0.3) 0.3 (0.3) 0.2 (0.2) 0.8 0.3 (0.3) 0.2 (0.1) 0.3 Median, N/person/py (IQR)* 0.17 (0.15–0.26) 0.16 (0.15–0.26) 0.17 (0.16–0.24) 0.18 0.17 (0.15–0.27) 0.16 (0.15–0.19) 0.11 IQR, interquartile range; py, person-year(s). *The mean and median number of events per year in individuals with at least 1 event during the period of study observation. care.diabetesjournals.org The ORIGIN Trial Investigators 25

Figure 1—Cumulative incidence of hypoglycemia. The proportions of participants with nonsevere (A and B) and severe (C and D) hypoglycemic events by allocation to glargine vs. standard care (A and C) and diabetes vs. no diabetes at baseline (B and D) are shown at yearly intervals. Numbers of measurements at each time point for the glargine treatment group (Gl) and standard care group (SC), diabetes group (DM+), and no-diabetes group (DM2) are shown at the top of each panel. glargine treatment versus standard years of education, hypertension, lower with hypoglycemia, separate fully ad- care. Both use of a sulfonylurea at study triglyceride and higher creatinine levels, justed models were generated for each enrollment and allocation to glargine and greater impairment of cognitive treatment group and used to estimate treatment were independently associ- function. Similar predictors were noted curves across the range of on-treatment ated with an increased risk of both non- in a supplemental analysis in which all HbA1c values for both nonsevere and se- severe and severe events. Specifically, univariate risk factors for either nonse- vere events (Fig. 2). Progressively higher HRs for nonsevere hypoglycemia were vere or severe hypoglycemia were in- on-treatment HbA1c values were associ- 2.07 (95% CI 1.78–2.40) for sulfonylurea cluded in a single multivariable model ated with a lower risk of nonsevere use and 4.53 (4.00–5.13) for glargine al- that was run to identify independent events in both treatment groups; a location, while for severe events they risk factors for each category of hypo- lower risk of severe events in the glar- were 1.35 (1.07–1.70) and 3.57 (2.80– glycemia (Supplementary Table 2). In gine group; and a higher risk of severe 4.55), respectively, for sulfonylurea use this unified model, nonsevere events events in the standard care group. and glargine allocation. Additional inde- themselves were independently associ- pendent predictors of hypoglycemia de- ated with severe hypoglycemia (HR 2.39 CONCLUSIONS pended on whether nonsevere or severe [95% CI 1.88–3.04], P , 0.001). These analyses provide a detailed de- events were being analyzed. For nonse- scription of the determinants of hypo- vere hypoglycemia, independent pre- Hypoglycemic Risk at Different Levels glycemia during a median follow-up dictors included younger age, South of On-Treatment HbA1c period of 6.2 years in people who at Asian ethnicity, depression, lower BMI, Ahighlysignificant interaction between baseline had moderately elevated glu- higher LDL level, diabetes at study en- updated HbA1c level and treatment allo- cose levels while taking only 0 or 1 oral rollment, use, and higher cation was noted for both nonsevere glucose-lowering drugs. They show that, baseline HbA1c level. In contrast, inde- (P = 0.004) and severe (P , 0.001) hy- regardless of the assigned therapy, the pendent predictors of severe hypoglyce- poglycemia. To further explore the re- absolute incidence of hypoglycemia of mia included older age, no more than 12 lationship of on-treatment HbA1c level any sort was low. Thus, despite a 26 Nonsevere and Severe Hypoglycemia in ORIGIN Diabetes Care Volume 38, January 2015

Table 2—Adjusted hazard of significant risk factors for confirmed nonsevere and the standard care group with respect to significant risk factors for severe hypoglycemia nonsevere hypoglycemia. However, the Nonsevere Severe relationship during standard care was reversed for severe hypoglycemia, with Variable HR (95% CI) P HR (95% CI) P higher achieved HbA1c levels associated Age (per year) 0.98 (0.98–0.99) ,0.001 1.04 (1.03–1.06) ,0.001 with a higher risk of severe hypoglyce- Ethnic South Asian 0.63 (0.42–0.95) 0.03 Not included mia. Such a phenomenon has been Education for 9–12 years Not included 1.81 (1.39–2.36) ,0.001 noted in other studies (12–14) and Hypertension Not included 1.51 (1.14–2.00) 0.004 may reflect an increased susceptibility Depression 1.28 (1.11–1.48) ,0.001 Not included to hypoglycemia due to sulfonylurea Allocation to receive use or (in ,10% of the standard care glargine 4.53 (4.00–5.13) ,0.001 3.57 (2.80–4.55) ,0.001 participants in the ORIGIN trial) use of Metformin 1.24 (1.06–1.44) 0.007 Not included other than glargine (9) among Sulfonylurea 2.07 (1.78–2.40) ,0.001 1.35 (1.07–1.70) 0.01 those people who did not easily main- BMI (per kg/m2) 0.97 (0.96–0.98) ,0.001 0.99 (0.97–1.01) 0.49 tain good glycemic control.

Baseline HbA1c Finally, nonsevere hypoglycemia itself level (per %) 1.24 (1.14–1.35) ,0.001 1.10 (0.94–1.28) 0.24 was an independent predictor of the oc- LDL level 1.12 (1.01–1.24) 0.03 Not included currence of severe hypoglycemia, a finding Triglyceride level Not included 0.84 (0.75–0.94) 0.003 that is not unexpected. It is unknown Creatinine level whether this association is due more to (per mg/dL) Not included 1.01 (1.01–1.02) ,0.001 the effects of unmeasured covariates, eGFR Not included 1.01 (1.00–1.01) 0.15 such as behavioral factors leading to glyce- Mini-Mental State Exam 1.00 (0.98–1.02) 0.88 0.96 (0.93–0.99) 0.02 mic variability, or to hypoglycemia-induced Prior diabetes 1.52 (1.21–1.92) ,0.001 1.25 (0.85–1.83) 0.25 unawareness of falling glucose levels. Fur-

Updated HbA1c level (per %) 0.85 (0.77–0.93) ,0.001 0.98 (0.82–1.16) 0.79 ther study of the relationships between Analyses for conformed nonsevere and severe hypoglycemia are based on separate frailty the occurrence of nonsevere hypoglyce- models that included independent variables with univariate P , 0.1 for each type of mia and subsequent severe hypoglycemia hypoglycemia in supplementary material. eGFR, estimated glomerular filtration rate. or its medical consequences is warranted. These findings add to a growing body of evidence suggesting that, rather than threefold higher rate of hypoglycemia, diabetes, a higher baseline HbA1c level, just representing two parts of a range of the annual incidences of confirmed and a lower HbA1c level achieved during below-normal glucose levels, hypogly- nonsevere and severe hypoglycemia in therapy were independently associated cemic events requiring or not requiring people allocated to insulin glargine– with nonsevere hypoglycemia. In con- third-party assistance tend to occur in mediated fasting normoglycemia were trast, the independent determinants of different situations, have differing impli- only 10% and 1% respectively. Non- severe hypoglycemia were older age, cations for clinical management, and severe hypoglycemia was more likely limited education, hypertension, renal should be evaluated separately. Nonse- to occur in people with diabetes than insufficiency, and mild cognitive impair- vere events can occur when otherwise in those without diabetes, regardless ment. These observations are also con- healthy people with type 2 diabetes are of whether normal fasting glucose levels sistent with those of other reports seeking nearly normal levels of glycemic were sought with basal insulin adminis- (5,13–17). Stated in clinically relevant control, especially with sulfonylurea or tration. The fact that the rates of hypo- terms, these analyses suggest that an insulin treatment. Although they did not glycemia were even lower among individual with a BMI of 25 kg/m2 would independently predict an increased risk people without diabetes than those be ;30% more likely to have nonsevere of CV events in the ORIGIN trial (8), they with diabetes suggests that, in addi- hypoglycemia than one with a BMI of may interfere with quality of life and the tion to promoting , diabe- 35 kg/m2, but with no difference in the attainment of glycemic treatment goals. tes may impair the ability to avoid risk of severe hypoglycemia. Also, an in- In contrast, because severe hypoglyce- hypoglycemia. dividual who is 70 years old would be mia is more likely among older people Allocation to treatment with glargine predicted to have ;40% less risk of non- and individuals with established medical and the use of a sulfonylurea at baseline severe hypoglycemia but 80% higher risk problems who are attempting to main- were independent risk factors for both of severe hypoglycemia compared with a tain excellent glycemic control, more nonsevere and severe hypoglycemia. person who is 50 years of age. conservative glucose-lowering regimens However, additional risk factors for non- Of note is the relationship between and targets should be considered for severe and severe hypoglycemia clearly the HbA1c level achieved during the trial, this population. The finding that severe differed, with no independent predic- and the risk of nonsevere and severe hypoglycemia during standard care was tors, other than glargine assignment events with the two forms of treatment. more likely when the HbA1c level was and prior use of a sulfonylurea, shared In the glargine group, lower achieved above the usual target levels, together by the two categories. Thus, consistent HbA1c levels were associated with a with the prior observation (8) that se- with other reports (11,12), younger age, higher risk of both types of hypoglyce- vere hypoglycemia in the ORIGIN trial lower BMI, depression, a history of mia. A similar relationship was noted for was two to three times more strongly care.diabetesjournals.org The ORIGIN Trial Investigators 27

Figure 2—Relation of risk of hypoglycemia to on-treatment HbA1c, adjusted incidence (percent per year) of nonsevere hypoglycemia (A) and severe hypoglycemia (B) by allocation to glargine (solid blue lines) or standard care (solid red lines) over the observed range of updated on-treatment HbA1c values, with 95% CIs (broken lines). linked to subsequent CV events in peo- severe hypoglycemia. The randomized McMaster University and Hamilton Health ple allocated to receive standard care treatment comparison permitted evalu- Sciences, Hamilton, Ontario, Canada); Linda rather than glargine therapy, suggests ation of two forms of treatment. Among G. Mellbin and Lars Ryden´ (Department of that severe hypoglycemia identifies a the limitations of the study are the post Medicine Solna, Karolinska Institute, Stock- highly vulnerable population at risk for hoc design of the analytic plan, the likeli- holm, Sweden); Julio Rosenstock (Dallas Di- serious medical problems. Similar ob- hood that some hypoglycemic events as- abetes and Endocrine Center at Medical City servations from other large outcomes sociated with a glucose level of #3.0 and University of Texas Southwestern trials (4,18–20)inpeoplewithtype2di- mmol/L (#54 mg/dL) were unreported Medical Center, Dallas, TX); and Jeffrey abetes provide further support for this by participants, and the lack of general- Probstfield (Department of Medicine, view. People who have experienced izability to people with type 1 diabetes University of Washington, Seattle, WA). even a single severe hypoglycemic epi- or a longer duration of type 2 diabetes. sode clearly deserve increased atten- The occurrence of glucose levels ,3.9 tion, with the aim of mitigating the risk mmol/L (,70 mg/dL), whether symptom- Acknowledgments. An international steering of both later severe hypoglycemia by atic or not, was not systematically committee designed and conducted the trial, and all data were collected and analyzed by the adjusting glucose-lowering medications ascertained in the ORIGIN trial, and so ORIGIN Project Office located at the Population and of other outcomes, including both the potential significance of such events Health Research Institute in Hamilton, Ontario, CV and non-CV events, by optimizing is not clarified by these findings. Canada. The authors thank Rose Hastings and cardioprotective and other preventive In summary, individuals with early dys- Russell Standley Memorial Trusts for their support therapies. These observations further glycemia enrolled in the ORIGIN trial had in the preparation of the manuscript. Funding. The ORIGIN Trial was funded by Sanofi suggest that the current nomenclature relatively low absolute rates of hypogly- and Pronova BioPharma, Norway, which provided of hypoglycemia, which describes a con- cemia that were nevertheless higher n-3 fatty acid supplements and placebo. tinuum based on glucose levels, might when normal glucose levels were tar- Duality of Interest. M.C.R. reports receiving be reconsidered to better distinguish geted with administration of insulin research grant support from Sanofi,AstraZeneca, between events that interfere with glargine and sulfonylureas. Moreover, Eli Lilly, and Novo Nordisk; honoraria for speaking from Sanofi; and honoraria for consulting from management but do not predict serious nonsevere and severe hypoglycemic Sanofi, AstraZeneca, Eli Lilly, Elcelyx, and Valeritas. outcomes and those that are strongly events differed in ways beyond glucose These dualities of interest have been reviewed linked with adverse outcomes, whether levels and have different implications for and managed by Oregon Health & Science Univer- causally or by association. individualizing therapy. sity. H.C.G. reports receiving consulting fees fi The strengths of this investigation in- from Sano , Novo Nordisk, Lilly, Bristol-Myers Squibb, Roche, AstraZeneca, and GlaxoSmithKline; clude the large and geographically di- Appendix lecture fees from Sanofi and Bayer; and support verse population studied, the median Writing Committee for the ORIGIN Trial for research or continuing education through follow-up period of .6 years, and (un- Investigators. Matthew C. Riddle (Depart- his institution from Sanofi, Lilly, Takeda, Novo like most prior studies of hypoglycemia) ment of Medicine, Oregon Health & Science Nordisk, Boehringer Ingelheim, Bristol-Myers fi Squibb, and AstraZeneca. H.J. declares no poten- the prospectively de ned data collec- University, Portland, OR); Hertzel C. Gerstein tial conflicts of interest relevant to this article. tion that provided enough events to al- and Hyejung Jung (Department of Medicine L.G.M. reports receiving research grant support low the comparison of nonsevere and and Population Health Research Institute, from the Swedish Heart Lung Foundation, the 28 Nonsevere and Severe Hypoglycemia in ORIGIN Diabetes Care Volume 38, January 2015

Swedish Diabetes Association, Merck, Bayer AG, Sessions of the American Diabetes Association, 11. Karl DM, Gill J, Zhou R, Riddle MC. Clinical and Sanofi; and lecture honoraria from Merck, San Francisco, CA, 13–17 June 2014. predictors of risk of hypoglycaemia during addi- Sanofi, Novartis, Bayer Schering, AstraZeneca, tion and titration of insulin glargine for type 2 Roche, and Eli Lilly. L.R. reports receiving References diabetes mellitus. Diabetes Obes Metab 2013; – research grant support from the Swedish 1. Zoungas S, Patel A, Chalmers J, et al.; 15:622 628 Heart Lung Foundation, the Swedish Diabe- ADVANCE Collaborative Group. Severe hypogly- 12. Miller CD, Phillips LS, Ziemer DC, Gallina DL, tes Association, Roche, and Bayer AG; and Cook CB, El-Kebbi IM. Hypoglycemia in patients cemia and risks of vascular events and death. lecture/consulting honoraria from Sanofi, with type 2 diabetes mellitus. Arch Intern Med N Engl J Med 2010;363:1410–1418 Bristol-Myers Squibb, AstraZeneca, and Merck. 2001;161:1653–1659 2. Frier BM, Schernthaner G, Heller SR. Hypo- J.R. reports receiving consulting fees from 13. Leese GP, Wang J, Broomhall J, et al.; glycemia and cardiovascular risks. Diabetes Care Sanofi, Novo Nordisk, Eli Lilly, GlaxoSmithKline, DARTS/MEMO Collaboration. Frequency of 2011;34(Suppl. 2):S132–S137 Takeda, Merck, Daiichi Sankyo, Janssen, Novartis, severe hypoglycemia requiring emergency treat- 3. Yakubovich N, Gerstein HC. Serious cardio- Boehringer Ingelheim, MannKind, Halozyme, ment in type 1 and type 2 diabetes: a population- Intarcia, and Lexicon; and research grants from vascular outcomes in diabetes: the role of hy- based study of health service resource use. – Merck, Pfizer, Sanofi, Novo Nordisk, Roche, poglycemia. Circulation 2011;123:342 348 Diabetes Care 2003;26:1176–1180 Bristol-Myers Squibb, Eli Lilly, GlaxoSmithKline, 4. Mellbin LG, Malmberg K, Waldenstrom A, 14. Davis TME, Brown SGA, Jacobs IG, Bulsara Takeda, Novartis, AstraZeneca, Amylin, Janssen, Wedel H, Ryden L. for the DIGAMI 2 investiga- M, Bruce DG, Davis WA. Determinants of severe Daiichi Sankyo, MannKind, Boehringer Ingelheim, tors. Prognostic implications of hypoglycaemic hypoglycemia complicating type 2 diabetes: the Intarcia, and Lexicon. J.P. reports receiving episodes during hospitalization for myocardial Fremantle diabetes study. J Clin Endocrinol honoraria for consulting from Sanofi and infarction in patients with type 2 diabetes: a re- Metab 2010;95:2240–2247 research grant support from Sanofi, Bristol-Myers port from the DIGAMI 2 trial. Heart 2009;95: 15. Miller ME, Bonds DE, Gerstein HC, et al.; Squibb, and AstraZeneca. No other potential 721–727 ACCORD Investigators. The effects of baseline conflicts of interest relevant to this article were 5. Kong APS, Yang X, Luk A, et al. Severe hypo- characteristics, glycaemia treatment approach, reported. glycemia identifies vulnerable patients with and glycated haemoglobin concentration on the These potential conflicts of interest have been type 2 diabetes at risk for premature death risk of severe hypoglycaemia: post hoc epidemi- reviewed and managed by Oregon Health & Sci- and all-site cancer: the Hong Kong diabetes reg- ological analysis of the ACCORD study. BMJ ence University. The authors acknowledge the istry. Diabetes Care 2014;37:1024–1031 2010;340:b5444 collaborative relationship with Sanofi and the 6. Gerstein H, Yusuf S, Riddle MC, Ryden L, 16. Whitmer RA, Karter AJ, Yaffe K, Population Health Research Institute, which Bosch J; Origin Trial Investigators. Rationale, de- Quesenberry CP Jr, Selby JV. Hypoglycemic epi- made the study and this analysis possible. sign, and baseline characteristics for a large in- sodes and risk of dementia in older patients Author Contributions. The current analyses ternational trial of cardiovascular disease with type 2 diabetes mellitus. JAMA 2009;301: were conceived and executed by the Writing prevention in people with dysglycemia: the 1565–1572 Committee. M.C.R. helped to lead the design of ORIGIN Trial (Outcome Reduction with an Initial 17. Bruderer SG, Bodmer M, Jick SS, Bader G, the analyses, had access to the results of the Glargine Intervention). Am Heart J 2008;155: Schlienger RB, Meier CR. Incidence of and risk fi analyses, wrote the rst draft of the manuscript, 26–32, e1–e6 factors for severe hypoglycaemia in treated d contributed to revisions of the manuscript, and 7. Gerstein HC, Bosch J, Dagenais GR, et al.; type 2 diabetes patients in the UK anested helped to make the decision to submit it for ORIGIN Trial Investigators. Basal insulin and car- case-control analysis. Diabetes Obes Metab 2014;16:801–811 publication. H.C.G., L.G.M., and L.R. helped to diovascular and other outcomes in dysglycemia. lead the design of the analyses, had access 18. Bonds DE, Miller ME, Bergenstal RM, et al. N Engl J Med 2012;367:319–328 to the results of the analyses, contributed to The association between symptomatic, severe 8. Mellbin LG, Ryden´ L, Riddle MC, et al.; revisions of the manuscript, and helped to make hypoglycaemia and mortality in type 2 diabetes: ORIGIN Trial Investigators. Does hypoglycaemia the decision to submit it for publication. H.J. retrospective epidemiological analysis of the increase the risk of cardiovascular events? A re- performed the analyses, had access to the ACCORD study. BMJ 2010;340:b4909 results of the analyses, contributed to revisions port from the ORIGIN trial. Eur Heart J 2013;34: 19. Seaquist ER, Miller ME, Bonds DE, et al.; – of the manuscript, and helped to make the 3137 3144 ACCORD Investigators. The impact of frequent decision to submit it for publication. J.R. and J.P. 9. ORIGIN Trial Investigators. Characteristics as- and unrecognized hypoglycemia on mortality in , had access to the results of the analyses, sociated with maintenance of mean A1C 6.5% the ACCORD study. Diabetes Care 2012;35:409– contributed to revisions of the manuscript, and in people with dysglycemia in the ORIGIN trial. 414 helped to make the decision to submit it for Diabetes Care 2013;36:2915–2922 20. Riddle MC, Ambrosius WT, Brillon DJ, et al.; publication. M.C.R. is the guarantor of this work 10. Frick U, Frick H, Langguth B, Landgrebe M, Action to Control Cardiovascular Risk in Diabe- and, as such, had full access to all the data in the Hubner-Liebermann¨ B, Hajak G. The revolving tes Investigators. Epidemiologic relationships study and takes responsibility for the integrity of door phenomenon revisited: time to readmis- between A1C and all-cause mortality during a the data and the accuracy of the data analysis. sion in 17’145 [corrected] patients with 37’697 median 3.4-year follow-up of glycemic treat- Prior Presentation. Parts of this study were hospitalisations at a German psychiatric hospi- ment in the ACCORD trial. Diabetes Care 2010; presented in abstract form at the 74th Scientific tal. PLoS One 2013;8:e75612 33:983–990