Content uploaded by Elbert S Huang
Author content
All content in this area was uploaded by Elbert S Huang on Apr 26, 2016
Content may be subject to copyright.
Incidence of Remission in Adults
With Type 2 Diabetes: The
Diabetes & Aging Study
Diabetes Care 2014;37:3188–3195 |DOI: 10.2337/dc14-0874
OBJECTIVE
To estimate the incidence of remission in adults with type 2 diabetes not treated
with bariatric surgery and to identify variables associated with remission.
RESEARCH DESIGN AND METHODS
We quantified the incidence of diabetes remission and examined its correlates among
122,781 adults with type 2 diabetes in an integrated healthcare delivery system.
Remission required the absence of ongoing drug therapy and was defined as follows:
1) partial: at least 1 year of subdiabetic hyperglycemia (hemoglobin A
1c
[HbA
1c
] level
5.7–6.4% [39–46 mmol/mol]); 2) complete: at least 1 year of normoglycemia (HbA
1c
level <5.7% [<39 mmol/mol]); and 3) prolonged: complete remission for at least 5
years.
RESULTS
The incidence density (remissions per 1,000 person-years; 95% CI) of partial,
complete, or prolonged remission was 2.8 (2.6–2.9), 0.24 (0.20–0.28), and 0.04
(0.01–0.06), respectively. The 7-year cumulative incidence of partial, complete, or
prolonged remission was 1.47% (1.40–1.54%), 0.14% (0.12–0.16%), and 0.007%
(0.003–0.020%), respectively. The 7-year cumulative incidence of achieving any
remission was 1.60% in the whole cohort (1.53–1.68%) and 4.6% in the subgroup
with new-onset diabetes (<2 years since diagnosis) (4.3–4.9%). After adjusting for
demographic and clinical characteristics, correlates of remission included age >65
years, African American race, <2 years since diagnosis, baseline HbA
1c
level <5.7%
(<39 mmol/mol), and no diabetes medication at baseline.
CONCLUSIONS
In community settings, remission of type 2 diabetes does occur without bariatric
surgery, but it is very rare.
It is widely believed that type 2 diabetes is a chronic progressive condition, which at
best can be controlled, but never cured (1), and that once treatment with glucose-
lowering medication is initiated, it is required indefinitely and is intensified over time
(2,3). However, a growing body of evidence from clinical trials and case-controlstudies
(4–6) has reported the remission of type 2 diabetes in certain populations, most
notably individuals who received bariatric surgery. A post hoc analysis from the Action
for Health in Diabetes (Look AHEAD) study, a randomized, controlled trial of an in-
tensive lifestyle intervention among adults with type 2 diabetes, found evidence of
remission in both the intensive lifestyle management, and the diabetes support and
education groups, suggesting that metabolic/bariatric surgery is not the only pathway
1
Division of Research, Kaiser Permanente
Northern California, Oakland, CA
2
Department of Health Policy, School of Public
Health and Health Services, The George Wash-
ington University, Washington, DC
3
Evolent Health, Arlington, VA
4
Department of Medicine, The University of
Chicago Medical Center, Chicago, IL
5
Chicago Center for Diabetes Translation Research,
The University of Chicago, Chicago, IL
6
Center for Translational and Policy Research
of Chronic Disease, The University of Chicago,
Chicago, IL
Corresponding author: Andrew J. Karter, andy.j.
karter@kp.org.
Received 7 April 2014 and accepted 25 August
2014.
This article contains Supplementary Data online
at http://care.diabetesjournals.org/lookup/
suppl/doi:10.2337/dc14-0874/-/DC1.
© 2014 by the American Diabetes Association.
Readers may use this article as long as the work
is properly cited, the use is educational and not
for profit, and the work is not altered.
Andrew J. Karter,
1
Shantanu Nundy,
2,3
Melissa M. Parker,
1
Howard H. Moffet,
1
and Elbert S. Huang
4,5,6
3188 Diabetes Care Volume 37, December 2014
EPIDEMIOLOGY/HEALTH SERVICES RESEARCH
to remission and that remission may also
be achieved through intensive lifestyle
management (7). In addition to a return
to normal or near-normal blood glucose
levels in the absence of glucose-lowering
treatment, these individuals have been
found to have biologic evidence of remis-
sion, including normalization of b-cell
function and hepatic insulin sensitivity
(8,9).
Recently, the American Diabetes As-
sociation (ADA) convened an expert
panel to define the term “cure of diabe-
tes”(10). The ADA consensus statement
borrowed from the cancer literature in
defining the term “remission”as achiev-
ing glycemia below the diabetic range in
the absence of active pharmacologic or
surgical therapy. It defines the following
three mutually exclusive states of remis-
sion: partial remission, which is subdia-
betic hyperglycemia of at least 1 year
(A1C level not diagnostic of diabetes
[5.7–6.4%; 39–46 mmol/mol], fasting
glucose level 100–125 mg/dL [5.6–6.9
mmol/L]); complete remission, which
is normoglycemia of at least 1 year
(A1C level in the normal range [,5.7%;
,39 mmol/mol], fasting glucose ,100
mg/dL [5.6 mmol/L]); and prolonged re-
mission (or “cure”), complete remission
of at least 5 years.
Despite the clinical relevance and im-
portance of remission, little is known
about the incidence of remission in com-
munity settings (11,12). Studies to date
have focused largely on remission after
gastricbypassorreliedondatafrom
clinical trials, which have limited gener-
alizability. Therefore, we conducted a
retrospective cohort study to describe
the incidence rates and variables associ-
ated with remission among adults with
type 2 diabetes who received usual care,
excluding bariatric surgery, in a large,
ethnically diverse population.
RESEARCH DESIGN AND METHODS
Study Cohort
Kaiser Permanente Northern California
(KPNC) is an integrated healthcare deliv-
ery system that provides comprehensive
medical services to .3 million members.
KPNC membership closely approximates
the general population ethnically and so-
cioeconomically, except for the extremes
of income distribution.
The source population was identified
from the KPNC Diabetes Registry (13,14).
Established in 1993, the registry is updated
annually by identifying all health plan
memberswithdiabetesfromfourauto-
mated databases using a validated algo-
rithm (15), as follows: primary hospital
discharge diagnosis of diabetes; $2out-
patient visit diagnoses of diabetes; any
prescription for a diabetes-specificmedi-
cation; or two abnormal outpatient labo-
ratory results (fasting glucose level $126
mg/dL; random or postchallenge [75 g]
glucose level $200 mg/dL; hemoglobin
A
1c
[HbA
1c
] level $6.5% [$48 mmol/mol])
from tests performed on separate days,
within a 3-year period, after excluding
those identified due to gestational di-
abetes (based on ICD-9 code 648.8) or
cases identifiedbasedonmedications
only where there were competing indica-
tions (e.g., HIV lipodystrophy, polycystic
ovary syndrome). Starting with a sampling
frame that included the 197,699 plan
members in the KPNC Diabetes Registry
as of 1 January 2005, we exclu ded subjects
based on the following criteria: 1)ifthey
lacked continuous KPNC membership with
pharmacy benefits since 1 January 2004
(no gap of $3months)(n= 12,507); 2)if
they were ,19 years of age as of 1 January
2005 (n= 1,401); 3)iftheyhadtype1di-
abetes (based on self-report or had diabe-
tes onset at ,30 years of age, treated with
insulin only and was never treated with
oral agents) (n= 6,625); 4) if they had
less than two HbA
1c
measurement results
during follow-up (n= 23,310); 5)ifthey
had no diabetes medication dispensed in
the year prior to baseline and had either
no HbA
1c
measurements or all HbA
1c
mea-
surements of ,6.5% (,48 mmol/mol)
(e.g., did not have recent evidence of clin-
ical diabetes such as pharmacologic treat-
ment or a recent HbA
1c
level in the
diabetic range) (n= 27,271); 6) if they
had a history of bariatric surgery, includ-
ing gastric bypass, sleeve gastrectomy, or
lap banding, or had undergone any of
these procedures during the study period
(see Supplementary Table 1 for complete
listing) (n= 1,564); or 7) if they had been a
long-termuser (.60 days) of oral steroids
in the year prior to baseline (n=2,240).
The remaining 122,781 eligible subjects
were the focus of our analysis.
Follow-up of this analytic cohort
members was initiated on 1 January
2005, and individuals were censored at
the first occurrence of one of the follow-
ing events: 1)ifanyoftheremission
case definitions were met; 2)ifthere
had been a gap of $3 months in either
membership or prescription benefits; 3)
if death had occurred from any cause; or
4) the end of the study (31 December
2011). We collected data on demo-
graphic characteristics (age, sex, race),
and laboratory data related to glycemic
control, pharmacy data, outpatient di-
agnoses, and inpatient hospitalizations,
including diagnoses and procedures. All
HbA
1c
tests were analyzed at a single
regional laboratory. This high-volume
Kaiser laboratory is licensed by the Cal-
ifornia Department of Health Services,
and is inspected and accredited by the
College of American Pathologists. Since
2002, the regional laboratory has used
the Diabetes Control and Complications
Trial standardization of HbA
1c
imple-
mented by the National Glycohemoglo-
bin Standardization Group. In 2011,
KPNC updated its assay for measuring
HbA
1c
. This change resulted in a slight
increase in test values compared with
the period of 2005–2011 and biased
our study toward more conservative es-
timates of remission. Survival status and
dates of death were captured from
KPNC records and the Death Data Files
from the California Department of Pub-
lic Health.
Case Definition
Our definitions of remission were based
on the 2009 ADA consensus statement
(10). “Partial remission”of diabetes was
defined as having two or more consec-
utive subdiabetic HbA
1c
measurements,
all of which were in the range of 5.7–
6.4% [39–46 mmol/mol] over a period
of at least 12 months. “Complete re-
mission”was defined as having two or
more consecutive normoglycemic HbA
1c
measurements, all of which were ,5.7%
[,39 mmol/mol] over a period of at
least 12 months. “Prolonged remission”
was defined as having two or more con-
secutive normoglycemic HbA
1c
mea-
surements, all of which were ,5.7%
[,39 mmol/mol] over a period of at
least 60 months. Each definition of
remission requires the absence of phar-
macologic treatment during the defined
observation period. Periods of no phar-
macologic therapy were defined as be-
ginning on the last date that a diabetes
medication prescription was filled plus
two times the daily supply last dis-
pensed, and ending when and if a
new diabetes medication prescription
was filled (patients who have had no
care.diabetesjournals.org Karter and Associates 3189
medication dispensed automatically
meet the criteria of having no pharma-
cologic therapy). By definition, patients
who met the criteria for complete re-
mission also met the definition for par-
tial remission. However, the categories
were mutually exclusive and, similar to
the oncologic literature (16), periods of
remission were categorized by the most
advanced category observed even
though they also qualified for a less ad-
vanced stage (i.e., prolonged remission
is more advanced than complete remis-
sion, and complete remission is more
advanced than partial remission). Thus,
patients who met the criteria for partial
remission for at least 1 year who, after
longer follow-up, eventually met the cri-
teria for complete remission were clas-
sified as having complete remission. A
fourth,summary,definition, “Any re-
mission,”was applied to patients who
met any of the above remission category
definitions.
To operationalize the ADA case defi-
nitions of remission (10), we made sev-
eral conservative modifications of the
definitions. We used an HbA
1c
level of
,5.7% [,39 mmol/mol] to define com-
plete and prolonged remission rather
than 6.0% [42 mmol/mol], as some stud-
ies have (17). This approach follows
from reference ranges used by the
KPNC centralized laboratory, which de-
fines normoglycemia as an HbA
1c
level
of ,5.7% [,39 mmol/mol] and predia-
betes as an HbA
1c
level of 5.7–6.4% [39–
46 mmol/mol], and is consistent with
national and international guidelines
for the diagnosis of prediabetes and di-
abetes (18,19). We also required at least
two HbA
1c
measurements at least 12 or
60 months apart for partial/complete
and prolonged remission, respectively
(5,7,20,21). Finally, our definitions
specified a minimum time interval be-
tween measurements rather than a
time interval from the first measure-
ment. We chose only HbA
1c
levels be-
cause fasting blood glucose levels are
rarely used at KPNC after the diagnosis
of diabetes.
Covariates
Using previously published methods
(15), the date that a patient first re-
ceived a diagnosis of diabetes was cal-
culated preferentially from self-report
or else from the date of inclusion in
the diabetes registry if this occurred at
least 6 months after joining KPNC as a
member; in the absence of these condi-
tions, the variable was set to “missing,”
which occurred in 14.2% of individuals.
Other covariates included comorbid
conditions, defined using standard out-
patient and inpatient Diagnosis-Related
Group and ICD-9 codes as well as labo-
ratory findings in the 12 months prior to
baseline (see Supplementary Table 1 for
details). BMI was obtained from the
most recent measurement available,
but because electronic data capture of
BMI was not fully implemented across
all KPNC medical centers until 2006,
these data are missing in 29.7% of par-
ticipants. Diabetes medications in the
year prior to baseline and since 2002
were obtained from outpatient phar-
macy dispensing. Finally, we used a val-
idated, standardized deprivation index
as an indicator of neighborhood-level
socioeconomic status (22,23).
Statistical Analyses
We estimated the incidence density
(number of remissions per 1,000 person-
years at risk) for partial, complete, and
prolonged remission separately, as well
as the incidence density of any re-
mission. In addition, we calculated the
7-year cumulative incidence of any re-
mission (reported as the percentage of
subjects who experienced remission
during a 7-year follow-up) using the
Kaplan-Meier method of estimating
the cumulative probability of an event
(24). Seven years was chosen as the
period for analysis because complete
remission requires at least 5 years of
data. Results were stratified by years
since diagnosis and baseline diabetes
therapy.
Cox proportional hazards models were
specified to identify significant variables
associated with the time to any remission
(partial, complete, or prolonged remis-
sion) (24). We also performed a subgroup
analysis of patients with new-onset dia-
betes (#2 years since diagnosis at base-
line). We examined unadjusted and
adjusted models based on a complete
case–only analysis. The adjusted model
included demographic factors (age at co-
hort entry, sex, race, deprivation index),
baseline risk factors (BMI, estimated glo-
merular filtration rate [eGFR], HbA
1c
level), comorbidities (cardiovascular dis-
ease, hypertension, dyslipidemia, dia-
betic retinopathy, diabetic neuropathy,
congestive heart failure), and diabetes
factors (oral or insulin use, years since di-
abetes diagnosis).
All analyses were conducted using SAS
version 9.3 (SAS Institute Inc., Cary, NC).
All statistical testing was two sided; the
level of significance was ,0.05. The study
was approved by the Institutional Review
Boards of Kaiser Permanente and The
University of Chicago.
RESULTS
As of 1 January 2005, 122,781 individuals
met our study criteria, yielding 709,005
person-years of total follow-up time.
The median duration of follow up in the
overall cohort was 7 years (25th percen-
tile 4.7; 75th percentile 7.0), and in the
new-onset cohort it was 7 years (25th
percentile 4.5; 75th percentile 7.0). A to-
tal of 11,510 subjects were censored be-
cause of having a gap of $3 months in
either membership or prescription bene-
fits within the first 2 years of follow-up.
The average age of participants was 62
years, 47.1% were female, and 51.6%
were nonwhite (Table 1). The mean (SD)
interval between HbA
1c
tests in the remis-
sion group was 256 days (139 days). The
mean interval (SD) between HbA
1c
tests
among patients not in the remission
group was 212 days (118 days). The me-
dian time since the diagnosis of diabetes
in our cohort was 5.9 years, and the av-
erage baseline HbA
1c
level was 7.4% [57
mmol/mol]. The 18,684 individuals
(15.2%) in the subset with new-onset di-
abetes, defined as #2 years since diagno-
sis, were younger, were more likely to
have their diabetes controlled by diet,
and had fewer comorbidities (Table 2).
The incidence densities of partial,
complete, and prolonged remission in
the full cohort were 2.8 (95% CI 2.6–
2.9), 0.24 (95% CI 0.20–0.28), and 0.04
(95% CI 0.01–0.06) cases per 1,000
person-years, respectively (Table 3).
The 7-year cumulative incidences of par-
tial, complete, and prolonged remission
were 1.5% (95% CI 1.4–1.5%), 0.14%
(95% CI 0.12–0.16%), and 0.01% (95%
CI 0.003–0.02%), respectively. The 7-
year cumulative incidence of any remis-
sion decreased with longer time since
diagnosis from a high of 4.6% (95% CI
4.3–4.9%) for individuals diagnosed
with diabetes in the past 2 years to a
low of 0.4% (95% CI 0.3–0.5%) in those
diagnosed .10 years ago. The 7-year
cumulative incidence of any remission
3190 Diabetes Remission Diabetes Care Volume 37, December 2014
was much lower for individuals using in-
sulin (0.05%; 95% CI 0.03–0.1%) or oral
agents (0.3%; 95% CI 0.2–0.3%) at base-
line compared with diabetes patients
not using medication at baseline (12%;
95% CI 12–13%).
In bivariate and multivariable analy-
ses, in the full cohort, any remission
was associated with older age, less
time since diagnosis, lower baseline
HbA
1c
level, and not taking diabetes
medications (P,0.0001) (Table 4).
Higher socioeconomic status, moderate
renal impairment, and no dyslipidemia
were also associated with remission in
bivariate analyses (P,0.0001), but
these associations attenuated in multi-
variable analyses (P,0.05). In addition,
African American race was associated
with remission in the multivariable anal-
ysis only (hazard ratio 1.4; 95% CI 1.15–
1.75; P= 0.001).
In the new-onset cohort, any remis-
sion was also associated with lower
baseline HbA
1c
level, and not taking di-
abetes medications in both bivariate
and multivariable analyses (P,
0.0001) (Supplementary Table 2). In ad-
dition, older age, higher socioeconomic
status, and moderate renal impairment
were associated with remission in bivar-
iate analyses (P,0.0001), but were at-
tenuated after adjustment (P,0.05). In
addition, African American race and no
dyslipidemia were associated with re-
mission only after adjustment.
CONCLUSIONS
In this large cohort of insured adults
with type 2 diabetes not treated with
bariatric surgery, we found that 1.5%
of individuals with recent evidence of
clinical diabetes achieved at least partial
remission over a 7-year period. If these
results were generalized to the 25.6 mil-
lion U.S. adults living with type 2 diabe-
tes in 2010 (25), they would suggest that
384,000 adults could experience remis-
sion over the next 7 years. However, the
rate of prolonged remission was ex-
tremely rare (0.007%), translating into
only 1,800 adults in the U.S. experienc-
ing remission lasting at least 5 years. To
provide context, 1.7% of the cohort
died, while only 0.8% experienced any
level of remission, during the calendar
year 2006. Thus, the chances of dying
were higher than the chances of any
remission.
Remission is common among pa-
tients with diabetes who have un-
dergone bariatric surgery (;70%
experienced remission within 5 years
after surgery) (5). To the best of our
knowledge, this is the firstreportofre-
mission in a usual care setting among
the broad population of adults with
type 2 diabetes (e.g., not those who un-
derwent bariatric surgery). Although
remission of type 2 diabetes is uncom-
mon, it does occur in patients who have
not undergone surgical interventions.
Moreover, we found evidence of remis-
sion, albeit rare, even in individuals
previously requiring oral antiglycemic
medication or insulin therapy. It is im-
portant to consider that these findings
were based on a conservative sampling
frame that excluded patients without
recent evidence of clinical diabetes at
baseline and using a more stringent
definition of remission than typically
used in the literature. These findings
challenge widespread assumptions
that type 2 diabetes is uniformly irre-
versible and progressive.
Table 1—Baseline demographics and diabetes characteristics
Characteristic
Full cohort*
(N= 122,781)
New-onset cohort†
(N= 18,684)
Age, mean (SD), years 61.8 (12.6) 56.5 (13.1)
Female sex 57,779 (47.1) 8,723 (46.7)
Race/ethnicity
African American 13,169 (11.3) 1,751 (10.5)
Asian 11,505 (9.9) 1,810 (10.8)
Filipino 8,793 (7.5) 1,469 (8.8)
Latino 17,391 (14.9) 2,798 (16.7)
White 56,493 (48.4) 7,825 (46.8)
Mixed 7,592 (6.5) 828 (5.0)
Other 1,894 (1.6) 249 (1.5)
Socioeconomic status: neighborhood deprivation index
1st quartile (least deprived) 22,370 (19.8) 3,301 (19.5)
2nd quartile 34,142 (30.2) 5,092 (30.0)
3rd quartile 33,122 (29.3) 5,138 (30.3)
4th quartile (most deprived) 23,363 (20.7) 3,434 (20.2)
HbA
1c
,5.7% [,39 mmol/mol] 4,807 (3.9) 1,288 (7.1)
5.7–6.4% [39–46 mmol/mol] 27,232 (22.3) 5,213 (28.6)
6.5–7.9% [47–63 mmol/mol] 59,286 (48.6) 8,225 (45.1)
8.0% + [$64 mmol/mol] 30,775 (25.2) 3,495 (19.2)
BMI, mean (SD), kg/m
2
31.1 (5.4) 31.5 (5.4)
eGFR, mean (SD), mL/min/1.73 m
2
77.5 (23.8) 85.2 (21.5)
Diabetes medication
None 13,639 (11.1) 5,133 (27.5)
Oral agents only 86,064 (70.1) 12,632 (67.6)
Any insulin 23,078 (18.8) 919 (4.9)
Diabetes duration, median (IQR), years 5.9 (2.9–11.2) 0.9 (0.6–1.4)
Data are expressed as n(%), unless otherwise indicated. IQR, interquartile range. *Missing data in
full cohort: race (N= 5,944), socioeconomic status (N= 9,784), HbA
1c
level (N= 681), diabetes
duration (N= 19,229), BMI (N= 36,513), eGFR (N=7,300).†New-onset diabetes defined as ,2
years since diagnosis. Missing data in new-onset cohort: race (N= 1,954), socioeconomic status
(N= 1,719), HbA
1c
level (N=463),BMI(N= 4,946), eGFR (N=1,054).
Table 2—Baseline comorbidities
Characteristic
Full cohort
(N= 122,781)
New-onset cohort*
(N= 18,684)
Hypertension 103,171 (84.0) 13,800 (73.9)
Hyperlipidemia 96,938 (79.0) 13,469 (72.1)
Cardiovascular disease 25,115 (20.5) 2,161 (11.6)
Diabetic retinopathy 20,305 (16.5) 724 (3.9)
Diabetic neuropathy 23,806 (19.4) 1,877 (10.1)
Congestive heart failure 8,289 (6.8) 724 (3.9)
Data are expressed as n(%). *New-onset diabetes defined as ,2 years since diagnosis.
care.diabetesjournals.org Karter and Associates 3191
Our study should be considered inlight
of the Look AHEAD study (7). Investiga-
tors reported the incidence of any remis-
sion (partial or complete) at 1 year was
11.5% in the intensive management arm,
and 2.0% in the diabetes support and ed-
ucation arm, compared with a 1.6% inci-
dence over 7 years in our study. Several
important differences between the study
populations and study design should be
noted. The Look AHEAD study presum-
ably enrolled patients who were willing
or motivated to make extensive lifestyle
changes, while our study population is
representative of a general adult diabetes
population. Patients in our study also
had a longer duration of diabetes and a
higher burden of comorbidity than those
enrolled in the Look AHEAD study. In ad-
dition, the two studies use different op-
erational definitions of remission. Future
studies should seek to validate and ex-
tend our findings in other populations,
including children and adolescents.
Consistent with prior literature, we
observed that the incidence of remis-
sion was not uniform across individuals.
Variables associated with remission
were similar to those identified in the
Look AHEAD study (7); they included
no glucose-lowering medication and
lower baseline HbA
1c
level. We also ob-
served an association of remission with
fewer years since diagnosis, although
the size of the effect weakened after
controlling for demographic, diabetes,
and health status covariatesdafinding
that may reflect differences between
the actual physiologic onset of diabetes
versus the point of clinical recognition
(i.e., diagnosis) (26), as well as the mul-
titude of factors that influence the nat-
ural history of b-cell function and mass
(8,27,28). These findings, combined
with our observation that remission
was more common in adults with lower
baseline HbA
1c
levels who had not been
previously treated with insulin or oral
hypoglycemic medications, provide in-
direct evidence of the potential benefits
of screening, diagnosis, and initiating in-
tensive lifestyle management before
the onset or early in the natural history
of the disease (18). Although lower BMI
was associated with remission in the un-
adjusted analysis, these associations did
not persist after controlling for covari-
ates. Baseline BMI was similarly not as-
sociated with remission in the Look
AHEAD study, which found that only a
change in BMI predicted remission.
These findings may have important im-
plications when considering treatment
options (lifestyle vs. surgical therapy)
for overweight and obese individuals
with diabetes.
We also identified a number of vari-
ables associated with higher rates of re-
mission that have not been previously
described, as follows: older age, African
American race, higher socioeconomic
status, renal impairment, and absence
of dyslipidemia. Higher rates of remis-
sion in adults over age 65 years are con-
sistent with evidence from the Diabetes
Prevention Program demonstrating
that older adults experienced the great-
est risk reduction in diabetes from life-
style modification (29,30), findings that
have particular salience given that the
prevalence of type 2 diabetes is highest
among the elderly (25). Although the
unadjusted analysis suggested no asso-
ciation between remission and African
American race, after controlling for co-
variates including neighborhood-level
socioeconomic status we found that Af-
rican American race was positively asso-
ciated with remission. Despite a higher
incidence of diabetes in African Amer-
icans (15), this finding complements
earlier reports (14,31) in our study
population of lower rates of certain
diabetes complications including heart
disease among African Americans.
Higher socioeconomic status was asso-
ciated with remission in both unad-
justed and adjusted analyses, although
the effect size was modest. The rates of
remission increased with renal impair-
ment; however, we did not see marked
differences across stages of chronic kid-
ney disease. Because insulin is normally
metabolized by the kidney (32), worsen-
ing renal function can lead to improved
glycemic control (33); however, this typ-
ically occurs at later stages of renal im-
pairment and cannot fully explain our
findings. Finally, higher rates of remission
among individuals without dyslipidemia
may reflect common biologic pathways
in the metabolic syndrome among
dyslipidemia, obesity, and impaired glu-
cose tolerance (34), and should be in-
vestigated in future studies, given the
high rates of comorbid dyslipidemia in
patients with type 2 diabetes.
More research is needed on how and
whether to refinethecasedefinitions
for diabetes remission. Considerable
Table 3—Incidence rates of remission for the full cohort and stratified by time since diagnosis
Total person-years
at risk*
Incident
events (n)
7-Year cumulative
incidence (95% CI)†
Incidence rate per 1,000
person-years (95% CI)
Any remission
All 586,725 1,761 1.60% (1.53–1.68) 3.00 (2.86–3.14)
Time since diagnosis ,2 years (n= 18,451) 88,473 776 4.55% (4.25–4.88) 8.77 (8.15–9.39)
Time since diagnosis 2–3 years (n= 18,127) 89,526 424 2.54% (2.31–2.79) 4.74 (4.29–5.19)
Time since diagnosis 4–5 years (n= 15,122) 75,304 228 1.67% (1.46–1.89) 3.03 (2.63–3.42)
Time since diagnosis 6–9 years (n= 20,270) 102,270 152 0.82% (0.70–0.96) 1.49 (1.25–1.72)
Time since diagnosis $10 years (n= 30,326) 147,333 98 0.37% (0.30–0.45) 0.67 (0.53–0.80)
Baseline diabetes therapy, no medication (n= 13,502) 62,009 1,538 12.33% (11.76–12.93) 24.80 (23.56–26.04)
Baseline diabetes therapy, OHA only (n= 84,968) 419,376 212 0.28% (0.24–0.32) 0.51 (0.44–0.57)
Baseline diabetes therapy, insulin (n= 22,625) 105,340 11 0.05% (0.03–0.10) 0.10 (0.04–0.17)
Partial remission 587,341 1,615 1.47% (1.40–1.54) 2.75 (2.62–2.88)
Complete remission 593,216 140 0.14% (0.12–0.16) 0.24 (0.20–0.28)
Prolonged remission 170,356 6 0.007% (0.003–0.02) 0.035 (0.007–0.063)
OHA, oral hypoglycemic agent. *Defined as the total number of person-years 12 months after cohort inception for partial and complete remission,
and after 60 months for prolong edre mission. †Defined as 100% minus the cumulative survival probability calculated using the Kaplan-Meier method.
3192 Diabetes Remission Diabetes Care Volume 37, December 2014
Table 4—Full cohort: unadjusted and adjusted Cox proportional hazard ratios for any remission
Characteristic
Any remission (n=1,761)
n(%)
Hazard ratio (95% CI)
Unadjusted Pvalue Fully adjusted* Pvalue
Sex
Men 820 (1.3) 1.0 (Ref) 1.0 (Ref)
Women 941 (1.7) 1.29 (1.17–1.41) ,0.0001 1.04 (0.91–1.19) 0.5532
Age, years
,45 81 (0.7) 1.0 (Ref) 1.0 (Ref)
45–65 551 (0.9) 1.28 (1.01–1.61) 0.0403 1.55 (1.04–2.31) 0.0323
$65 1,129 (2.2) 3.02 (2.41–3.78) ,0.0001 2.54 (1.67–3.86) ,0.0001
Race/ethnicity
White 780 (1.4) 1.0 (Ref) 1.0 (Ref)
African American 189 (1.5) 1.02 (0.87–1.2) 0.7931 1.42 (1.15–1.75) 0.001
Asian 162 (1.4) 0.99 (0.84–1.18) 0.9315 0.91 (0.72–1.14) 0.407
Filipino 80 (0.9) 0.64 (0.51–0.8) 0.0001 0.75 (0.56–1.01) 0.0551
Latino 160 (0.9) 0.66 (0.56–0.78) ,0.0001 1.05 (0.84–1.31) 0.6587
Mixed race 83 (1.1) 0.78 (0.62–0.98) 0.0318 0.97 (0.73–1.3) 0.8497
Other 17 (0.9) 0.66 (0.41–1.06) 0.0844 1.09 (0.56–2.11) 0.8083
Socioeconomic status: neighborhood deprivation index
1st quartile (least deprived) 398 (1.8) 1.55 (1.33–1.81) ,0.0001 1.28 (1.03–1.61) 0.0288
2nd quartile 522 (1.5) 1.34 (1.15–1.55) 0.0001 1.08 (0.87–1.35) 0.4599
3rd quartile 463 (1.4) 1.24 (1.06–1.44) 0.0056 1.14 (0.91–1.41) 0.2485
4th quartile (most deprived) 259 (1.1) 1.0 (Ref) 1.0 (Ref)
Time since diagnosis
,2 years 776 (4.2) 13.2 (10.7–16.29) ,0.0001 2.19 (1.63–2.95) ,0.0001
2–3 years 424 (2.3) 7.21 (5.79–8.98) ,0.0001 1.68 (1.24–2.28) 0.0008
3–10 years 380 (1.1) 3.28 (2.62–4.09) ,0.0001 1.6 (1.19–2.15) 0.0017
$10 years 98 (0.3) 1.0 (Ref) 1.0 (Ref)
Baseline HbA
1c
level
,5.7% (,39 mmol/mol) 131 (2.8) 24.46 (16.84–35.51) ,0.0001 19.2 (11.79–31.27) ,0.0001
5.7–6.4% (39–46 mmol/mol) 628 (2.3) 20.46 (14.55–28.75) ,0.0001 6.61 (4.23–10.32) ,0.0001
6.5–7.9% (47–63 mmol/mol) 963 (1.6) 14.24 (10.16–19.95) ,0.0001 2.7 (1.74–4.2) ,0.0001
$8% ($64 mmol/mol) 35 (0.1) 1.0 (Ref) 1.0 (Ref)
Diabetes medication in the past 1 year
None 1,538 (11.4) 239.75 (132.5–433.81) ,0.0001 132.51 (61.62–284.97) ,0.0001
Oral agents only 212 (0.2) 4.99 (2.72–9.15) ,0.0001 3.71 (1.71–8.03) 0.0009
Any insulin 11 (0) 1.0 (Ref) 1.0 (Ref)
Diabetes medication since 2002
None 1,447 (12.3) 128.07 (85.56–191.7) ,0.0001
Oral agents only 290 (0.3) 3.37 (2.22–5.11) ,0.0001
Any insulin 24 (0.1) 1.0 (Ref)
BMI, kg/m
2
,25 219 (1.9) 1.53 (1.29–1.82) ,0.0001 1 (0.79–1.25) 0.973
$25 to ,30 435 (1.7) 1.33 (1.15–1.54) 0.0001 0.83 (0.69–1) 0.0504
$30 to ,35 290 (1.3) 1.02 (0.87–1.2) 0.7886 0.8 (0.66–0.96) 0.0184
$35 308 (1.3) 1.0 (Ref) 1.0 (Ref)
CKD stage†
0 + 1 349 (0.9) 1.0 (Ref) 1.0 (Ref)
2 808 (1.6) 1.8 (1.59–2.04) ,0.0001 1.1 (0.91–1.33) 0.3475
3 467 (2) 2.25 (1.95–2.58) ,0.0001 1.33 (1.06–1.67) 0.0143
4 48 (2) 2.45 (1.81–3.31) ,0.0001 2.11 (1.37–3.26) 0.0008
5 12 (1.6) 1.9 (1.07–3.39) 0.0283 2.51 (1.14–5.51) 0.0221
No hypertension 262 (1.4) 0.91 (0.79–1.03) 0.1373 0.97 (0.79–1.19) 0.7791
Hypertension 1,499 (1.5) 1.0 (Ref) 1.0 (Ref)
No dyslipidemia 445 (1.7) 1.28 (1.15–1.42) ,0.0001 1.21 (1.03–1.43) 0.0216
Dyslipidemia 1,316 (1.4) 1.0 (Ref) 1.0 (Ref)
No cardiovascular disease 1,392 (1.4) 0.91 (0.81–1.02) 0.1174 0.91 (0.76–1.08) 0.2781
Cardiovascular disease 369 (1.5) 1.0 (Ref) 1.0 (Ref)
No diabetic retinopathy 1,677 (1.7) 3.86 (3.1–4.81) ,0.0001 1.11 (0.81–1.52) 0.5129
Diabetic retinopathy 84 (0.4) 1.0 (Ref) 1.0 (Ref)
Continued on p. 3194
care.diabetesjournals.org Karter and Associates 3193
uncertainty lies in the appropriate gly-
cemic thresholds for partial and com-
plete remission, the choice of glycemic
measure (fasting blood glucose vs.
HbA
1c
level), and the number and timing
of measurements. Studies of different
definitions of remission in bariatric sur-
gery suggest that the choice of thresh-
olds can have a significant impact on
estimates of remission rates (20,35), al-
though the frequency of measurement
remains unexplored. Given the growing
importance of this area of research, op-
erational definitions of remission for dif-
ferent study designs are needed, and,
once established, should be considered
in future diabetes clinical trials and epi-
demiologic studies as a clinical outcome
measure.
Our study also has important implica-
tions for clinical decision making. With-
out an empiric trial of withdrawal of
glucose-lowering medication in a type 2
diabetes patient with subdiabetic glu-
cose levels or normoglycemia, it is im-
possible to ascertain whether their
medication regimen is still needed. Clini-
cal inertia, although primarily used to
refer to the failure of clinicians to initi-
ate or intensify therapy (36), may also
be applied to underdiagnosis of remis-
sion if doctors do not test a patient’s
ability to maintain metabolic control
without medications. Such an empiric
trial, if successful, may reduce the po-
tential adverse effects of diabetes med-
ications and reduce healthcare costs,
while improving patient outcomes, a
hypothesis that should be tested in
future clinical trials.
Several limitations need to be acknowl-
edged. First, these data were collected
from routine clinical practice rather than
from a prescribed study protocol with
regular collection intervals. As a result,
the frequency of HbA
1c
testing varied
widely, particularly in our population of
interest, which consisted of individuals
with subdiabetic HbA
1c
levels. This limita-
tion is a direct result of ambiguity in clin-
ical guidelines about the intensity of
diabetes management required for these
individuals. Second, the criteria by which
diabetes was first identified differed for
subjects who experienced remission ver-
sus those who did not. The patients who
eventually went into remission had simi-
lar likelihoods of having two or more out-
patient diagnoses of diabetes as their
basis for being identified as having diabe-
tes when compared with subjects not
going into remission (96% vs. 97%, re-
spectively). However, patients who expe-
rienced remission were less likely than
patients who did not experience remis-
sion to be first identified due to a
diabetes-related medication (29% vs.
93%, respectively), abnormal labora-
tory test results (68% vs. 93%, respec-
tively), or hospital discharge (4% vs. 9%,
respectively). This suggests that patients
who eventually experienced remission
had less advanced diabetes at the onset
and more often received diagnoses from
their provider without laboratory evi-
dence. Third,although the consensus def-
inition of remission includes both fasting
glucose and HbA
1c
measures, this study
used only HbA
1c
measures because fast-
ing glucose measures are not routinely
collected after an established diabetes di-
agnosis by physicians at KPNC. Fourth,
during the period of follow-up in 2011,
KPNC updated its assay for measuring
HbA
1c
. This change increased test values
compared with the period of 2005–2011,
which would tend to bias our study to-
ward more conservative estimates of re-
mission. Fifth, because electronic capture
of BMI occurred gradually over the base-
line period, BMI was missing in 30% of
participants. However, the proportion of
patients with missing BMI data was rela-
tively constant across states of remission,
as follows: nonremission (30%), partial re-
mission (29%), complete remission (31%),
and prolonged remission (33%). While
not eliminating the potential for bias,
this greatly reduces the concern. Sixth,
our results are partially dependent on
methodological decisions about opera-
tionalizing definitions of remission. In
general, we favored a conservative ap-
proach. As discussed above, the appropri-
ate definitions of diabetes remission
remain an area of considerable debate
(10), which this study seeks to partially
inform.
Despite these limitations, to our
knowledge, this is the first study of the
epidemiology of remission in type 2 di-
abetes in adults not treated with bariat-
ric surgery. Our analysis shows that
remission is rare and variable. The likeli-
hood of remission is more common
among individuals with early-onset
diabetes and those not treated with
glucose-lowering medications at the
point of diabetes diagnosis. Although
rare, remission can also occur in individ-
uals with more severe diabetes and
those previously treated with insulin.
This information will help providers dis-
cuss the prognosis of diabetes in their
newly diagnosed patients. The large
differences in remission rates from
our community setting, compared with
those from lifestyle intervention trials
like Look AHEAD, should inform a wider
debate about the appropriate strategies
for managing diabetes.
Acknowledgments. S.N. dedicates this study
to his mother, Anju Nundy, who has type 2
diabetes.
Funding. This research was funded by the
National Institute of Diabetes and Digestive and
Kidney Diseases (NIDDK) (grant R01-DK-081796),
Table 4—Continued
Characteristic
Any remission (n=1,761)
n(%)
Hazard ratio (95% CI)
Unadjusted Pvalue Fully adjusted* Pvalue
No diabetic neuropathy 1,498 (1.5) 1.32 (1.16–1.51) ,0.0001 0.84 (0.7–1.01) 0.0689
Diabetic neuropathy 263 (1.1) 1.0 (Ref) 1.0 (Ref)
No congestive heart failure 1,630 (1.4) 0.78 (0.65–0.93) 0.0068 0.88 (0.68–1.14) 0.3312
Congestive heart failure 131 (1.6) 1.0 (Ref) 1.0 (Ref)
CKD, chronickidneydisease; Ref, referencevalue.*Fully adjusted model in cludes all vari ables listed in the table. †St age 0 + 1, eGFR $90 mL/min/1.73 m
2
;
stage 2, eGFR 60–89 mL/min/1.73 m
2
; stage 3, eGFR 30–59 mL/min/1.73 m
2
; stage 4, eGFR 15–29 mL/min/1.73 m
2
; stage 5, eGFR ,15 mL/min/1.73 m
2
.
3194 Diabetes Remission Diabetes Care Volume 37, December 2014
and NIDDK Centers for Diabetes Translation Re-
search at The University of Chicago (grant P30-
DK-092949) and Kaiser Permanente (grant
P30-DK-092924). A.J.K. was also supported by
the Agency for Healthcare Research and Qual-
ity Health Services Research Training grant
T32-HS-00084.
The funders had no role in the design and
conduct of the study; the collection, manage-
ment, analysis, and interpretation of the data;
and the preparation, review, or approval of the
manuscript.
Duality of Interest. No potential conflicts of
interest relevant to this article were reported.
Author Contributions. A.J.K. helped to acquire
the funding and sponsorship for the study;
acquired the data; oversaw the analysis; contrib-
uted to the study design, the analysis and
interpretation of the data, and the critical review
and revision of the manuscript; and drafted the
revised manuscript. S.N. conceived the study;
contributed to the study design, the analysis and
interpretation of the data, and the critical review
and revision of the manuscript; and drafted the
original manuscript. M.M.P. contributed to the
study design, the analysis and interpretation of
the data, and the critical review and revision of
the manuscript. H.H.M. contributed to the study
design, the analysis and interpretation of the
data, and critical review and revision of the
manuscript and also contributed administrative
support. E.S.H. acquired the funding and sponsor-
ship for the study and contributed to the study
design, the analysis and interpretation of the
data, and the critical review and revision of the
manuscript. A.J.K. and M.M.P. are the guarantors
of this work and, as such, had full access to all the
data in the study and take responsibility for the
integrity of the data and the accuracy of the data
analysis.
References
1. Saudek CD. Can diabetes be cured? Pote ntial
biological and mechanical approaches. JAMA
2009;301:1588–1590
2. UK Prospective Diabetes Study (UKPDS)
Group. Intensive blood-glucose control with
sulphonylureas or insulin compared with con-
ventional treatment and risk of complications
in patients with type 2 diabetes (UKPDS 33).
Lancet 1998;352:837–853
3. Turner RC, Cull CA, Frighi V, Holman RR; UK
Prospective Diabetes Study (UKPDS) Group. Gly-
cemic control with diet, sulfonylurea, metfor-
min, or insulin in patients with type 2 diabetes
mellitus: progressive requirement for multiple
therapies (UKPDS 49). JAMA 1999;281:2005–
2012
4. Mingrone G, Panunzi S, De Gaetano A, et al.
Bariatric surgery versus conventional medical
therapy for type 2 diabetes. N Engl J Med
2012;366:1577–1585
5. Arterburn DE, Bogart A, Sherwood NE, et al.
A multisite study of long-term remission and
relapse of type 2 diabetes mellitus following
gastric bypass. Obes Surg 2013;23:93–102
6. Buchwald H, Avidor Y, Braunwald E, et al.
Bariatric surgery: a systematic review and
meta-analysis. JAMA 2004;292:1724–1737
7. Gregg EW, Chen H, Wagenknecht LE, et al;
Look AHEAD Research Group. Association of an
intensive lifestyle intervention with remission
of type 2 diabetes. JAMA 2012;308:2489–2496
8. Lim EL, Hollingsworth KG, Aribisala BS, Chen
MJ, Mathers JC, Taylor R. Reversal of type 2 di-
abetes: normalisation of beta cell function in
association with decreased pancreas and liver
triacylglycerol. Diabetologia 2011;54:2506–
2514
9. Taylor R. Type 2 diabetes: etiology and re-
versibility. Diabetes Care 2013;36:1047–1055
10. Buse JB, Caprio S, Cefalu WT, et al. How do
we define cure of diabetes? Diabetes Ca re 2009;
32:2133–2135
11. Boyle JP, Thompson TJ, Gregg EW, Barker
LE, Williamson DF. Projection of the year 2050
burden of diabetes in the US adult population:
dynamic modeling of incidence, mortality, and
prediabetes prevalence. Popul Health Metr
2010;8:29
12. Wild S, Roglic G, Green A, Sicree R, King H.
Global prevalence of diabe tes: estimates for the
year 2000 and projections for 2030. Diabetes
Care 2004;27:1047–1053
13. Selby JV, Ray GT, Zhang D, Colby CJ. Excess
costs of medical care for patients with diabetes
in a managed care population. Diabetes Care
1997;20:1396–1402
14. Karter AJ, Ferrara A, Liu JY, Moffet HH,
Ackerson LM, Selby JV. Ethnic disparities in di-
abetic complications in an insured population.
JAMA 2002;287:2519–2527
15. Karter AJ, Schillinger D, Adams AS, et al.
Elevated rates of diabetes in Pacific Islanders
and Asian subgroups: the Diabetes Study of
Northern California (DISTANCE). Diabetes Care
2013;36:574–579
16. Ozer H, George SL, Schiffer CA, et al. Pro-
longed subcutaneous administration of recombi-
nant alpha 2b interferon in patients with
previously untreated Philadelphia chromosome-
positive chronic-phase chronic myelogenous leu-
kemia: effect on remission duration and survival:
Cancer and Leukemia Group B study 8583. Blood
1993;82:2975–2984
17. Dixon JB, Chuang LM, Chong K, et al. Pre-
dicting the glycemic response to gastric bypass
surgery in patients with type 2 diabetes. Diabe-
tes Care 2013;36:20–26
18. American Diabetes Association. Standards
of medical care in diabetesd2013. Diabetes
Care 2013;36(Suppl. 1):S11–S66
19. World Health Organization. Use of glycated
haemoglobin (HbA1c) in the diagnosis of diabe-
tes mellitus: abbreviated report of a WHO con-
sultation [article online], 2011. Available from
www.who.int/diabetes/publications/report-
hba1c_2011.pdf. Accessed April 22, 2013
20. Blackstone R, Bunt JC, Cortes MC,
Sugerman HJ. Type 2 diabetes after gastric by-
pass: remission in five models usi ng HbA1c, fast-
ing blood glucose, and medication status. Surg
Obes Relat Dis 2012;8:548–555
21. National Kidney Foundation Kidney Disease
Outcomes Quality Initiative. KDOQI clinical prac-
tice guidelines and clinical practice recommenda-
tions for diabetes and chronic kidney disease.Am
J Kidney Dis 2007;49(Suppl. 2):S12–S154
22. Messer LC, Laraia BA, Kaufman JS, et al. The
development of a standardized neighborhood
deprivation index. J Urban Health 2006;83:
1041–1062
23. Laraia BA, Karter AJ, Warton EM, Sch illinger
D, Moffet HH, Adler N. Place matters: neighbor-
hood deprivation and cardiometabolic risk fac-
tors in the Diabetes Stud y of Northern California
(DISTANCE). Soc Sci Med 2012;74:1082–1090
24. Collett D. Modelling Survival Data in Medi-
cal Research. 2nd ed. Boca Raton, FL, Chapman
& Hall/CRC, 2003
25. Centers for Disease Control and Prevention .
2011 National Diabetes Fact Sheet: National Es-
timates and General Information on Diabetes
and Prediabetes in the United States. Atlanta,
GA, U.S. Department of Health and Human Ser-
vices, Centers for Disease Control and Preven-
tion, 2011
26. Harris MI, Klein R, Welborn TA, Knuiman
MW. Onset of NIDDM occurs at least 4-7 yr
before clinical diagnosis. Diabetes Care 1992;
15:815–819
27. Clauson P, Linnarsson R, Gottsater A,
Sundkvist G, Grill V. Relationships between di-
abetes duration, metabolic control and beta-
cell function in a representative population of
type 2 diabetic patients in Sweden. Diabet Med
1994;11:794–801
28. Butler AE, Janson J, Bonner-Weir S, Ritzel R,
Rizza RA, Butler PC. Beta-cell deficit and in-
creased beta-cell apoptosisinhumanswith
type 2 diabetes. Diabetes 2003;52:102–110
29. Crandall J, Schade D, Ma Y, et al.; Diabetes
Prevention Program Research Group. The influ-
ence of age on the effects of lifestyle modifica-
tion and metformin in prevention of diabetes. J
Gerontol A Biol Sci Med Sci 2006;61:1075–1081
30. Ahmad LA, Crandall JP. Type 2 diabetes pre-
vention: a review. Clin Diabetes 2010;28:53–59
31. Kanaya AM, Adler N, Moffet HH, et al. Het-
erogeneity of diabetes outcomes among asians
and pacific islanders in the US: the diabetes
study of northern california (DISTANCE). Diabe-
tes Care 2011;34:930–937
32. Rabkin R, Ryan MP, Duckworth WC. The re-
nal metabolis m of insulin. Diabetologia 1984;
27:351–357
33. Cavanaugh KL. Diabetes management is-
sues for patients with chronic kidney disease.
Clin Diabetes 2007;25:90–97
34. Alberti KG, Zimmet P, Shaw J; IDF Epidemi-
ology Task Force Consensus Group. The meta-
bolic syndromeda new worldwide definition.
Lancet 2005;366:1059–1062
35. Ramos-Levi AM, Cabrerizo L, Mat´
ıaP,
S´
anchez-Pernaute A, Torres AJ, Rubio MA.
Which criteria should be used to define type 2
diabetes remission after bariatric surgery? BMC
Surg 2013;13:8
36. Phillips LS, Branch WT, Cook CB, et al. Clin-
ical inertia. Ann Intern Med 2001;135:825–834
care.diabetesjournals.org Karter and Associates 3195