Achieving Good Glycemic Control: Initiation of New Antihyperglycemic Therapies in Patients With Type 2 Diabetes From the Kaiser Permanente Northern California Diabetes Registry

, , , , , ,
The American Journal of Managed Care, April 2005, Volume 11, Issue 4

Objective: To compare the effectiveness of antihyperglycemictherapies in type 2 diabetic patients with poor glycemic control(baseline glycosylated hemoglobin [HbA1c ] > 8%).

Study Design: Longitudinal (cohort) study.

Methods: Study patients were 4775 type 2 diabetic patientswho initiated new antihyperglycemic therapies and maintainedthem for up to 1 year. The study setting was Kaiser PermanenteNorthern California Medical Group, an integrated, prepaid,healthcare delivery organization. Treatment regimens were 1 ormore of the following: insulin, thiazolidinediones, sulfonylureas,biguanides (metformin), or other less frequently used options(including meglitinides or α-glucosidase inhibitors).

Results: In this cohort, the mean HbA1c was 9.9% when therapywas initiated. Within 1 year, there was a drop of 1.3 percentagepoints in the mean HbA1c (to 8.6%), and 18% of new initiatorsachieved HbA1c values of ≤7%. After adjusting for baseline clinicaldifferences, the proportion of patients treated to goal was greatestamong those receiving thiazolidinediones in combination (24.6%-25.7%) or a regimen of metformin and insulin (24.9%), while theleast success was experienced by those receiving sulfonylureasalone (12.5%) or insulin-sulfonylureas regimens (10.9%). Theprobability of achieving the target goal was most strongly predictedby the level of glycemic control before initiation, but patientbehaviors (eg, frequent self-monitoring, lower rates of missedappointments) also were strongly associated with greater levels ofcontrol.

Conclusion: Overall, therapy initiation resulted in an impressivepopulation-level benefit. However, since most new initiatorsstill had not achieved good control within 12 months, careful monitoringand prompt therapy intensification remain important.

(Am J Manag Care. 2005;11:262-270)



The importance of maintaining tight blood glucosecontrol for the prevention of microvascular complications(eg, retinopathy, nephropathy) is wellestablished.1,2 For almost half of a century, insulin andsulfonylureas were the only treatment options for diabetesin the United States. In 1959, phenformin, abiguanide, was introduced in the US market but wasremoved in 1977 because of concerns regarding lacticacidosis.3 Metformin (another biguanide), while availableearlier in other countries, did not reach the USmarket until 1995. In the last decade, 3 new therapeuticclasses (alpha-glucosidase inhibitors, thiazolidinediones,and meglitinides) have been introduced.4Although numerous randomized, placebo-controlledclinical trials have evaluated the of these medicationsalone or in combinations,5 the relative of the whole spectrum of pharmacologic options innonexperimental settings has rarely been assessed. TheAmerican Diabetes Association (ADA) recommendstreating diabetic patients to achieve a glycemic targetof glycosylated hemoglobin (HbA1c) ≤7%,6 but the abilityof diabetes therapies to achieve this goal in usual practiceis poorly understood. We studied 4775 type 2 diabeticpatients with poor glycemic control (HbA1c > 8%) who initiatednew treatment regimens (index therapy) in theKaiser Permanente Northern California Medical Group(Kaiser Permanente) during 1999-2000. We comparedthe proportions of patients who achieved good glycemiccontrol (HbA1c ≤7%) within 3-12 months after initiatingthe most commonly used therapies.



Kaiser Permanente, an integrated, nonprofit, group-practice,prepaid healthcare delivery organization, providescomprehensive medical services to more than 3million members (as of January 2000) throughoutnorthern California, including the San Francisco Bayand Sacramento metropolitan areas, or ~25% to 30% ofthe region's population. Care is provided by approximately4400 physicians of The Permanente MedicalGroup at 17 hospitals and 152 medical offices. TheKaiser Permanente members are predominantlyemployed or retired individuals and their families, andclosely approximate the general population ethnicallyand socioeconomically except for the extreme tails ofthe income distribution.7-9

Source Population

In 1993, Kaiser Permanente established the KaiserPermanente Northern California Diabetes Registry. Thisregistry included 116 344 diabetic patients on January1, 1999; it has an estimated sensitivity of ~99% and a2% false-positive rate. The registry is updated annuallyby identifying all health plan members with diabetesfrom automated databases for pharmacy, laboratory,hospitalization records, and outpatient diagnoses. Themethods used in the Kaiser Permanente diabetes registryhave been described previously.8,10-13

Cohort Identification

We identified all diabetes registry members who initiateda new diabetes therapy between June 1, 1999, andMay 31, 2000; had been diagnosed with diabetes for atleast 1 year before initiation of this therapy; and had afull year of Kaiser membership with pharmacy benefitsafter initiation. As a group, individuals who initiated newdiabetes therapies differed from those who maintainedongoing therapies in terms of glycemic control, diseaseseverity, and patient characteristics and behaviors.14Thus, we relied on the "new user" design15 that restrictsthe study cohort to individuals who initiated new therapies;we used only those glycemic outcomes thatoccurred after therapy initiation but before therapyswitching or discontinuation, while controlling forpatient characteristics present before initiation. Bydesign, type 1 patients were not eligible because theywould not switch therapeutic classes once established oninsulin. We excluded diabetic patients with end-stagerenal disease (n = 3177) from our sampling framebecause of this disease's impact on insulin clearance(and thus glycemic control) and clearance of sulfonylureas,and because end-stage renal disease is a contraindicationfor metformin therapy.16 All those withoutcontinuous health plan membership (n = 4752) or withouta drug benefit (n = 2574) at any time during thestudy period were excluded to minimize misclassificationof subjects who may have filled prescriptions innon-Kaiser pharmacies. Because a single-tiered pharmacybenefit was in place at the time of the study, the out-of-pocket costs were uniform across therapeutic classesregardless of whether medications were branded orgeneric. Twenty-seven percent (n = 23 501) of diabetichealth plan members initiated new diabetes therapiesbetween June 1, 1999, and May 31, 2000, after excludingthose who were diagnosed with diabetes less than 12months before initiating the new (index) therapy. Ofthese new users of therapy, 8333 had HbA1c measuredboth during the 12-month window before initiation andthe 3-to 12-month period after initiation and before discontinuingor modifying the index therapy. Of these eligiblenew users, 4775 (57%) had poor glycemic controlbefore initiation, at levels that were above Kaiser's recommendedaction level (HbA1c > 8%). This group of4775 poorly controlled new users of therapy formedthe basis for this study and all analyses that follow.

Pharmacotherapeutic Exposure


The exposures of interest were new prescriptions forthe 12 most commonly prescribed monotherapy andcombination regimens (index therapies): (1) sulfonylureamonotherapy, the reference category; (2) metforminmonotherapy; (3) thiazolidinedione monotherapy; (4)insulin monotherapy; (5) metformin plus sulfonylurea;(6) metformin plus insulin; (7) sulfonylurea plus thiazolidinedione;(8) sulfonylurea plus insulin; (9) thiazolidinedioneplus insulin; (10) metformin plus sulfonylurea plusinsulin; (11) metformin plus sulfonylurea plus thiazolidinedione;and (12) "other" (meglitinides or α-glucosidaseinhibitors as monotherapy or as part of combinationtherapies). The exposure baseline date (index date) wasthe date on which the first prescription of the index therapywas dispensed. To ensure that patients were trulystarting a new regimen and maintaining that index therapy,we required that there was (1) at least 1 refill afterthe index date and any time within the follow-up periodfor each medication in the index therapy; (2) no evidenceof utilization of the index therapy during the 12months before the index date. These restrictions allowedus to minimize misclassification of patients as starting acombination therapy when they were in fact switchingfrom a single therapeutic class to a different class (andthus may have overlapping prescriptions for more than 1class during the transition).

We ended collection of our outcome (HbA1c ) valuesat the first occurrence of any of the following: the end ofthe study (up to 12 months after initiation of therapy),discontinuation of the index therapy, or modification ofthe index therapy. Therapy modification includedswitching from the index therapy to another therapy,adding an additional therapy to the index therapy, ordropping 1 of the components included in the indexcombination therapy. The date of treatment discontinuationwas calculated as the earliest date of the firstsupply of a new medication or, when medication wasdiscontinued, the date when the last supply plus a90-day grace period would be used up after the lastrecorded refill.

Treatment Response


For each individual, we assessed whether good control(HbA1c ≤7%) was achieved during the 3-to 12-monthwindow subsequent to the start of the index therapy,based on the last recorded HbA1c value in this follow-upperiod. We did not utilize HbA1c results from the first 3months of the index therapy to allow for initial dose titrationand physiologic adjustments to the new medication.For the new users who discontinued (230 out of 4775 or5%) or modified (1951 out of 4775 or 41%) the index therapywithin 12 months of its initiation, we restrictedanalysis to end points (HbA1c values) collected thatchange in therapy occurred. (In other words, we excludedHbA1c values assayed after a new user discontinued ormodified the index therapy.) Thus, only measures ofglycemic control that could be most directly linked withthe index therapy were used in analyses. HbA1c levelswere obtained from Kaiser's laboratory database, and allassays were conducted at Kaiser's centralized laboratorywith high-performance liquid chromatography.

Case-mix Adjustment

We previously observed that even among new users,there may be substantial variation in glycemic control,disease severity, and patient characteristics and behaviors,depending on the type of diabetes therapy that wasinitiated.14 Such differences may confound crude statisticalestimates. We therefore made case-mix adjustmentsto our statistical models for a wide range ofcovariates. These included age and sex, and the last previousHbA1c measurement and diabetes regimen(including no medication taken) before initiation of thenew therapy. Also included were the following covariates,assessed during the calendar year before indextherapy year: number of outpatient visits, standard diabetesprocesses of care (at least 1 annual visit to a primarycare physician, a dilated eye exam, andmeasurement of low-density lipoprotein cholesterol);type of primary care provider (endocrinologist vsother primary care provider); rate of missed scheduledoutpatient appointments; prescription copaymentamount; number of emergency room visits; number ofophthalmology exams; and frequency of self-monitoringof blood glucose (SMBG), based on glucose test striputilization.8,13

For a subanalysis, we used additional covariate datacaptured by a self-administered questionnaire or acomputer-assisted telephone interview in 1994-1997.Eighty-three percent (n = 77 726) of the 94 024 noninstitutionalizedhealth plan members in the diabetes registry(as of 1995) responded to that survey, which askedabout the daily number of insulin injections, use ofexercise and diet as diabetes treatments, time since diabetesdiagnosis, body mass index, smoking history, educationalattainment, and self-identified race/ethnicity.Neighborhood-level socioeconomic status was assessedby geocoding each member's address, and linking thegeocoded address to associated census-block groupaverage annual per capita income and proportion of residentsin a working-class profession. To assess whethermodel estimates based on the full cohort were robust,we compared them with estimates based on subanalysesof survey responders, with further covariate adjustment.In these subanalyses, we included time since diabetesdiagnosis, body mass index, smoking history,educational attainment, and self-identified race/ethnicity(58.0% non-Latino white, 12.4% African American,11.1% Asian, 9.1% Latino, 0.9% Pacific Islander, 0.7%Native American, 0.4% other, and 7.5% multiethnic), inaddition to all the covariates included in the full model(see above).

Analytic Methods

In addition to a crude (unadjusted) assessment, weused multivariate logistic regression models to assessthe probability of reaching glycemic control after initiatingnew therapies. All variables except for 1 (prebaselineHbA1c) were specified in their categorical form toconform to model linearity assumptions. Because of thestrong, linear relationship between prior HbA1c valuesand glycemic control in the follow-up, we included prebaselineHbA1c in its continuous form. Using data fromthe subset that returned a detailed health survey during1994-1997, we conducted additional analyses to assesswhether further adjustment for self-reported attributes(time since diabetes diagnosis, body mass index, smokinghistory, educational attainment, and self-identifiedrace/ethnicity) added important information. Ratherthan rely on adjusted odds ratios, which would yieldbiased estimators of effect given the common dependentvariables,17 we derived the adjusted (conditional)probability of achieving good control from the logisticregression model.


Study Subject Characteristics

This cohort of poorly controlled new users (Table l)was typical of the general diabetes patient population interms of age, sex, and use of health services. Most subjectswere cared for by a personal primary careprovider, had a relatively low pharmacy copayment,and practiced SMBG. This study cohort had much poorerglycemic control (mean HbA1c = 9.9%; SD = 1.5%)than that observed both in the source population of diabeticpatients and in all new users (not just the subjectswith poorly controlled diabetes selected forour cohort). This higher mean HbA1c wasexpected given that we excluded thosewhose HbA1c measurements were below 8%and selected exclusively new initiators (whomost likely failed to respond to previoustherapy). As a comparison, during this timeperiod, 30.2% of the general Kaiser diabeticpopulation had HbA1c ≤7% (mean HbA1c =8.2%; SD = 1.9%), and 19.1% of all new usershad HbA1c ≤7% (mean baseline HbA1c =8.8%; SD = 1.9%).

Most new users were treated withmonotherapies (52.6%) or no medication(11.9%) before initiation of the new therapy;the majority (69.0%) of this new-user cohortinitiated combination therapies (Table 2).The most common therapy before initiationwas sulfonylurea monotherapy (41.1%),while the most commonly initiated therapywas sulfonylurea plus metformin (38.7%).Ninety-one percent of cohort subjects treatedwith 1 oral agent transitioned into combinationtherapies with 2 or more oralagents. Similarly, 98.3% of the patients utilizinginsulin monotherapy added an oralagent to their insulin regimen rather thandiscontinuing insulin completely for oralagents. Most patients who originally weretaking an oral agent plus insulin wereswitched to other combination therapiesincluding insulin, or dropped oral agentsand relied on insulin monotherapy; however,relatively few discontinued the use ofinsulin.

Proportion Achieving Good Control

Among patients with poorly controlleddiabetes (HbA1c > 8%) who initiated newtherapies, 18.4% (95% confidence interval[CI] = 17.3%, 19.4%) achieved goodglycemic control (HbA1c ≤7%) during the 3to 12 months after initiation and maintenanceof their index therapy. Their meanHbA1c went from 9.9% (SD = 1.5%) beforeinitiation to 8.6% (SD = 1.7%) after initiation,a drop of ~1.3 percentage points in thesample means. Post-initiation levels ofglycemic control brought these poorly controllednew users closer to the overallsource population mean (HbA1c = 8.2%; SD= 1.9%); 30.2% of the source populationhad HbA1c values of ≤7%. In this cohort of4775 poorly controlled new users, 41% of the cohorthad the index therapy modified within 1 year afterinitiation, and 5% had discontinued the index therapy.

In unadjusted analyses, patients initiating sulfonylurea,metformin, or thiazolidinedione monotherapy orcombination therapy with sulfonylurea plus thiazolidinedionewere the most successful at achieving goodglycemic control (Table 3). Because the choice of initialtherapy should be dictated by a patient's condition, weassessed differences in the probability of achieving goodcontrol across the various diabetes regimens afteradjusting for patients' prior glycemic control in additionto other relevant attributes. These case mix-adjustedlogistic models included age, sex, pre-initiationHbA1c value, previous diabetes therapy, primary carephysician specialty, outpatient visit attendance, frequencyof SMBG (based on test strip consumption),amount of drug benefit copayment, number of annualoutpatient visits, antihypertensive and antilipemic therapies,and indicators for emergency room visits anddilated eye exams in the prior year. Sulfonylurea wasspecified as the reference group index therapy.

The effect of adjustment was substantial and consistentwith prescribing patterns. The 2 first-line therapies(sulfonylurea and metformin), prescribed commonly topatients with milder diabetes, were strongly associatedwith achieving good control in unadjusted models.However, their performance was greatly attenuatedafter adjusting for prior glycemic levels and diseaseseverity. Similarly, most of the combination therapies,which are presumably reserved for more advanced diabetes,had relatively low unadjusted but higher adjustedprobability of achieving good control. This patternof "confounding by indication"18 and the magnitude ofthe effect illustrate the importance of careful modeladjustment when assessing pharmacotherapeuticeffectiveness.

Different index therapies produced different levels ofglycemic control. Among subjects initiating monotherapy,only metformin users had a significantly greateradjusted probability of achieving good control thanusers of the reference therapy, sulfonylurea (17% vs12%). Initiators of thiazolidinedione monotherapy alsowere more likely to achieve good control (32%); however,this estimate was not statistically significant, probablydue to insufficient power because thiazolidinedioneis rarely prescribed as a monotherapy. Among subjectsinitiating combination therapy, users of the metforminplus-insulin combination and all combinations includingthiazolidinedione were significantly more likely(often exceeding a 2-fold increase) to achieve goodglycemic control compared with users of sulfonylureamonotherapy.




Behavioral factors also were predictive of achievinggood glycemic control. More frequent SMBG and satisfactoryappointment-keeping behavior (low rate of missingscheduled outpatient appointments) both wereassociated with a significant and graded increase ingood control after adjusting for initiated therapies andall of the covariates in the full model discussed above.The adjusted proportions of subjects (95% CI) achievinggood control ranged from 13.4% (reference, 11.3%,15.8%) with no practice of SMBG, 15.5% (14.0%, 17.2%;= .13) with some but less than daily practice of SMBG,and 18.8% (16.9%, 20.9%; = .0008) with daily SMBGpractice. The likelihood of achieving good control wasgreatest among those missing fewer scheduled outpatientappointments. In adjusted models, 17.0% (15.8%,18.2%) of those who missed fewer than 30% of their outpatientappointments achieved good control comparedwith 11.2% (8.7%, 14.2%; = .0009) of those who missed30% or more of their outpatient appointments.

A further analysis among the subset of the 67% (n= 3190) of the study cohort who responded to a previoushealth survey (1994-1997) allowed us to furtheradjust for self-reported case mix and severity-indicatingvariables including time since diabetes diagnosis,body mass index, smoking history, educational attainment,and self-identified race/ethnicity. This subanalysisresulted in no substantive differences andyielded the same conclusions suggested by analysis ofthe full cohort.







Of subjects treated with the 11 most commonly usedtherapeutic regimens, those who received the insulinsensitizers (thiazolidinediones and metformin) weremost likely to achieve good control, particularly whenthese agents were used in combination with insulin oranother oral agent. Treatments that were significantlymore effective than the first-line therapy, sulfonylureamonotherapy (12.46% achieved good control), weremetformin monotherapy (17.1% achieved good control;= .04), sulfonylurea plus thiazolidinedione (24.6%achieved good control; = .002), metformin plusinsulin (24.9% achieved good control; = .0008), thiazolidinedioneplus insulin (25.7% achieved good control;= .007), and the triple combination thiazolidinedioneplus metformin plus sulfonylurea (25.1% achieved goodcontrol; = .0007). No therapy was significantly lesseffective than sulfonylurea monotherapy.

Despite the impressive response to initiated therapy,82% of the new users failed to achieve good control and54% percent still had HbA1c measurements that exceeded 8% (the level at which ADA recommends action) 3 to12 months after initiation, and before subsequent therapychanges. This suggests the importance of intensivefollow-up after initiating new therapy and prompt therapyintensification when needed. Forty percent of thisnew-user cohort had additional therapy modificationswithin the year after therapy initiation, suggesting thatproviders are tracking therapy response and takingrapid action. Additionally, behavioral factors, includingSMBG frequency and outpatient appointment attendance,were strongly predictive of good control.

Our study, one of the few assessments of real-worldeffectiveness, compared all currently available diabetespharmacotherapies within a single population.In this observational study, patients did not achievethe level of control reported in randomized clinical trials.However, it is important to note that our studystipulated poor baseline control as an eligibility criterion.In the United Kingdom Prospective DiabetesStudy (UKPDS), a randomized, controlled trial (RCT)with arms including intensive regimens of behavioral,pharmacological, and diet therapy, ~50% (47%-52% forany medication) of patients randomized to eitherinsulin or sulfonylurea monotherapy maintained goodcontrol after 3 years, but the proportion declined progressivelyto 20% to 28% after 9 years.19 It is unclearhow much of the efficacy in the UKPDS was attributableto the additional clinical attention common in clinicaltrials. A UKPDS substudy of patients allocated totreatment with sulfonylurea monotherapy reportedthat inadequacy of this therapy contributed to theprogressive failure; 53% required additional insulintherapy within 6 years of follow-up.20

Randomized, controlled trials are considered thegold standard for evidence of efficacy, but they havelimitations. These trials include highly selected populationsreceiving special clinical attention and usuallyevaluate a single medication rather than multiple medicationtherapies.21,22 Although these are necessary constraintsin experimental settings, they limitgeneralizability.23 Moreover, low levels of medicationadherence may explain why the effectiveness observedin clinical practice24 usually falls short of the efficacydemonstrated in RCTs. For this reason, there is a growingskepticism regarding RCT results, creating a barrierto early adoption of new evidence-based recommendations.25 Thus, real-world effectiveness studies provideimportant complementary information.26,27

The observed proportion of patients achieving goodglycemic control is lower than the proportion achievingthe recommended level of control for other chronicconditions (eg, hypertension, hyperlipidemia). Estimatesof the proportion of hypertensive patients whoachieved well-controlled blood pressure (≤140/90)ranged from 27% to 61% (27% in the US National Healthand Nutrition Examination Survey III, 1991-1994; 35%in the New York managed care sample, 199828; 61% inthe metropolitan New York City sample, 199929),although a rate of 74% has been achieved in RCTs.30Estimates of good lipid control (low-density lipoproteincholesterol ≤100 mg/dL) range from 41.7%31 to88.5%.32 This suggests the currently available pharmacotherapiesfor hypertension and dyslipidemia mayhave greater relative effectiveness than antihyperglycemicagents.

Previous studies have shown that initiation of newdiabetes therapies (switching or augmenting) occursfrequently, perhaps driven by the need for intensificationof therapy.26 In this study population, ~27% ofthe diabetic patients initiated new therapy regimensduring the 1-year observation period. In the cohortwe studied (new users with poorly controlled type 2diabetes), 5% discontinued therapy and 41% modifiedtherapy (switched or augmented the index therapy)within 12 months. The UKPDS demonstrated that diabetesis a progressive disorder, requiring a stepwisetherapy intensification, with transitions from diet tomonotherapy, to combination therapy, and eventuallyto insulin.19 UKPDS findings also suggested that mostpatients fail to respond to sulfonylurea therapy as betacell dysfunction increases, and suggested the need toadd insulin or other therapies long before maximaldoses become inadequate.20 We observed a dramaticshift from monotherapy-dominated regimens used in theprebaseline period to the use of combination therapies.This trend toward diabetes polypharmacy, particularlywith inclusion of insulin sensitizers to address insulinresistance, has been noted in other populations.4,33

Some limitations are worth noting. We were unableto assess differences in effectiveness among patientswho had not completed a HbA1c test during the 3-to12-month window after initiation. Among subjectswho discontinued the initiated therapy within 12months, we only included measures of glycemic controlassayed before the point of therapy modificationor discontinuation. Thus, this was not an intent-to-treatanalysis that included all initiators. Therefore,our study slightly underestimated the proportion ofsubjects failing to achieve control relative to a cohortthat would have included the ~5% of subjects whodiscontinued their index therapies before having aHbA1c test. The mean post-initiation HbA1c valuewould have been slightly higher (8.7% rather than8.6%) if we had included those few subjects who discontinuedtherapy before having a post-initiationHbA1c test. The percentage failing to achieve goodcontrol observed in this study thus may be viewed asconservative, further reinforcing the public healthmessage suggesting the importance of intensive postinitiationfollow-up. Additionally, we were unable toassess adherence to diet and exercise recommendations,both of which could play a role in achievingglycemic goals.

Thiazolidinedione use was low (9.5% of our newusers) during the study observation period, but hasincreased steadily since its introduction to the Kaiserformulary in April 1997. This low usage was notbecause of a patient financial barrier associated withbranded therapies. All patients included in this studyhad single-tier drug benefits; thus, there was no out-of-pocketcost difference between initiating therapy withone versus another of the therapeutic classes.

Between-therapy comparisons of the proportionsachieving good glycemic control are not interpretableas causal effects of therapy, as in the case of clinicaltrial results. We assumed that observed therapy initiationsoccurred for a variety of reasons, including (1) theprovider decided to prescribe a new therapy because oflow effectiveness, side effects, or lack of medication-takingcompliance with the preceding therapy; or (2)the patient discontinued the preceding therapy of hisor her own accord. We were unable to distinguishbetween these causes. Case-mix differences due toassociation between diabetes severity and choice oftreatment distort findings (confounding by indication34),so that more intense therapy is associated withpoorer glycemic control. Although controlling for preinitiationHbA1c values, previous therapy, and durationof diabetes did alter our findings, in some cases dramatically,additional residual confounding was expected.Nonetheless, both unadjusted and adjustedestimates of real-world effectiveness associated withpharmacotherapy in a clinical setting provide animportant benchmark for evaluation, given the dauntingarray of therapies and their combinations availablefor patients with diabetes.

Several unique strengths of this study are worthmentioning. These findings come from a large sourcepopulation (more than 3 million patients), which isalmost one third of the population of NorthernCalifornia. The rich data available in the Kaiser electronicrecords facilitated statistical adjustment for confoundingvariables usually unavailable in claimsdatabases. Levels of control in our diabetes populationare like those in other published studies, which likelymakes our findings generalizable to insured individualswith diabetes. The proportion in good control in thewhole Kaiser diabetes population (30%) is consistentwith that reported by other studies: 32% of patientswere reported in good control in the Type 2 DiabetesPatient Outcomes Research Team study,35 and 26.5% ofinsulin-treated and 37.7% of oral agent-treated patientswere reported in good control in the US National Healthand Nutrition Examination Survey III.36

If, as we observed, the majority of patients need toachieve better control, this suggests that new pharmacotherapeuticmodalities with greater effectiveness areneeded, but also points out the existence of provider-andpatient-related barriers to achieving control. Amultifactorial approach that integrates pharmacologicoptions with patient self-management, clinical initiative,and social support has been shown to provide optimummanagement of diabetes.37 We found evidence ofthe importance of patient behaviors: significantly betterglycemic control was associated with frequentSMBG. Previous studies in this same population indicatedthat this self-management practice was underutilized,13 despite being associated with better glucosecontrol.8 We also noted that frequently missed medicalappointments were associated with poorer control inthis and a previous study in this population,11 highlightingthe importance of continuity of care andpatient adherence factors. The efficacy of behavioralinterventions focusing on diet and exercise for patientswith diabetes has been demonstrated previously,38,39,40-43 although the effectiveness is frequently limited bylow levels of adherence. Exploration of novel behavioralapproaches (eg, stress management44) also mayprove useful.

Clinical inertia (ie, failure of healthcare providers toinitiate or intensify therapy when indicated) has beenidentified as a significant obstacle to effective diseasemanagement.25,45,46 A previous study based on a populationwith a similar form of integrated care47 showed thatbefore new therapies were initiated, levels of HbA1c weretypically closer to 9% rather than 8%. We had similarfindings. After expanding the analysis to the whole diabeticpopulation (not just the new users with poor controlwe selected for this study), the mean baseline HbA1c was 8.7% before initiating new therapy. Earlier initiationof new therapy (ie, before HbA1c greatly exceeded 7%)likely would have resulted in a larger proportion ofpatients being brought into good control.

Before 2003, the ADA guidelines stipulated an actionlevel (HbA1c = 8%), above which therapy intensificationwas recommended. This action level was 1 percentagepoint higher than the glycemic target (HbA1c = 7%) alsorecommended by the ADA. This gap between targetand action level created a gray zone that could havepotentially reduced the clinical attention given to patientswith HbA1c between 7% and 8%. The eliminationof the action level from the ADA recommendations(2003 and after) may stimulate prompter intensificationof therapy for patients with borderline HbA1c (7%-8%).Moreover, in addition to specifying the glycemic targetof HbA1c &#8804;7%, ADA now further recommends that evenmore stringent goals (<6%) should be considered on anindividual basis, given that epidemiologic evidencehas failed to detect a limit below which further loweringdoesn't confer clinical benefits. The American Associationof Clinical Endocrinologists (AACE) and theAmerican College of Endocrinology recommend a goalof "HbA1c level of 6.5% or less" in their MedicalGuidelines for the Management of Diabetes Mellitus:The AACE System of Intensive Diabetes Self-Management. The Action to Control CardiovascularRisk in Diabetes (ACCORD) study currently is evaluatingthe risks and benefits of such near-normalization ofblood sugars.

It is worth noting that, since the time of this study,the proportion of patients achieving good glycemic controlhas increased steadily in the source population forthis study (Kaiser Permanente Northern CaliforniaDiabetes Registry) from around 30% during the study(1999-2000) to more than 50% in 2004 (unpublisheddata). This favorable trend is likely attributable to moreaggressive therapy intensification and increased use ofcombination therapy and insulin-sensitizing agents.


Among patients with poorly controlled diabetes, initiationof combination therapies that included thiazolidinedioneor a regimen of metformin plus insulinresulted in the highest proportion of patients achievinggood glycemic control, while monotherapy sulfonylureasresulted in the lowest proportion. Patient self-management,particularly SMBG and appointment-keepingbehavior, also played an important role. However, themajority of patients still had suboptimal glycemic control3 to 12 months after initiating even the most effectivetreatment options. This suggests the need forincreased vigilance among providers to promptly identifyfailures to achieve good control after initiating newtherapies and aggressive stepwise therapy intensificationwhen initial treatments fail.47,48


We would like to thank Avanish Mishra, PharmD, PhD, for helpful discussionsregarding conceptualization of this manuscript.

From The Division of Research, Kaiser Permanente, Oakland, Calif.

This study was supported by Pfizer Pharmaceuticals Group, Aventis Pharmaceuticals,The American Diabetes Association, and Kaiser Foundation Research Institute.

Address correspondence to: Andrew J. Karter, PhD, The Division of Research, KaiserPermanente, 2000 Broadway, Oakland, CA 94612. E-mail:

Diabetes Care.

1. Diabetes Control and Complications Trial (DCCT): results of feasibility study.The DCCT Research Group. 1987;10:1-19.

Diabetes Res Clin Pract.

2. Ohkubo Y, Kishikawa H, Araki E, et al. Intensive insulin therapy prevents theprogression of diabetic microvascular complications in Japanese patients withnon-insulin-dependent diabetes mellitus: a randomized prospective 6-year study.1995;28:103-117.

Diabetes Care.

3. Wysowski DK, Armstrong G, Governale L. Rapid increase in the use of oralantidiabetic drugs in the United States, 1990-2001. 2003;26:1852-1855.


4. Inzucchi SE. Oral antihyperglycemic therapy for type 2 diabetes: scientificreview. 2002;287:360-372.

Arch Intern Med.

5. Johnson JL, Wolf SL, Kabadi UM. Efficacy of insulin and sulfonylurea combinationtherapy in type II diabetes. A meta-analysis of the randomized placebo-controlledtrials. 1996;156:259-264.

Diabetes Care.

6. American Diabetes Association. Clinical practice recommendations. 1999.1999;22(suppl 1).

Am J Public Health.

7. Krieger N. Overcoming the absence of socioeconomic data in medical records:validation and application of a census-based methodology. 1992;82:703-710.

Am J Med.

8. Karter AJ, Ackerson LM, Darbinian JA, et al. Self-monitoring of blood glucoselevels and glycemic control: the Northern California Kaiser Permanente Diabetesregistry. 2001;111:1-9.

Am J Public Health.

9. Hiatt RA, Friedman GD. Characteristics of patients referred for treatment ofend-stage renal disease in a defined population. 1982;72:829-833.


10. Karter AJ, Ferrara A, Liu JY, Moffet HH, Ackerson LM, Selby JV. Ethnic disparitiesin diabetic complications in an insured population. 2002;287:2519-2527.

Med Care.

11. Karter AJ, Parker MM, Moffet HH, et al. Missed appointments and poorglycemic control: an opportunity to identify high-risk diabetic patients. 2004;42:110-115.

Diabetes Care.

12. Selby JV, Karter AJ, Ackerson LM, Ferrara A, Liu J. Developing a predictionrule from automated clinical databases to identify high-risk patients in a large populationwith diabetes. 2001;24:1547-1555.

Diabetes Care.

13. Karter AJ, Ferrara A, Darbinian J, Ackerson LM, Selby JV. Self-monitoring ofblood glucose: language and financial barriers in a managed care population withdiabetes. 2000;23:477-483.

Diabetes Care.

14. Karter AJ, Ahmed AT, Liu J, et al. Use of thiazolidinediones and risk of heartfailure in people with type 2 diabetes: a retrospective cohort study. Response toDelea et al. 2004;27:850-851.

Am J Epidemiol.

15. Ray WA. Evaluating medication effects outside of clinical trials: new-userdesigns. 2003;158:915-920.

16. Bristol-Myers Squibb Company. Glucophage (metformin hydrochloride)tablets. Prescribing information [package insert]. Princeton, NJ: Bristol-MyersSquibb Co; 2004.

Modern Epidemiology.

17. Rothman KJ, Greenland S. 1st ed. Philadelphia, Pa:Lippincott-Raven; 1998.

Am J Epidemiol.

18. Salas M, Hofman A, Stricker BH. Confounding by indication: an example ofvariation in the use of epidemiologic terminology. 1999;149:981-983.


19. Turner RC, Cull CA, Frighi V, Holman RR. Glycemic control with diet, sulfonylurea,metformin, or insulin in patients with type 2 diabetes mellitus: progressiverequirement for multiple therapies (UKPDS 49). UK Prospective DiabetesStudy (UKPDS) Group. 1999;281:2005-2012.

Diabetes Care.

20. Wright A, Burden AC, Paisey RB, Cull CA, Holman RR. Sulfonylurea inadequacy:efficacy of addition of insulin over 6 years in patients with type 2 diabetes in theUK Prospective Diabetes Study (UKPDS 57). 2002;25:330-336.


21. Kramer MS, Shapiro SH. Scientific challenges in the application of randomizedtrials. 1984;252:2739-2745.


22. Gurwitz JH, Col NF, Avorn J. The exclusion of the elderly and women fromclinical trials in acute myocardial infarction. 1992;268:1417-1422.



23. Riegelman R, Verme D, Rochon J, El-Mohandes A. Interaction and interventionmodeling: predicting and extrapolating the impact of multiple interventions. 2002;12:151-156.

Med Care.

24. Robin DM, Giordani PJ, Lepper HS, Croghan TW. Patient adherence andmedical treatment outcomes: a meta-analysis. 2002;40:794-811.

Ann Intern Med.

25. Phillips LS, Branch WT, Cook CB, et al. Clinical inertia. 2001;135:825-834.

Diabetes Care.

26. Boccuzzi SJ, Wogen J, Fox J, Sung JC, Shah AB, Kim J. Utilization of oralhypoglycemic agents in a drug-insured US population. 2001;24:1411-1415.


27. Colwell JA. Controlling type 2 diabetes: are the benefits worth the costs?1997;278:1700.

Am J Manag Care.

28. DiTusa L, Luzier AB, Jarosz DE, Snyder BD, Izzo JLJ. Treatment of hypertensionin a managed care setting. 2001;7:520-524.


29. Cheng JW, Kalis MM, Feifer S. Patient-reported adherence to guidelines ofthe Sixth Joint National Committee on Prevention, Detection, Evaluation, andTreatment of High Blood Pressure. 2001;21:828-841.

Curr Control Trials Cardiovasc Med.

30. Thijs L, Staessen JA, Beleva S, et al. How well can blood pressure be controlled?Progress report on the Systolic Hypertension in Europe Follow-Up Study(Syst-Eur 2). 2001;2:298-306.

Am J Cardiol.

31. Insull W, Kafonek S, Goldner D, Zieve F. Comparison of efficacy and safetyof atorvastatin (10mg) with simvastatin (10mg) at six weeks. ASSET Investigators.2001;87:554-559.

Clin Ther.

32. Chung N, Cho SY, Choi DH, et al. STATT: a titrate-to-goal study of simvastatinin Asian patients with coronary heart disease. Simvastatin Treats Asians toTarget. 2001;23:858-870.


33. Fonseca V, Rosenstock J, Patwardhan R, Salzman A. Effect of metformin androsiglitazone combination therapy in patients with type 2 diabetes mellitus: a randomizedcontrolled trial. 2000;283:1695-1702.

Med Care.

34. Poses RM, Smith WR, McClish DK, Anthony M. Controlling for confoundingby indication for treatment. Are administrative data equivalent to clinical data?1995;33:AS36-46.


35. Hayward RA, Manning WG, Kaplan SH, Wagner EH, Greenfield S. Startinginsulin therapy in patients with type 2 diabetes: effectiveness, complications, andresource utilization. 1997;278:1663-1669.

Diabetes Care.

36. Harris MI, Eastman RC, Cowie CC, Flegal KM, Eberhardt MS. Racial and ethnicdifferences in glycemic control of adults with type 2 diabetes. 1999;22:403-408.



37. Wolpert HA, Anderson BJ. Metabolic control matters: why is the message lostin the translation? The need for realistic goal-setting in diabetes care. 2001;24:1301-1303.

Diabetes Care.

38. Delahanty LM, Halford BN. The role of diet behaviors in achieving improvedglycemic control in intensively treated patients in the Diabetes Control andComplications Trial. 1993;16:1453-1458.

Acta Diabetol.

39. Domenech MI, Assad D, Mazzei ME, Kronsbein P, Gagliardino JJ. Evaluationof the effectiveness of an ambulatory teaching/treatment programme for noninsulindependent (type 2) diabetic patients. 1995;32:143-147.

N Engl J Med.

40. Hu FB, Manson JE, Stampfer MJ, et al. Diet, lifestyle, and the risk of type 2diabetes mellitus in women. 2001;345:790-797.


Engl J Med.

41. Tuomilehto J, Lindstrom J, Eriksson JG, et al. Prevention of type 2 diabetesmellitus by changes in lifestyle among subjects with impaired glucose tolerance. 2001;344:1343-1350.

Diabetes Care.

42. Evidence-Based Nutrition Principles and Recommendations for the Treatmentand Prevention of Diabetes and Related Complications. 2002;25:S50-S60.

Ann Intern Med.

43. Van Dam RM, Rimm EB, Willett WC, Stampfer MJ, Hu FB. Dietary patternsand risk for type 2 diabetes mellitus in US men. 2002;136:201-209.

Diabetes Care.

44. Surwit RS, van Tilburg MA, Zucker N, et al. Stress management improveslong-term glycemic control in type 2 diabetes. 2002;25:30-34.

Diabetes Care.

45. Cook CB, Ziemer DC, El-Kebbi IM, et al. Diabetes in urban African-Americans, XVI: overcoming clinical inertia improves glycemic control in patientswith type 2 diabetes. 1999;22:1494-1500.

Eff Clin Pract.

46. Mottur-Pilson C, Snow V, Bartlett K. Physician explanations for failing tocomply with "best practices." 2001;4:207-213.

Am J Manag Care.

47. Brown JB, Nichols GA. Slow response to loss of glycemic control in type 2diabetes mellitus. 2003;9:213-217.

N Engl J Med.

48. Nathan DM. Clinical practice. Initial management of glycemia in type 2 diabetesmellitus. 2002;347:1342-1349.