Statins and the Risk of DiabetesEvidence From a LargePopulation-Based Cohort StudyDiabetes Care 2014372225ndash2232 | DOI 102337dc13-2215

OBJECTIVE

To investigate the relationship between adherence with statin therapy and therisk of developing diabetes

RESEARCH DESIGN AND METHODS

The cohort comprised 115709 residents of the Italian Lombardy region who werenewly treated with statins during 2003 and 2004 Patients were followed from theindex prescription until 2010 During this period patients who began therapy withan antidiabetic agent or were hospitalized for a main diagnosis of type 2 diabeteswere identified (outcome) Adherence was measured by the proportion of dayscovered (PDC) with statins (exposure) A proportional hazards model was fitted toestimate hazard ratios (HRs) and 95 CIs for the exposure-outcome associationafter adjusting for several covariates A set of sensitivity analyses was performedto account for sources of systematic uncertainty

RESULTS

During follow-up 11154 cohort members experienced the outcome Comparedwith patients with very-low adherence (PDC lt25) those with low (26ndash50)intermediate (51ndash75) and high (Dagger75) adherence to statin therapy had HRs(95 CIs) of 112 (106ndash118) 122 (114ndash127) and 132 (126ndash139) respectively

CONCLUSIONS

In a real-world setting the risk of new-onset diabetes rises as adherence withstatin therapy increases Benefits of statins in reducing cardiovascular eventsclearly overwhelm the diabetes risk

Hydroxymethylglutaryl-CoA reductase inhibitors known as statins are the mosteffective drugs for the treatment of hypercholesterolemia (1) and their importantrole in primary and secondary prevention of cardiovascular (CV) morbidity andmortality is well established (2) even among patients with type 2 diabetes (3)However during the past decade concern has been raised about the use of statinsand the development of diabetes (4) This did not appear to be the case in the posthoc analysis of one of the earliest trials on the CV protective effects of statins theWOSCOPS (West of Scotland Coronary Prevention Study) which suggested thatpravastatin might reduce the risk of diabetes (5) In the more recent JUPITER (Jus-tification for the Use of Statins in Primary Prevention an Intervention Trial Evalu-ating Rosuvastatin) trial rosuvastatin was found to be associated with an increasedrisk of physician-reported diabetes as well as with a small though significant in-crease in HbA1c levels (6) Two meta-analyses of clinical trials found that among

1Department of Statistics and QuantitativeMethods Division of Biostatistics Epidemiologyand Public Health University of Milano-BicoccaMilan Italy2Operative Unit of Territorial Health Services Re-gion Lombardia Milan Italy3Department of Pharmacological and Biomolec-ular Sciences Centre of Epidemiology and Pre-ventive Pharmacology (SEFAP) University ofMilano Milan Italy4IRCCS Multimedica Sesto San Giovanni MilanItaly5Department of Health Science University ofMilano-Bicocca Milan Italy6IRCCS Istituto Auxologico Italiano Milan Italy

Corresponding author Giovanni Corrao giovannicorraounimibit

Received 19 September 2013 and accepted 17April 2014

This article contains Supplementary Data onlineat httpcarediabetesjournalsorglookupsuppldoi102337dc13-2215-DC1

copy 2014 by the American Diabetes AssociationReaders may use this article as long as the workis properly cited the use is educational and notfor profit and the work is not altered

Giovanni Corrao1 Buthaina Ibrahim1

Federica Nicotra1 Davide Soranna1

Luca Merlino2 Alberico L Catapano34

Elena Tragni3 Manuela Casula3

Guido Grassi45 and Giuseppe Mancia56

Diabetes Care Volume 37 August 2014 2225

EPIDEM

IOLO

GYH

EALTH

SERVICES

RESEA

RCH

patients treated with statins the risk ofdeveloping diabetes was higher than inthose treated with placebo (78) Itshould be noted however that noneof these trials were designed to lookfor diabetes and that the meta-analysesused a range of methods to detect theconditionBecause statins are among the most

commonly prescribed drugs and diabe-tes affects a substantial and increasingproportion of the population world-wide efforts aimed at elucidating themagnitude of the association betweenuse of statins and diabetes risk have ma-jor implications for public health Toaddress this issue we used the data pro-vided by the health-care utilization da-tabases of Lombardy a region of Italywith 10 million inhabitants This largecohort study investigated whether in-creasing levels of adherence with statintherapy increases the risk of developingphysician-diagnosed diabetes and esti-mated the magnitude of the dose-response relationship Controlling forsources of systematic uncertainty wasof particular concern in this study

RESEARCH DESIGN AND METHODS

Health-Care Utilization Database ofLombardyThe data used for the current studywere retrieved from the health-care uti-lization databases of Lombardy a regionof Italy that accounts for16 of Italyrsquospopulation The Italian population iscovered by the National Health Service(NHS) and in Lombardy an automatedsystem of databases collects a variety ofinformation including 1) an archive ofresidents who receive NHS assistancereporting demographic and administra-tive data 2) a database on diagnosis atdischarge from public or private hospi-tals of the region 3) a database on out-patient drug prescriptions reimbursableby the NHS and 4) a database on out-patient visits including visits in special-ist ambulatory care and diagnosticlaboratories accredited by the NHS Foreach patient we linked these databasesthrough a single identification code Fulldetails of the procedures are reportedelsewhere (9)

Cohort Selection and Follow-upThe target population comprised all NHSbeneficiaries aged 40ndash80 years living inLombardy According to the 2011 Italiancensus this population was 5097075

individuals Of these we identified pa-tients for whom at least one prescrip-tion of statins was dispensed during2003 and 2004 and the first dispensa-tion was defined as the index prescrip-tion To make data relevant to the studyaim four patient categories were ex-cluded 1) patients who received oneor more prescriptions of a statin within3 years before the index prescription 2)patients who received at least one anti-diabetic agent or were hospitalized for aprimary or secondary diagnosis of dia-betes within 3 years before the indexprescription 3) patients who did notreach at least 1 year of follow-up and4) patients who received only one dis-pensation of statins during the first yearafter the date of index prescription Theremaining patients represented thestudy cohort Each member of the co-hort accumulated person-years offollow-up from the date of index pre-scription until the earliest among thedates of new-onset diabetes (see nextsection) or censoring (ie death fromany cause emigration or 31 December2010 [end of follow-up])

Outcome IdentificationCohort members who experienced atleast one sign suggestive of the occur-rence of diabetes during follow-up wereidentified The date of the first dispen-sation of an antidiabetic agent (Anatom-ical Therapeutic Chemical ClassificationSystem code A10) or of hospitalizationwith a primary diagnosis of diabetes(ICD-9 code 250) was assumed as thedate of outcome onset whichever wasearlier However because of the possi-bility that one isolated dispensation ofan antidiabetic agent might be an inap-propriate initial drug treatment in a sec-ondary analysis we used amore-specificdiagnostic criterion by requiring that atleast three antidiabetic prescriptions bedispensed to consider a patient as hav-ing the outcome

Assessment of Exposure to StatinsWe identified all prescriptions dis-pensed to the cohort members duringfollow-up The period covered by a pre-scription was calculated from the num-ber of tablets in the dispensed canistersassuming a treatment schedule of onetablet per day (10) For overlapping pre-scriptions the individual was assumedto have completed the former one be-fore starting the second Adherence to

therapy was assessed as the cumulativenumber of days during which the medi-cationwas available divided by the num-ber of days of follow-up a quantityreferred to as the proportion of dayscovered (PDC) (11) Patients were cate-gorized as having very-low (PDC25)low (PDC 25ndash49) intermediate (PDC50ndash74) and high (PDC $75) adher-ence To avoid the arbitrary nature ofthis categorization in a secondary analy-sis we used quartile categories of PDC forthe entire period of follow-up to defineincreasing levels of exposure to statins

PDC categories were also consideredaccording to the potency of the statinsdispensed during follow-up Based on asystematic review and meta-analysis ofrandomized controlled trials that quan-tified the effect of statins on serum LDLcholesterol concentration (12) treat-ment with high-potency statins was de-fined $10 mg rosuvastatin $20 mgatorvastatin and $40 mg simvastatinwhereas all other statin treatmentswere defined as low potency

CovariatesFor each cohort member data also in-cluded 1) sex age and type of statins(atorvastatin fluvastatin pravastatinrosuvastatin and simvastatin) at the in-dex prescription 2) concomitant use ofother drugs during follow-up (antihyper-tensive nitrate digitalis antiarrhythmiccorticosteroid oral contraceptive cyclo-sporine antipsychotic and antidepres-sant agents) 3) hospitalizations for CVdiseases in the 3 years before the indexprescription 4) the Charlson comorbidityindex score (13) which was calculatedfrom the diagnostic information availablefrom inpatient charts in the 3 years be-fore and 1 year after the date of the indexprescription and 5) the number of serumcholesterol tests and specialist visits dur-ing the first year after the date of the in-dex prescription

Data AnalysisThe x2 test was used for the trend andlinear regression model when appropri-ate to test differences in demographicand clinical characteristics with increas-ing levels of statin adherence The inci-dence rate of diabetes was calculatedby the ratio between the number ofcases of incident diabetes and the num-ber of person-years accumulated duringfollow-up

2226 Statins and the Risk of Diabetes Diabetes Care Volume 37 August 2014

Cox proportional hazards regressionwas used to estimate the hazard ratio(HR) and 95 CI for the association be-tween exposure to statins and time of di-abetes onset The predictor variables ofinterest were the factors constructed ac-cording to the categories of PDC using thevery-low-adherence category as refer-ence Because drug exposure may varyover time assessmentof its value requiresproper accounting for the cumulative andvarying nature of the measure This wasdone by including adherence categoriesas time-dependent variables in the modelThe relationship of the risk of diabe-

tes with increasing exposure to statinsas a whole as well as to low-potency andhigh-potency statins separately was as-sessed using both unadjusted and ad-justed HRs Adjustment included thesereported covariates as time-fixed (iethose measured at index prescription)or time-dependent (concomitant useof other drugs) factors Trends in HRswere tested when feasible according tothe statistical significance of the regres-sion coefficient of the recorded variablesobtained by scoring the correspondingcategories The null hypothesis of equal-ity in the regression coefficients of low-potency and high-potency statins wastested by the z test

Sensitivity AnalysesTo verify the robustness of the findingsthe following sensitivity analyses wereperformed 1) As mentioned previouslywe adopted various ways for categoriz-ing exposure to statins (predefined orquartile categories of PDC) or for iden-tifying incident diabetes (one to threeprescriptions of antidiabetic drugs dur-ing follow-up) and 2) we checked thepossibility that our estimates were af-fected by detection bias (14) (ie thatlong-term exposure to statins and amore-regular prescription renewalimply a more regular use of primarycare services triggering the search fordiabetes) Detection bias was investi-gated by estimating the statinndashincidentdiabetes relationship in the cohortmembers stratified according to thenumber of cholesterol tests and outpa-tient specialist visits during the first yearof follow-up The rationale is that therelationship would move toward thatexpected under the null among patientswho did not use laboratory analysisservices andor attend specialist visits

Finally although our estimates wereadjusted for a number of factors andbecause relevant clinical features werenot available in the databases we ad-dressed the potential bias generatedby unmeasured confounders For exam-ple we considered obesity as a poten-tially important unmeasured confounderbecause obesity is a major predictor ofdiabetes (15) and can influence adher-ence to statin therapy (1617) whichmeans that failing to adjust for its effectscan move the results toward or awayfrom the null We made a quantitativeassessment of the obesity-related biasby using the Monte Carlo sensitivity anal-ysis (18) We set 1) the prevalence of theobesity among users of statins to 375(19) 2) the risk of diabetes in obese pa-tients to threefold higher than in normalweight patients (15) and 3) the odds ofobese patients belonging to the highestPDC category from twofold lower to two-fold higher than that of normal weightpatients (1617) The Monte Carlo sensi-tivity analysis corrected the observed HRsfor the bias factor calculated from thepreviously mentioned data and tookinto account random uncertainty for ad-justing estimates Full details on theMonte Carlo sampling procedure andcomputing method are reported else-where (20) For all hypotheses tested atwo-tailed P 005 was consideredsignificant

RESULTS

PatientsThe distribution of the exclusion crite-ria is shown in Supplementary Fig 1The 115709 patients included in thestudy cohort accumulated 748049person-years of observation with anaverage follow-up of 64 years perpatient

Table 1 shows the characteristics ofthe entire cohort as well as stratificationby adherence to statin therapy At theindex date the mean age was 62years and 49 were men Almost onein three patients began therapy withatorvastatin or simvastatin Most pa-tients were cotreated with antihyper-tensive drugs whereas cotreatmentwith other drugs was less frequentMore than 30 of the cohort was hos-pitalized previously for CV diseasesOther chronic comorbidities were iden-tified in more than one in four patientsDuring the first year of follow-up almost

one in five patients and6 of patientsdid not have a serum cholesterol test orattend a specialist visit respectively

During follow-up 11154 cohortmembers experienced the study out-come with an incidence rate of 149new cases of diabetes per 1000 person-years More than one-half (573)had very-low or low adherence to statintherapy Very-low or low adherencewasmore common in women in patientswho began therapy with fluvastatinor pravastatin and among patients whodid not use other drugs were not hos-pitalized previously for CV disease ordid not have other chronic comorbiditiesPatients with a higher number of serumcholesterol tests and specialist visits hadhigher adherence to statin therapy

Adherence to Statin Therapy and Riskof DiabetesThere was a continuous and significanttrend toward an increase of diabetesrisk as adherence to statin therapy in-creased both in the unadjusted and inthe adjusted risk models (58 [95CI 51ndash66] vs 32 [26ndash39] for highadherence vs very-low adherence re-spectively) (Fig 1A and B) Althoughhigh-potency statins showed a greateraction on diabetes risk than low-potencystatins (Fig 1C and D) there was no sta-tistical evidence that the effect of statinson diabetes risk differed according totheir potency (Ptrend = 0372)

The relationship did not substantiallychange by 1) varying the criteria for cat-egorization of adherence with statintherapy or 2) considering a more de-manding criterion for diabetes diagno-sis In particular taking the first categoryof exposure as the reference the HR(95 CI) of diabetes increased to 111(100ndash123) 123 (116ndash131) and 137(129ndash145) by increasing quartiles ofPDC and to 119 (111ndash127) 135(126ndash145) and 153 (144ndash164) bythe alternative criterion for diabetesdiagnosis

Detection BiasThe combined action of adherence tostatin therapy and frequency of choles-terol tests and outpatient specialist vis-its on the risk of diabetes is shown in Fig2 Among patients with very-low adher-ence to statin therapy the risk of diabe-tes decreased with the number of serumcholesterol tests but increased as thenumber of specialist visits increased

carediabetesjournalsorg Corrao and Associates 2227

Cox proportional hazards regressionwas used to estimate the hazard ratio(HR) and 95 CI for the association be-tween exposure to statins and time of di-abetes onset The predictor variables ofinterest were the factors constructed ac-cording to the categories of PDC using thevery-low-adherence category as refer-ence Because drug exposure may varyover time assessmentof its value requiresproper accounting for the cumulative andvarying nature of the measure This wasdone by including adherence categoriesas time-dependent variables in the modelThe relationship of the risk of diabe-

tes with increasing exposure to statinsas a whole as well as to low-potency andhigh-potency statins separately was as-sessed using both unadjusted and ad-justed HRs Adjustment included thesereported covariates as time-fixed (iethose measured at index prescription)or time-dependent (concomitant useof other drugs) factors Trends in HRswere tested when feasible according tothe statistical significance of the regres-sion coefficient of the recorded variablesobtained by scoring the correspondingcategories The null hypothesis of equal-ity in the regression coefficients of low-potency and high-potency statins wastested by the z test

Sensitivity AnalysesTo verify the robustness of the findingsthe following sensitivity analyses wereperformed 1) As mentioned previouslywe adopted various ways for categoriz-ing exposure to statins (predefined orquartile categories of PDC) or for iden-tifying incident diabetes (one to threeprescriptions of antidiabetic drugs dur-ing follow-up) and 2) we checked thepossibility that our estimates were af-fected by detection bias (14) (ie thatlong-term exposure to statins and amore-regular prescription renewalimply a more regular use of primarycare services triggering the search fordiabetes) Detection bias was investi-gated by estimating the statinndashincidentdiabetes relationship in the cohortmembers stratified according to thenumber of cholesterol tests and outpa-tient specialist visits during the first yearof follow-up The rationale is that therelationship would move toward thatexpected under the null among patientswho did not use laboratory analysisservices andor attend specialist visits

Finally although our estimates wereadjusted for a number of factors andbecause relevant clinical features werenot available in the databases we ad-dressed the potential bias generatedby unmeasured confounders For exam-ple we considered obesity as a poten-tially important unmeasured confounderbecause obesity is a major predictor ofdiabetes (15) and can influence adher-ence to statin therapy (1617) whichmeans that failing to adjust for its effectscan move the results toward or awayfrom the null We made a quantitativeassessment of the obesity-related biasby using the Monte Carlo sensitivity anal-ysis (18) We set 1) the prevalence of theobesity among users of statins to 375(19) 2) the risk of diabetes in obese pa-tients to threefold higher than in normalweight patients (15) and 3) the odds ofobese patients belonging to the highestPDC category from twofold lower to two-fold higher than that of normal weightpatients (1617) The Monte Carlo sensi-tivity analysis corrected the observed HRsfor the bias factor calculated from thepreviously mentioned data and tookinto account random uncertainty for ad-justing estimates Full details on theMonte Carlo sampling procedure andcomputing method are reported else-where (20) For all hypotheses tested atwo-tailed P 005 was consideredsignificant

RESULTS

PatientsThe distribution of the exclusion crite-ria is shown in Supplementary Fig 1The 115709 patients included in thestudy cohort accumulated 748049person-years of observation with anaverage follow-up of 64 years perpatient

Table 1 shows the characteristics ofthe entire cohort as well as stratificationby adherence to statin therapy At theindex date the mean age was 62years and 49 were men Almost onein three patients began therapy withatorvastatin or simvastatin Most pa-tients were cotreated with antihyper-tensive drugs whereas cotreatmentwith other drugs was less frequentMore than 30 of the cohort was hos-pitalized previously for CV diseasesOther chronic comorbidities were iden-tified in more than one in four patientsDuring the first year of follow-up almost

one in five patients and6 of patientsdid not have a serum cholesterol test orattend a specialist visit respectively

During follow-up 11154 cohortmembers experienced the study out-come with an incidence rate of 149new cases of diabetes per 1000 person-years More than one-half (573)had very-low or low adherence to statintherapy Very-low or low adherencewasmore common in women in patientswho began therapy with fluvastatinor pravastatin and among patients whodid not use other drugs were not hos-pitalized previously for CV disease ordid not have other chronic comorbiditiesPatients with a higher number of serumcholesterol tests and specialist visits hadhigher adherence to statin therapy

Adherence to Statin Therapy and Riskof DiabetesThere was a continuous and significanttrend toward an increase of diabetesrisk as adherence to statin therapy in-creased both in the unadjusted and inthe adjusted risk models (58 [95CI 51ndash66] vs 32 [26ndash39] for highadherence vs very-low adherence re-spectively) (Fig 1A and B) Althoughhigh-potency statins showed a greateraction on diabetes risk than low-potencystatins (Fig 1C and D) there was no sta-tistical evidence that the effect of statinson diabetes risk differed according totheir potency (Ptrend = 0372)

The relationship did not substantiallychange by 1) varying the criteria for cat-egorization of adherence with statintherapy or 2) considering a more de-manding criterion for diabetes diagno-sis In particular taking the first categoryof exposure as the reference the HR(95 CI) of diabetes increased to 111(100ndash123) 123 (116ndash131) and 137(129ndash145) by increasing quartiles ofPDC and to 119 (111ndash127) 135(126ndash145) and 153 (144ndash164) bythe alternative criterion for diabetesdiagnosis

Detection BiasThe combined action of adherence tostatin therapy and frequency of choles-terol tests and outpatient specialist vis-its on the risk of diabetes is shown in Fig2 Among patients with very-low adher-ence to statin therapy the risk of diabe-tes decreased with the number of serumcholesterol tests but increased as thenumber of specialist visits increased

carediabetesjournalsorg Corrao and Associates 2227

Of note evidence of a trend toward in-creasing risk of diabetes as adherence tostatin therapy increased was also ob-served in patients who did not receivecholesterol tests (top panel) or outpa-tient specialist visits (bottom panel)

Unmeasured Confounding

Figure 3 shows the risk of diabetes associ-atedwith treatment adherence (exposure-outcome HR) after adjustment for anunmeasured confounder expected tobe from twofold less to twofold morefrequent among patients with high

treatment adherence (confounder-exposure odds ratio [OR]) As expectedthe trend in exposure-outcome HR wasprogressively driven toward the null asthe confounder-exposure OR increased(ie as the confounder prevalence pro-gressively increased among patientswith high adherence) The significantexposure-outcome excess risk associ-ated with high adherence observed inthe main analysis became nonsignifi-cant (113 [100ndash128]) whether theconfounder-exposure OR was $184(eg obese patients in the lowest

adherence category had a prevalenceodds 184-fold higher than those in thehighest adherence category) Howeverthe point estimate of exposure-outcomeassociation was never annulled even forthe highest considered disparity of treat-ment adherence between obese andnormal weight patients

CONCLUSIONS

The results show that compared withpatients who used statins for a smallportion of the follow-up period thosewho exhibited a more continuous use

Table 1mdashSelected characteristics of cohort members according to categories of adherence to statin therapy

PDC category

25 25ndash50 50ndash75 $75 Entire cohort

All patients 42921 (371) 23357 (202) 21251 (184) 28180 (244) 115709 (100)

Age mean (SD) 620 (99) 627 (93) 625 (91) 627 (90) 624 (94)

Male sex 17425 (406) 10533 (451) 10873 (512) 17427 (618) 56258 (486)

First-line therapy with statinsAtorvastatin 13601 (317) 7344 (314) 6859 (323) 9938 (353) 37742 (326)Fluvastatin 5311 (124) 2224 (95) 2105 (99) 2613 (93) 12253 (106)Pravastatin 7539 (176) 4173 (179) 3627 (171) 4135 (147) 19474 (168)Rosuvastatin 3507 (82) 2040 (87) 1811 (85) 2238 (79) 9596 (83)Simvastatin 12963 (302) 7576 (324) 6849 (322) 9256 (328) 36644 (317)

Concomitant use of other drugsAntihypertensive drugsACE inhibitors 16935 (395) 10588 (453) 10701 (504) 16271 (577) 54495 (471)ARBs 11572 (270) 6935 (297) 6663 (314) 8933 (317) 34103 (295)CCBs 11756 (274) 7390 (316) 7184 (338) 10738 (381) 37068 (320)b-Blockers 13211 (308) 9335 (400) 9819 (462) 15881 (564) 48246 (417)Thiazides 171 (04) 121 (05) 116 (05) 187 (07) 595 (05)Others 10394 (242) 6400 (274) 6120 (288) 9081 (322) 31995 (277)

Nitrates andor digitalis 4265 (99) 3391 (145) 3915 (184) 7751 (275) 19322 (167)Antiarrhythmics 2595 (60) 1772 (76) 1725 (81) 2770 (98) 8862 (77)High-dose corticosteroids 6755 (157) 3693 (158) 3237 (152) 4224 (150) 17909 (155)Oral contraceptives 2268 (53) 1148 (49) 917 (43) 860 (31) 5193 (45)Cyclosporine 147 (03) 81 (03) 75 (04) 123 (04) 426 (04)Antipsychotics 1082 (25) 487 (21) 357 (17) 470 (17) 2396 (21)Antidepressants 7855 (183) 4177 (179) 3580 (168) 4399 (156) 20011 (173)

History of CV disease 10848 (253) 7622 (326) 8168 (384) 14967 (531) 41607 (360)

Charlson comorbidity index score0 35712 (832) 18200 (779) 15515 (730) 16857 (598) 86284 (746)1 3716 (87) 2844 (122) 3350 (158) 6898 (245) 16808 (145)2 1985 (46) 1310 (56) 1321 (62) 2487 (88) 7103 (61)$3 1508 (35) 1003 (43) 1065 (50) 1938 (69) 5514 (48)

Number of cholesterolemia tests0 11639 (271) 4760 (204) 4094 (193) 5696 (202) 26189 (226)1 15365 (358) 8252 (353) 7272 (342) 9725 (345) 40614 (351)2 9966 (232) 6331 (271) 5971 (281) 7594 (269) 29862 (258)$3 5951 (139) 4014 (172) 3914 (184) 5165 (183) 19044 (165)

Number of outpatient specialist visits0 3310 (77) 1424 (61) 1151 (54) 1319 (47) 7204 (62)1 4091 (95) 1923 (82) 1597 (75) 2024 (72) 9635 (83)2 4390 (102) 2226 (95) 1927 (91) 2432 (86) 10975 (95)$3 31130 (725) 17784 (761) 16576 (780) 22405 (795) 87895 (760)

New cases of diabetesn 3128 2463 2260 3303 11154Rate (every 1000 person-years) 111 165 166 184 149

Data are n () unless otherwise indicated ARB angiotensin receptor blocker CCB calcium channel blocker

2228 Statins and the Risk of Diabetes Diabetes Care Volume 37 August 2014

of statins had a 32 excess risk of new-onset diabetes Furthermore the resultsshow that an increased risk of new-onset diabetes is related to all availablestatins at all doses The results failed toshow however significant evidencethat high-potency statins exert a stron-ger action on diabetes developmentthan low-potency statins In summarythis population-based study supportsthe notion that continuous use of statinsis associated with a nonmarginal in-crease in the risk of developing diabetesin a real-life setting

Comparison With Available EvidenceThe current findings confirm and extendthe results of the JUPITER trial whichwas the first to find that new-onset di-abetes was 27 more common in pa-tients treated with statins than inthose receiving placebo (5) as well asthe results of the PROSPER (ProspectiveStudy of Pravastatin in the Elderly at Risk)trial which showed that new-onset dia-betes increased by 30 in the pravastatingroup compared with the placebo group(21) Even more clear-cut risk excesses(+20 and +48) have been reportedfrom observational investigations (81923)

Meta-analyses of clinical trials howevershowed a weaker increased risk of new-onset diabetes by10 (7823) Finallythe SPARCL (Stroke Prevention by Ag-gressive Reduction in Cholesterol Levels)trial showed that high-dose atorvastatintreatment comparedwith placebo is associ-atedwith a 37excess risk of diabetes (24)

The current findings differ from thoseof a large cohort study bolstering thesuggestion that statin-induced diabetesis likely a class effect (8) and of a meta-analysis of clinical trials showing that in-tensive- versus moderate-dose statintherapy is associated with a greaterrisk of diabetes (25) Two large random-ized trials however consistently showedthat atorvastatin 80mg daily is associatedwith a weak and nonsignificant excess ofdiabetes compared with atorvastatin10 mg daily (HR 110 [094ndash129]) or sim-vastatin 20 mg daily (HR 119 [098ndash143]) (24)

PlausibilityThe mechanisms underlying the diabe-togenic influence of statins are incom-pletely understood Atorvastatin andsimvastatin have been shown to de-crease glucose uptake in adipocyte cell

lines (26) and insulin sensitivity (27) re-spectively Simvastatin and atorvastatinhave been shown to decrease insulin se-cretion in b-cells (28) The inhibition ofisoprenoid synthesis may explain someof the observed dysglycemic effects ofstatins (26) Finally it has been hypoth-esized that statins may influence muscleor liver insulin sensitivity directly butthere is no specific experimental evi-dence to support this hypothesis (29)

Strengths and WeaknessesThe current study is unique in severalrespects First the investigation wasbased on data from a very large unse-lected population which was made pos-sible because of a cost-free health-caresystem for virtually all Italian citizens Sec-ond the drug prescription database pro-vides highly accurate data because reportof prescriptions by the pharmacies is es-sential for reimbursement andfiling of anincorrect report about dispensed drugshas legal consequences (30) Third pa-tients were identified from the point ofthe initial lipid-lowering therapy and thecomplete sequence of the subsequentprescriptions for statins was availableFourth we were able to include patients

Figure 1mdashEffect of adherence to statin therapy on the HRs for diabetes HR was estimated according to Cox proportional hazards model A and BCrude and adjusted estimates of all statins C and D Adjusted estimates of high-potency and low-potency statins Adjustments were made for age(continuous) sex first-line statin therapy concomitant use of other drugs history of CV disease and categories of the Charlson comorbidity indexscore (see Table 1 for the complete list of covariates included in the model)

carediabetesjournalsorg Corrao and Associates 2229

of statins had a 32 excess risk of new-onset diabetes Furthermore the resultsshow that an increased risk of new-onset diabetes is related to all availablestatins at all doses The results failed toshow however significant evidencethat high-potency statins exert a stron-ger action on diabetes developmentthan low-potency statins In summarythis population-based study supportsthe notion that continuous use of statinsis associated with a nonmarginal in-crease in the risk of developing diabetesin a real-life setting

Comparison With Available EvidenceThe current findings confirm and extendthe results of the JUPITER trial whichwas the first to find that new-onset di-abetes was 27 more common in pa-tients treated with statins than inthose receiving placebo (5) as well asthe results of the PROSPER (ProspectiveStudy of Pravastatin in the Elderly at Risk)trial which showed that new-onset dia-betes increased by 30 in the pravastatingroup compared with the placebo group(21) Even more clear-cut risk excesses(+20 and +48) have been reportedfrom observational investigations (81923)

Meta-analyses of clinical trials howevershowed a weaker increased risk of new-onset diabetes by10 (7823) Finallythe SPARCL (Stroke Prevention by Ag-gressive Reduction in Cholesterol Levels)trial showed that high-dose atorvastatintreatment comparedwith placebo is associ-atedwith a 37excess risk of diabetes (24)

The current findings differ from thoseof a large cohort study bolstering thesuggestion that statin-induced diabetesis likely a class effect (8) and of a meta-analysis of clinical trials showing that in-tensive- versus moderate-dose statintherapy is associated with a greaterrisk of diabetes (25) Two large random-ized trials however consistently showedthat atorvastatin 80mg daily is associatedwith a weak and nonsignificant excess ofdiabetes compared with atorvastatin10 mg daily (HR 110 [094ndash129]) or sim-vastatin 20 mg daily (HR 119 [098ndash143]) (24)

PlausibilityThe mechanisms underlying the diabe-togenic influence of statins are incom-pletely understood Atorvastatin andsimvastatin have been shown to de-crease glucose uptake in adipocyte cell

lines (26) and insulin sensitivity (27) re-spectively Simvastatin and atorvastatinhave been shown to decrease insulin se-cretion in b-cells (28) The inhibition ofisoprenoid synthesis may explain someof the observed dysglycemic effects ofstatins (26) Finally it has been hypoth-esized that statins may influence muscleor liver insulin sensitivity directly butthere is no specific experimental evi-dence to support this hypothesis (29)

Strengths and WeaknessesThe current study is unique in severalrespects First the investigation wasbased on data from a very large unse-lected population which was made pos-sible because of a cost-free health-caresystem for virtually all Italian citizens Sec-ond the drug prescription database pro-vides highly accurate data because reportof prescriptions by the pharmacies is es-sential for reimbursement andfiling of anincorrect report about dispensed drugshas legal consequences (30) Third pa-tients were identified from the point ofthe initial lipid-lowering therapy and thecomplete sequence of the subsequentprescriptions for statins was availableFourth we were able to include patients

Figure 1mdashEffect of adherence to statin therapy on the HRs for diabetes HR was estimated according to Cox proportional hazards model A and BCrude and adjusted estimates of all statins C and D Adjusted estimates of high-potency and low-potency statins Adjustments were made for age(continuous) sex first-line statin therapy concomitant use of other drugs history of CV disease and categories of the Charlson comorbidity indexscore (see Table 1 for the complete list of covariates included in the model)

carediabetesjournalsorg Corrao and Associates 2229

without previous clinical evidence (drugtreatment andor hospitalization) of dia-betes so that the data relate to the effectof statin use on new-onset of diabetesFinally a number of sensitivity analyseswere performed which increased the ro-bustness of the findingsThe study has a number of potential

limitations First evaluation of statin ad-herencewasbasedonpharmacy-dispensinginformation This method assumes thatthe PDC by a prescription corresponds tothe proportion of days of medicationuse Small insignificant differences be-tween the assessed number of dis-pensed pills and the actual pill countwere reported by a study investigatingadherence to statin therapy over 12months (31) Furthermore data on dis-pensing history have been shown to beconsistent with other adherence mea-sures drug serum levels and clinicaldrug effects (32) Nevertheless the useofmedication dispensing as ameasure of

adherence remains a source of uncer-tainty in our estimates

Second because of privacy regula-tions identification codes of prescrip-tion records were not available foranalysis so drug-based diagnoses ofdiabetes cannot be scrutinized and vali-dated However it seems highly unlikelythat diagnostic errors could differen-tially affect patients according to theiradherence to statins Moreover theadherence-diabetes relationship wasconfirmed and indeed amplified when amore stringent diagnostic criterion fornew-onset diabeteswasusedwhich raisesthe possibility that the statin-dependentrisk of developing diabetes might begreater than that resulting from the pri-mary diagnostic criterion we adopted Itshould also be mentioned that the im-proved diagnostic specificity obtainedwith the stricter criterion minimizes biasof a risk estimate that includes diagnos-tic misclassification

Third our estimates might be af-fected by detection bias that is patientswith long-term adherence to statin ther-apy may have been more likely toreceive a diagnosis of diabetes How-ever we found a clear relationship be-tween adherencewith statins and risk ofdiabetes even among patients who didnot undergo laboratory examinations oroutpatient specialist visits Thus the ex-cess risk of diabetes also concerned pa-tients with low use of health servicesmaking it unlikely that detection biasexplains the main findings

Finally whether the observed find-ings are a result of our inability to fullyaccount for higher adherence to statintherapy in patients at higher risk of di-abetes is the more relevant question ininterpreting the findings Baseline fast-ing serum glucose and other compo-nents of the metabolic syndrome suchas triglyceride level BMI and hyperten-sion (24) might be more frequent in

Figure 2mdashCombined action of adherence to statin therapy and number of cholesterolemia tests (top panel) and outpatient specialist visits (bottompanel) on the HR for diabetes HRwas estimated according to Cox proportional hazards model Estimates are adjusted for age (continuous) sex first-line statin therapy cotreatment with other drugs history of CV disease and categories of the Charlson comorbidity index score (see Table 1 for thecomplete list of covariates included in the model)

2230 Statins and the Risk of Diabetes Diabetes Care Volume 37 August 2014

individuals with high rather than low ad-herence to statin treatment Becauseour databases have a limited amountof clinical information we dealt withconfounding in several ways First onlystatin users were included in the cohortto compare the duration of statin ther-apy whereas inclusion of nonuserswould be for observational investiga-tions (819) In this way the potentialfor confounding is reduced by activelycomparing patients with the same indi-cation at baseline (33) Second our es-timates were adjusted for a number ofavailable demographic therapeutic andclinical characteristics such as cotreat-ment with antihypertensive and otheragents history of CV disease and cate-gories of Charlson comorbidity indexscore Third the statin-diabetes rela-tionship was observed within eachstratum of frequency of laboratory ex-aminations and outpatient specialist vis-its Because medical attention might beconsidered a proxy of clinical profile andother unmeasured risk factors for diabe-tes this further protects our conclu-sions although residual confoundingcannot be excluded For this reasonwe also performed a sensitivity analysisand showed that the association be-tween adherence to statin treatmentand the risk of diabetes was not an-nulled after correction for an unmea-sured confounder of great importancefor the development of diabetes obe-sity It should be emphasized that

beyond obesity no other unmeasuredfactor is able to annul the investigatedrelationship even if it is characterizedby very-high prevalence (ie up to375) strongly affects the outcome(ie up to threefold increased diabetesrisk) and is markedly more diffuseamong adherent patients (ie up totwofold more than in those with littleadherence to statin therapy) In suchconditions an increased risk of diabetesweaker than that found in our mainanalysis is expected This possibly ex-plains the stronger association generallyreported by observational investiga-tions than reported in clinical trials

Implications for Benefits of StatinTreatmentAssuming that 1) the diabetes incidencerate among patients with very-low ad-herence to statin therapy represents thebaseline rate of new-onset diabetes be-cause diabetes is unlikely to developduring very-short-term statin use and2) diabetes develops in patients withhigh adherence to statins at a rate132-fold faster than the baseline ratewe expected that 280 patients wouldhave to be continuously treated toinduce one case of diabetes everyyear In a meta-analysis of seven ran-domized controlled trials of statin useversus control in patients with diabetesstatins decreased major CV and cere-brovascular events every 40 patientstreated (34)

Consistent with our calculation ameta-analysis reported that treating255 patients with statins for 4 yearswould induce one case of diabetesbut in the meantime it would prevent54 coronary deaths or myocardial in-farctions for each millimolar reductionin serum LDL cholesterol (6) Thus theCV protection offered by statins ap-pears to markedly outweigh the in-creased incidence of diabetes althoughthe adverse effect of diabetes onCV risk (35) suggests that the originalstatin-dependent protection might de-cline with time The decline might be lessthan predicted based on epidemiolog-ical data on diabetes and CV morbidi-ty and mortality however becausewhether the prognostic value of drug-induced diabetes is equivalent to that ofnative diabetes is still uncertain (36)This has been reported to be the case insome studies (3738) whereas in otherstudies a few years or even long-termexposure to diabetes induced by antihy-pertensive drugs was not found to in-crease CV morbid or fatal events (2439)

ConclusionThis large population-based cohortstudy extends earlier findings of an in-creased risk of diabetes with statin ther-apy by providing evidence of a clear-cutassociation between adherence to sta-tin therapy and risk of new-onset diabe-tes in a real-world setting It appearsfrom event-based investigations that

Figure 3mdashChanges in HRs (and 95 CIs) for the risk of diabetes (outcome) associated with high adherence to statin therapy (exposure) after externaladjustment for obesity (confounder) External adjustment was performed by means of Monte Carlo sensitivity analysis Adjusted estimates wereobtained by means of setting 1) the prevalence of obesity among statin users to 375 (12) 2) the risk of diabetes in obese patients to threefoldhigher than in normal weight patients (25) and 3) the odds of obese patients in the highest PDC category from twofold lower to twofold higher thanthat of normal weight patients (2627)

carediabetesjournalsorg Corrao and Associates 2231

individuals with high rather than low ad-herence to statin treatment Becauseour databases have a limited amountof clinical information we dealt withconfounding in several ways First onlystatin users were included in the cohortto compare the duration of statin ther-apy whereas inclusion of nonuserswould be for observational investiga-tions (819) In this way the potentialfor confounding is reduced by activelycomparing patients with the same indi-cation at baseline (33) Second our es-timates were adjusted for a number ofavailable demographic therapeutic andclinical characteristics such as cotreat-ment with antihypertensive and otheragents history of CV disease and cate-gories of Charlson comorbidity indexscore Third the statin-diabetes rela-tionship was observed within eachstratum of frequency of laboratory ex-aminations and outpatient specialist vis-its Because medical attention might beconsidered a proxy of clinical profile andother unmeasured risk factors for diabe-tes this further protects our conclu-sions although residual confoundingcannot be excluded For this reasonwe also performed a sensitivity analysisand showed that the association be-tween adherence to statin treatmentand the risk of diabetes was not an-nulled after correction for an unmea-sured confounder of great importancefor the development of diabetes obe-sity It should be emphasized that

beyond obesity no other unmeasuredfactor is able to annul the investigatedrelationship even if it is characterizedby very-high prevalence (ie up to375) strongly affects the outcome(ie up to threefold increased diabetesrisk) and is markedly more diffuseamong adherent patients (ie up totwofold more than in those with littleadherence to statin therapy) In suchconditions an increased risk of diabetesweaker than that found in our mainanalysis is expected This possibly ex-plains the stronger association generallyreported by observational investiga-tions than reported in clinical trials

Implications for Benefits of StatinTreatmentAssuming that 1) the diabetes incidencerate among patients with very-low ad-herence to statin therapy represents thebaseline rate of new-onset diabetes be-cause diabetes is unlikely to developduring very-short-term statin use and2) diabetes develops in patients withhigh adherence to statins at a rate132-fold faster than the baseline ratewe expected that 280 patients wouldhave to be continuously treated toinduce one case of diabetes everyyear In a meta-analysis of seven ran-domized controlled trials of statin useversus control in patients with diabetesstatins decreased major CV and cere-brovascular events every 40 patientstreated (34)

Consistent with our calculation ameta-analysis reported that treating255 patients with statins for 4 yearswould induce one case of diabetesbut in the meantime it would prevent54 coronary deaths or myocardial in-farctions for each millimolar reductionin serum LDL cholesterol (6) Thus theCV protection offered by statins ap-pears to markedly outweigh the in-creased incidence of diabetes althoughthe adverse effect of diabetes onCV risk (35) suggests that the originalstatin-dependent protection might de-cline with time The decline might be lessthan predicted based on epidemiolog-ical data on diabetes and CV morbidi-ty and mortality however becausewhether the prognostic value of drug-induced diabetes is equivalent to that ofnative diabetes is still uncertain (36)This has been reported to be the case insome studies (3738) whereas in otherstudies a few years or even long-termexposure to diabetes induced by antihy-pertensive drugs was not found to in-crease CV morbid or fatal events (2439)

ConclusionThis large population-based cohortstudy extends earlier findings of an in-creased risk of diabetes with statin ther-apy by providing evidence of a clear-cutassociation between adherence to sta-tin therapy and risk of new-onset diabe-tes in a real-world setting It appearsfrom event-based investigations that

Figure 3mdashChanges in HRs (and 95 CIs) for the risk of diabetes (outcome) associated with high adherence to statin therapy (exposure) after externaladjustment for obesity (confounder) External adjustment was performed by means of Monte Carlo sensitivity analysis Adjusted estimates wereobtained by means of setting 1) the prevalence of obesity among statin users to 375 (12) 2) the risk of diabetes in obese patients to threefoldhigher than in normal weight patients (25) and 3) the odds of obese patients in the highest PDC category from twofold lower to twofold higher thanthat of normal weight patients (2627)

carediabetesjournalsorg Corrao and Associates 2231

benefits of statins in reducing CV eventsclearly overwhelm the diabetes risk

Duality of Interest GM discloses consultancyagreements with Boehringer Ingelheim andNovartis and participation in speakers bureausfor Bayer Boehringer Ingelheim Merck Sharp ampDohmeManar International Novartis RecordatiSanofi Sankyo and Servier No other potentialconflicts of interest relevant to this article werereportedAuthorContributionsGC contributed to theinitial study idea interpretation of the resultsand drafting of the manuscript BI contributedto the protocol preparation of the data set foranalysis data analysis and interpretation of theresults FN contributed to the sensitivityanalysis and interpretation of the results DSET and MC contributed to the protocol andinterpretation of the results LM contributedto abstracting the data and authorizing their useand the interpretation of the results ALCGG and GM contributed to the interpreta-tion of the pharmacological and clinical pro-spective results and review of the manuscriptGC is the guarantor of this work and as suchhad full access to all the data in the study andtakes responsibility for the integrity of the dataand the accuracy of the data analysis

References1 Catapano AL Reiner Z De Backer G et alEuropean Society of Cardiology (ESC) EuropeanAtherosclerosis Society (EAS) ESCEAS guide-lines for the management of dyslipidaemiasThe Task Force for the Management ofDyslipidaemias of the European Society of Car-diology (ESC) and the European AtherosclerosisSociety (EAS) Atherosclerosis 20112173ndash462 Baigent C Keech A Kearney PM et al Cho-lesterol Treatment Trialistsrsquo (CTT) CollaboratorsEfficacy and safety of cholesterol-lowering treat-ment prospective meta-analysis of data from90056 participants in 14 randomised trials ofstatins Lancet 20053661267ndash12783 Kearney PM Blackwell L Collins R et alCholesterol Treatment Trialistsrsquo (CTT) Collabo-rators Efficacy of cholesterol-lowering therapyin 18686 people with diabetes in 14 randomisedtrials of statins ameta-analysis Lancet 2008371117ndash1254 Colbert JD Stone JA Statin use and the riskof incident diabetes mellitus a review of theliterature Can J Cardiol 201228581ndash5895 Ridker PM Danielson E Fonseca FAH et alJUPITER Study Group Rosuvastatin to preventvascular events in men and women with ele-vated C-reactive protein N Engl J Med 20083592195ndash22076 Sattar N Preiss D Murray HM et al Statinsand risk of incident diabetes a collaborativemeta-analysis of randomised statin trials Lan-cet 2010375735ndash7427 Mills EJ Wu P Chong G et al Efficacy andsafety of statin treatment for cardiovascular dis-ease a network meta-analysis of 170255 pa-tients from 76 randomized trials QJM 2011104109ndash124

8 Zaharan NL Williams D Bennett K Statinsand risk of treated incident diabetes in a pri-mary care population Br J Clin Pharmacol2013751118ndash11249 Corrao G Cesana G Merlino L Pharmaco-epidemiological research and the linking of elec-tronic healthcare databases available in theItalian region of Lombardy Biomed Stat ClinEpidemiol 20082111ndash12510 Helin-Salmivaara A Lavikainen P KorhonenMJ et al Long-term persistence with statintherapy a nationwide register study in FinlandClin Ther 2008302228ndash224011 Andrade SE Kahler KH Frech F Chan KAMethods for evaluation of medication adher-ence and persistence using automated data-bases Pharmacoepidemiol Drug Saf 200615565ndash574 discussion 575ndash57712 LawMRWald NJ Rudnicka AR Quantifyingeffect of statins on low density lipoprotein cho-lesterol ischaemic heart disease and strokesystematic review and meta-analysis BMJ2003326142313 Charlson ME Charlson RE Peterson JCMarinopoulos SS Briggs WM Hollenberg JPThe Charlson comorbidity index is adapted topredict costs of chronic disease in primarycare patients J Clin Epidemiol 2008611234ndash124014 Delgado-Rodrıguez M Llorca J Bias J Epi-demiol Community Health 200458635ndash64115 Lindstrom J Tuomilehto J The diabetes riskscore a practical tool to predict type 2 diabetesrisk Diabetes Care 200326725ndash73116 Huther J von Wolff A Stange D et al In-complete medication adherence of chronicallyill patients in German primary care Patient Pre-fer Adherence 20137237ndash24417 Mazzaglia G Ambrosioni E Alacqua Met al Adherence to antihypertensive medica-tions and cardiovascular morbidity amongnewly diagnosed hypertensive patients Circula-tion 20091201598ndash160518 Greenland S Sensitivity analysis MonteCarlo risk analysis and Bayesian uncertainty as-sessment Risk Anal 200121579ndash58319 Culver AL Ockene IS Balasubramanian Ret al Statin use and risk of diabetes mellitus inpostmenopausal women in the WomenrsquosHealth Initiative Arch Intern Med 2012172144ndash15220 Corrao G Nicotra F Parodi A et al Externaladjustment for unmeasured confounders im-proved drug-outcome association estimatesbased on health care utilization data J Clin Epi-demiol 2012651190ndash119921 Shepherd J Blauw GJ Murphy MB et alPROSPER Study Group PROspective Study ofPravastatin in the Elderly at Risk Pravastatinin elderly individuals at risk of vascular disease(PROSPER) a randomised controlled trial Lan-cet 20023601623ndash163022 Carter AA Gomes TG Camacho X JuurlinkDN Shah BR Mamdani MM Risk of incidentdiabetes among patients treated with statinspopulation based study BMJ 2013346f261023 Rajpathak SN Kumbhani DJ Crandall JBarzilai N Alderman M Ridker PM Statin ther-apy and risk of developing type 2 diabetesa meta-analysis Diabetes Care 2009321924ndash1929

24 Waters DD Ho JE DeMicco DA et al Pre-dictors of new-onset diabetes in patients treatedwith atorvastatin results from 3 large random-ized clinical trials J Am Coll Cardiol 2011571535ndash154525 Preiss D Seshasai SR Welsh P et al Risk ofincident diabetes with intensive-dose com-pared with moderate-dose statin therapya meta-analysis JAMA 20113052556ndash256426 Takaguri A Satoh K Itagaki M Tokumitsu YIchihara K Effects of atorvastatin and pravastat-in on signal transduction related to glucose up-take in 3T3L1 adipocytes J Pharmacol Sci 200810780ndash8927 Koh KK Quon MJ Han SH et al Differentialmetabolic effects of pravastatin and simvastatinin hypercholesterolemic patients Atherosclero-sis 2009204483ndash49028 Ishikawa M Okajima F Inoue N et al Dis-tinct effects of pravastatin atorvastatin andsimvastatin on insulin secretion from a beta-cell line MIN6 cells J Atheroscler Thromb200613329ndash33529 MaT Tien L FangC-L LiouYS JongGP Statinsand new-onset diabetes a retrospective longitudi-nal cohort study Clin Ther 2012341977ndash198330 Strom BL Overview of automated data-bases in pharmacoepidemiology In Pharmaco-epidemiology 3rd ed Strom BL Ed ChichesterUK Wiley 2000 p 219ndash22231 Lee JK Grace KA Foster TG et al Howshould we measure medication adherence inclinical trials and practice Ther Clin Risk Manag20073685ndash69032 Steiner JF Prochazka AV The assessment ofrefill compliance using pharmacy records meth-ods validity and applications J Clin Epidemiol199750105ndash11633 Schneeweiss S Avorn J A review of uses ofhealth care utilization databases for epidemio-logic research on therapeutics J Clin Epidemiol200558323ndash33734 Chen YH Feng B Chen ZW Statins for pri-mary prevention of cardiovascular and cerebro-vascular events in diabetic patients withoutestablished cardiovascular diseases a meta-analysis Exp Clin Endocrinol Diabetes 2012120116ndash12035 LaRosa JC He J Vupputuri S Effect of stat-ins on risk of coronary disease a meta-analysisof randomized controlled trials JAMA 19992822340ndash234636 Mancia G Grassi G Zanchetti A New-onsetdiabetes and antihypertensive drugs J Hypertens2006243ndash1037 Aksnes TA Kjeldsen SE Rostrup M Omvik PHua TA Julius S Impact of new-onset diabetesmellitus on cardiac outcomes in the Valsartan Anti-hypertensive Long-term Use Evaluation (VALUE)trial population Hypertension 200750467ndash47338 Verdecchia P Reboldi G Angeli F et al Ad-verse prognostic significance of new diabetes intreated hypertensive subjects Hypertension200443963ndash96939 Kostis JB Wilson AC Freudenberger RSCosgrove NM Pressel SL Davis BR SHEP Collab-orative Research Group Long-term effect ofdiuretic-based therapy on fatal outcomes insubjects with isolated systolic hypertensionwith and without diabetes Am J Cardiol 20059529ndash35

2232 Statins and the Risk of Diabetes Diabetes Care Volume 37 August 2014

Copyright of Diabetes Care is the property of American Diabetes Association and its contentmay not be copied or emailed to multiple sites or posted to a listserv without the copyrightholders express written permission However users may print download or email articles forindividual use

Copyright of Diabetes Care is the property of American Diabetes Association and its contentmay not be copied or emailed to multiple sites or posted to a listserv without the copyrightholders express written permission However users may print download or email articles forindividual use