The Unseen Hand: AI-Based Prescribing Decision Support Tools and the Evaluation of Drug Safety and Effectiveness

The use of artificial intelligence (AI)-based tools to guide prescribing decisions is full of promise and may enhance patient outcomes. These tools can perform actions such as choosing the 'safest' medication, choosing between competing medications, promoting de-prescribing or even predicting non-adherence. These tools can exist in a variety of formats; for example, they may be directly integrated into electronic medical records or they may exist in a stand-alone website accessible by a web browser. One potential impact of these tools is that they could manipulate our understanding of the benefit-risk of medicines in the real world. Currently, the benefit risk of approved medications is assessed according to carefully planned agreements covering spontaneous reporting systems and planned surveillance studies. But AI-based tools may limit or even block prescription to high-risk patients or prevent off-label use. The uptake and temporal availability of these tools may be uneven across healthcare systems and geographies, creating artefacts in data that are difficult to account for. It is also hard to estimate the 'true impact' that a tool had on a prescribing decision. International borders may also be highly porous to these tools, especially in cases where tools are available over the web. These tools already exist, and their use is likely to increase in the coming years. How they can be accounted for in benefit-risk decisions is yet to be seen.

Application of an Innovative Data Mining Approach Towards Safe Polypharmacy Practice in Older Adults

INTRODUCTION

Polypharmacy is common and is associated with higher risk of adverse drug event (ADE) among older adults. Knowledge on the ADE risk level of exposure to different drug combinations is critical for safe polypharmacy practice, while approaches for this type of knowledge discovery are limited. The objective of this study was to apply an innovative data mining approach to discover high-risk and alternative low-risk high-order drug combinations (e.g., three- and four-drug combinations).

METHODS

A cohort of older adults (≥ 65 years) who visited an emergency department (ED) were identified from Medicare fee-for-service and MarketScan Medicare supplemental data. We used International Classification of Diseases (ICD) codes to identify ADE cases potentially induced by anticoagulants, antidiabetic drugs, and opioids from ED visit records. We assessed drug exposure data during a 30-day window prior to the ED visit dates. We investigated relationships between exposure of drug combinations and ADEs under the case-control setting. We applied the mixture drug-count response model to identify high-order drug combinations associated with an increased risk of ADE. We conducted therapeutic class-based mining to reveal low-risk alternative drug combinations for high-order drug combinations associated with an increased risk of ADE.

RESULTS

We investigated frequent high-order drug combinations from 8.4 million ED visit records (5.1 million from Medicare data and 3.3 million from MarketScan data). We identified 5213 high-order drug combinations associated with an increased risk of ADE by controlling the false discovery rate at 0.01. We identified 1904 high-order, high-risk drug combinations had potential low-risk alternative drug combinations, where each high-order, high-risk drug combination and its corresponding low-risk alternative drug combination(s) have similar therapeutic classes.

CONCLUSIONS

We demonstrated the application of a data mining technique to discover high-order drug combinations associated with an increased risk of ADE. We identified high-risk, high-order drug combinations often have low-risk alternative drug combinations in similar therapeutic classes.

Changing patterns in medication prescription for gestational diabetes during a time of guideline change in the USA: a cross-sectional study

OBJECTIVE

To define patterns of prescription and factors associated with choice of pharmacotherapy for gestational diabetes mellitus (GDM), namely metformin, glyburide and insulin, during a period of evolving professional guidelines.

DESING

Cross-sectional study.

SETTING

US commercial insurance beneficiaries from Market-Scan (late 2015 to 2018).

STUDY DESIGN

We included women with GDM, singleton gestations, 15-51 years of age on pharmacotherapy. The exposure was pharmacy claims for metformin, glyburide and insulin.

MAIN OUTCOMES

Pharmacotherapy for GDM with either oral agent, metformin or glyburide, compared with insulin as the reference, and secondarily, consequent treatment modification (addition and/or change) to metformin, glyburide or insulin.

RESULTS

Among 37 762 women with GDM, we analysed data from 10 407 (28%) with pharmacotherapy, 21% with metformin (n = 2147), 48% with glyburide (n = 4984) and 31% with insulin (n = 3276). From late 2015 to 2018, metformin use increased from 17 to 29%, as did insulin use from 26 to 44%, whereas glyburide use decreased from 58 to 27%. By 2018, insulin was the most common pharmacotherapy for GDM; metformin was more likely to be prescribed by 9% compared with late 2015/16, but glyburide was less likely by 45%. Treatment modification occurred in 20% of women prescribed metformin compared with 2% with insulin and 8% with glyburide.

CONCLUSIONS

Insulin followed by metformin has replaced glyburide as the most common pharmacotherapy for GDM among a privately insured US population during a time of evolving professional guidelines. Further evaluation of the relative effectiveness and safety of metformin compared with insulin is needed.

TWEETABLE ABSTRACT

Insulin followed by metformin has replaced glyburide as the most common pharmacotherapy for gestational diabetes mellitus in the USA.

A multistate transition model for statin-induced myopathy and statin discontinuation

The overarching goal of this study was to simultaneously model the dynamic relationships among statin exposure, statin discontinuation, and potentially statin-related myopathic outcomes. We extracted data from the Indiana Network of Patient Care for 134,815 patients who received statin therapy between January 4, 2004, and December 31, 2008. All individuals began statin treatment, some discontinued statin use, and some experienced myopathy and/or rhabdomyolysis while taking the drug or after discontinuation. We developed a militate model to characterize 12 transition probabilities among six different states defined by use or discontinuation of statin and its associated myopathy or rhabdomyolysis. We found that discontinuation of statin therapy was common and frequently early, with 44.4% of patients discontinuing therapy after 1 month, and discontinuation is a strong indicator for statin-induced myopathy (risk ratio, 10.8; p < 0.05). Women more likely than men (p < 0.05) and patients aged 65 years and older had a higher risk than those aged younger than 65 years to discontinue statin use or experience myopathy. In conclusion, we introduce an innovative multistate model that allows clear depiction of the relationship between statin discontinuation and statin-induced myopathy. For the first time, we have successfully demonstrated and quantified the relative risk of myopathy between patients who continued and discontinued statin therapy. Age and sex were two strong risk factors for both statin discontinuation and incident myopathy.

Random control selection for conducting high-throughput adverse drug events screening using large-scale longitudinal health data

Case-control design based high-throughput pharmacoinformatics study using large-scale longitudinal health data is able to detect new adverse drug event (ADEs) signals. Existing control selection approaches for case-control design included the dynamic/super control selection approach. The dynamic/super control selection approach requires all individuals to be evaluated at all ADE case index dates, as the individuals' eligibilities as control depend on ADE/enrollment history. Thus, using large-scale longitudinal health data, the dynamic/super control selection approach requires extraordinarily high computational time. We proposed a random control selection approach in which ADE case index dates were matched by randomly generated control index dates. The random control selection approach does not depend on ADE/enrollment history. It is able to significantly reduce computational time to prepare case-control data sets, as it requires all individuals to be evaluated only once. We compared the performance metrics of all control selection approaches using two large-scale longitudinal health data and a drug-ADE gold standard including 399 drug-ADE pairs. The F-scores for the random control selection approach were between 0.586 and 0.600 compared to between 0.545 and 0.562 for dynamic/super control selection approaches. The random control selection approach was ~ 1000 times faster than dynamic/super control selection approach on preparing case-control data sets. With large-scale longitudinal health data, a case-control design-based pharmacoinformatics study using random control selection is able to generate comparable ADE signals than the existing control selection approaches. The random control selection approach also significantly reduces computational time to prepare the case-control data sets.

Dietary Patterns of Insulinemia, Inflammation and Glycemia, and Pancreatic Cancer Risk: Findings from the Women's Health Initiative

BACKGROUND

Pancreatic cancer risk is increasing in countries with high consumption of Western dietary patterns and rising obesity rates. We examined the hypothesis that specific dietary patterns reflecting hyperinsulinemia (empirical dietary index for hyperinsulinemia; EDIH), systemic inflammation (empirical dietary inflammatory pattern; EDIP), and postprandial glycemia [glycemic index (GI); glycemic load (GL)] are associated with pancreatic cancer risk, including the potential modifying role of type 2 diabetes (T2D) and body mass index (BMI).

METHODS

We calculated dietary scores from baseline (1993-1998) food frequency questionnaires among 129,241 women, 50-79 years-old in the Women's Health Initiative. We used multivariable-adjusted Cox regression to estimate HRs and 95% confidence intervals (95% CI) for pancreatic cancer risk.

RESULTS

During a median 19.9 years of follow-up, 850 pancreatic cancer cases were diagnosed. We observed no association between dietary scores and pancreatic cancer risk overall. However, risk was elevated among participants with longstanding T2D (present >3 years before pancreatic cancer diagnosis) for EDIH. For each 1 SD increment in dietary score, the HRs (95% CIs) were: EDIH, 1.33 (1.06-1.66); EDIP, 1.26 (0.98-1.63); GI, 1.26 (0.96-1.67); and GL, 1.23 (0.96-1.57); although interactions were not significant (all P _interaction >0.05). Separately, we observed inverse associations between GI [0.86 (0.76-0.96), P _interaction = 0.0068] and GL [0.83 (0.73-0.93), P _interaction = 0.0075], with pancreatic cancer risk among normal-weight women.

CONCLUSIONS

We observed no overall association between the dietary patterns evaluated and pancreatic cancer risk, although women with T2D appeared to have greater cancer risk.

IMPACT

The elevated risk for hyperinsulinemic diets among women with longstanding T2D and the inverse association among normal-weight women warrant further examination.

Insulinemic and Inflammatory Dietary Patterns Show Enhanced Predictive Potential for Type 2 Diabetes Risk in Postmenopausal Women

OBJECTIVE

The empirical dietary index for hyperinsulinemia (EDIH) and empirical dietary inflammatory pattern (EDIP) scores assess the insulinemic and inflammatory potentials of habitual dietary patterns, irrespective of the macronutrient content, and are based on plasma insulin response or inflammatory biomarkers, respectively. The glycemic index (GI) and glycemic load (GL) assess postprandial glycemic potential based on dietary carbohydrate content. We tested the hypothesis that dietary patterns promoting hyperinsulinemia, chronic inflammation, or hyperglycemia may influence type 2 diabetes risk.

RESEARCH DESIGN AND METHODS

We calculated dietary scores from baseline (1993-1998) food frequency questionnaires among 73,495 postmenopausal women in the Women's Health Initiative, followed through March 2019. We used multivariable-adjusted Cox regression to estimate hazard ratios (HRs) and 95% CIs for type 2 diabetes risk. We also estimated multivariable-adjusted absolute risk of type 2 diabetes.

RESULTS

During a median 13.3 years of follow-up, 11,009 incident cases of type 2 diabetes were diagnosed. Participants consuming the most hyperinsulinemic or proinflammatory dietary patterns experienced greater risk of type 2 diabetes; HRs (95% CI) comparing highest to lowest dietary index quintiles were EDIH 1.49 (1.32-1.68; P _trend < 0.0001) and EDIP 1.45 (1.29-1.63; P _trend < 0.0001). The absolute excess incidence for the same comparison was 220 (EDIH) and 271 (EDIP) cases per 100,000 person-years. GI and GL were not associated with type 2 diabetes risk: GI 0.99 (0.88-1.12; P _trend = 0.46) and GL 1.01 (0.89-1.16; P _trend = 0.30).

CONCLUSIONS

Our findings in this diverse cohort of postmenopausal women suggest that lowering the insulinemic and inflammatory potentials of the diet may be more effective in preventing type 2 diabetes than focusing on glycemic foods.

The Insulinemic, Inflammatory, and Glycemic Potential of the Diet in Relation to Risk of Type 2 Diabetes

Objectives

Dietary patterns that promote chronic systemic inflammation, hyperinsulinemia, or hyperglycemia may influence type 2 diabetes (T2D) risk. We evaluated an empirical dietary index for hyperinsulinemia (EDIH), empirical dietary inflammatory pattern (EDIP), glycemic index (GI), and glycemic load (GL), and risk of T2D among US postmenopausal women. EDIH and EDIP assess the insulinemic or inflammatory potential of habitual diets, irrespective of macronutrient content, and are based on plasma concentrations of insulin response or inflammatory biomarkers, respectively. The GI and GL assess postprandial glycemic potential based on carbohydrate content of the diet.

Methods

We calculated dietary scores from baseline food frequency questionnaires among 73,495 participants aged 50–79 years in the Women's Health Initiative Observational Study and Clinical Trials. We used multivariable-adjusted Cox regression to estimate hazard ratios (HR) and 95% confidence intervals (95% CI) for risk of T2D according to quintiles of dietary scores.

Results

There were 11,009 incident cases of T2D during a median 13.3 years of follow-up. In multivariable-adjusted analyses, participants in the highest dietary score quintiles (consuming the most hyperinsulinemic, proinflammatory, or hyperglycemic diets) were at highest risk of T2D compared to those in the lowest quintiles: EDIH: HR, 1.54(1.37, 1.74); Ptrend &lt; .0001; EDIP: HR, 1.45 (1.29, 1.64); Ptrend &lt; .0001). GI and GL were not associated with T2D risk: GI: HR, 0.99 (0.88, 1.12); Ptrend = 0.94; GL: HR, 0.98 (0.85, 1.12); Ptrend = 0.32. In subgroup analyses, associations of EDIH and EDIP with T2D risk were stronger among overweight or obese than normal-weight women (Pinteraction: EDIH = 0.02, EDIP = 0.003), and findings did not significantly vary by race/ethnicity.

Conclusions

In this large sample of postmenopausal women, hyperinsulinemic, and pro-inflammatory dietary patterns were associated with higher risk of T2D, more so among overweight and obese women, whereas dietary glycemic potential was not associated with T2D risk.

Funding Sources

NCI grant # R00CA207736 and the WHI program is funded by NHLBI grant #s HHSN268201600018C, HHSN268201600001C, HHSN268201600002C, HHSN268201600003C, and HHSN268201600004C.

Challenges facing drug utilization research in the Latin American region

PURPOSE

The International Society of Pharmacoepidemiology (ISPE) in collaboration with the Latin America Drug Utilization Research Group (LatAm DURG), the Medicines Utilization Research in Africa (MURIA) group, and the Uppsala Monitoring Center, is leading an initiative to understand challenges to drug utilization research (DUR) in the Latin American (LatAm) and African regions with the goal of communicating results and proposing solutions to these challenges in four scientific publications. The purpose of this first manuscript is to identify the main challenges associated with DUR in the LatAm region.

METHODS

Drug utilization (DU) researchers in the LatAm region voluntarily participated in multiple discussions, contributed with local data and reviewed successive drafts and the final manuscript. Additionally, we carried out a literature review to identify the most relevant publications related to DU studies from the LatAm region.

RESULTS

Multiple challenges were identified in the LatAm region for DUR including socioeconomic inequality, access to medical care, complexity of the healthcare system, limited investment in research and development, limited institutional and organization resources, language barriers, limited health education and literacy. Further, there is limited use of local DUR data by decision makers particularly in the identification of emerging health needs coming from social and demographic transitions.

CONCLUSIONS

The LatAm region faces challenges to DUR which are inherent in the healthcare and political systems, and potential solutions should target changes to the system.

Association of Race and Ethnicity With Late-Life Depression Severity, Symptom Burden, and Care

Importance

Knowledge gaps persist regarding racial and ethnic variation in late-life depression, including differences in specific depressive symptoms and disparities in care.

Objective

To examine racial/ethnic differences in depression severity, symptom burden, and care.

Design, Setting, and Participants

This cross-sectional study included 25 503 of 25 871 community-dwelling older adults who participated in the Vitamin D and Omega-3 Trial (VITAL), a randomized trial of cancer and cardiovascular disease prevention conducted from November 2011 to December 2017. Data analysis was conducted from June to September 2018.

Exposure

Racial/ethnic group (ie, non-Hispanic white; black; Hispanic; Asian; and other, multiple, or unspecified race).

Main Outcomes and Measures

Depressive symptoms, assessed using the Patient Health Questionnaire-8 (PHQ-8); participant-reported diagnosis, medication, and/or counseling for depression. Differences across racial/ethnic groups were evaluated using multivariable zero-inflated negative binomial regression to compare PHQ-8 scores and multivariable logistic regression to estimate odds of item-level symptom burden and odds of depression treatment among those with diagnosed depression.

Results

There were 25 503 VITAL participants with adequate depression data (mean [SD] age, 67.1 [7.1] years) including 12 888 [50.5%] women, 17 828 [69.9%] non-Hispanic white participants, 5004 [19.6%] black participants, 1001 [3.9%] Hispanic participants, 377 [1.5%] Asian participants, and 1293 participants [5.1%] who were categorized in the other, multiple, or unspecified race group. After adjustment for sociodemographic, lifestyle, and health confounders, black participants had a 10% higher severity level of PHQ-8 scores compared with non-Hispanic white participants (rate ratio [RR], 1.10; 95% CI, 1.04-1.17; P < .001); Hispanic participants had a 23% higher severity level of PHQ-8 scores compared with non-Hispanic white participants (RR, 1.23; 95% CI, 1.10-1.38; P < .001); and participants in the other, multiple, or unspecified group had a 14% higher severity level of PHQ-8 scores compared with non-Hispanic white participants (RR, 1.14; 95% CI, 1.04-1.25; P = .007). Compared with non-Hispanic white participants, participants belonging to minority groups had 1.5-fold to 2-fold significantly higher fully adjusted odds of anhedonia (among black participants: odds ratio [OR], 1.76; 95% CI, 1.47-2.11; among Hispanic participants: OR, 1.96; 95% CI, 1.43-2.69), sadness (among black participants: OR, 1.31; 95% CI, 1.07-1.60; among Hispanic participants: OR, 2.09; 95% CI, 1.51-2.88), and psychomotor symptoms (among black participants: OR, 1.77; 95% CI, 1.31-2.39; among Hispanic participants: OR, 2.12; 95% CI, 1.28-3.50); multivariable-adjusted odds of sleep problems and guilt appeared higher among Hispanic vs non-Hispanic white participants (sleep: OR, 1.24; 95% CI, 1.01-1.52; guilt: 1.84; 95% CI, 1.31-2.59). Among those with clinically significant depressive symptoms (ie, PHQ-8 score ≥10) and/or those with diagnosed depression, black participants were 61% less likely to report any treatment (ie, medications and/or counseling) than non-Hispanic white participants after adjusting for confounders (adjusted OR, 0.39; 95% CI, 0.27-0.56).

Conclusions and Relevance

In this cross-sectional study, significant racial and ethnic differences in late-life depression severity, item-level symptom burden, and depression care were observed after adjustment for numerous confounders. These findings suggest a need for further examination of novel patient-level and clinician-level factors underlying these associations.

Protocol for studying racial/ethnic disparities in depression care using joint information from participant surveys and administrative claims databases: an observational cohort study

INTRODUCTION

Current evidence indicates that older racial/ethnic minorities encounter disparities in depression care. Because late-life depression is common and confers major adverse health consequences, it is imperative to reduce disparities in depression care. Thus, the primary objectives of this protocol are to: (1) quantify racial/ethnic disparities in depression treatment and (2) identify and quantify the magnitude of these disparities accountable for by a multifactorial combination of patient, provider and healthcare system factors.

METHODS AND ANALYSIS

Data will be derived from the Vitamin D and Omega-3 Trial-Depression Endpoint Prevention (VITAL-DEP) study, a late-life depression prevention ancillary study to the VITAL trial. A total of 25 871 men and women, aged 50+ and 55+ years, respectively, were randomised in a 2×2 factorial randomised trial of heart disease and cancer prevention to receive vitamin D and/or fish oil for 5 years starting from 2011. Most participants were aged 65+ years old at randomisation. Medicare claims data for over 19 000 VITAL/VITAL-DEP participants were linked to conduct our study.The major study outcomes are depression treatment (antidepressant use and/or receipt of psychotherapy services) and adherence to medication treatment (antidepressant adherence and acceptability). The National Academy of Medicine framework for studying racial disparities was leveraged to select patient-level, provider-level and healthcare system-level variables and to address their potential roles in depression care disparities. Blinder-Oaxaca regression decomposition methods will be implemented to quantify and identify correlates of racial/ethnic disparities in depression treatment and adherence.

ETHICS AND DISSEMINATION

This study received Institutional Review Board (IRB) approval from the Partners Healthcare (PHS) IRB, protocol# 2010P001881. We plan to disseminate our results through publication of manuscripts patient engagement activities, such as study newsletters regularly sent out to VITAL participants, and presentations at scientific meetings.

TRIAL REGISTRATION NUMBER

NCT01696435.

A review of the Ghana National Health Insurance Scheme claims database: possibilities and limits for drug utilization research

BACKGROUND

There are inadequate data on prescribed drug utilization in Sub-Saharan Africa (SSA). Drug utilization research (DUR) in this region is hampered by lack of access to databases that capture prescribed drug utilization such as health insurance claims, electronic medical records and disease registries. The primary objective of this MiniReview was to describe the content of the NHIS claims database in the context of the health care system in Ghana. We will also review the possibilities and limitations of analysing this novel database for drug utilization research (DUR) in Ghana.

METHODS

We reviewed the history, composition of the database, coverage and health systems in Ghana. To demonstrate the application of the NHIS claims database for DUR, we reviewed the NHIS' drug formulary (NHIS medicines' list), assessed and quantified the utilization of the top 25 most commonly prescribed medicines and their distributions by age, sex, region of residence and by MDCs.

RESULTS

As of December 2014, about 40% (~10.5 million) of the Ghanaian population were active beneficiaries of NHIS. There were 1.43 million unique patients in the NHIS claims database who received services from about 81 providers located in 9 out of the 10 regions in Ghana. The mean age of this sample of beneficiaries was 31 (standard deviation, 22) years, a third of whom were aged <18 years old. Nearly, 2 out of every 3 beneficiaries were females. On average, there were approximately 3 outpatient visits per beneficiary in 2015. There were about 522 unique drugs on the NHIS medicine list. Overall, analgesic was the most prescribed class of medicine (mostly paracetamol and diclofenac). Antimalarials, artemether-lumefantrine, were observed as the second most prescribed medicines followed by anti-infectives (metronidazole) and antihypertensives (amlodipine).

CONCLUSION

The Ghana NHIS claims database is a great resource for DUR. This database could also be extended to facilitate pharmacoepidemiological and other health services' research especially if transformed into one of the existing standardized common data models.

Changes in prescribing and healthcare resource utilization after FDA Drug Safety Communications involving zolpidem-containing medications

PURPOSE

Products containing the sedative/hypnotic zolpidem were subject to Drug Safety Communications (DSCs) in January and May 2013 describing the risk of next-morning impairment and recommending lower starting doses particularly for women. This study aimed to assess whether zolpidem DSCs were associated with prescribing-pattern changes between January 2011 and December 2013.

METHODS

We assessed overall dispensings of zolpidem-containing products between January 2011 and December 2013 by conducting a time-series analysis. Analyses were stratified by gender because the DSC contained gender-specific information. Participants were patients drawn from the Optum Clinformatics data source of commercially insured people in the USA. We evaluated changes in mean prescribed dose of the two drugs and health care utilization metrics.

RESULTS

Each month of the study, more than 80 000 patients received a zolpidem-containing product and approximately one-tenth as many received eszopiclone. The two DSCs did not affect the downward trajectory of new zolpidem prescriptions. However, there was an increase in use of lower-dose forms of zolpidem (30% increase, p < 0.001), coupled with a reduction in higher-dose forms (13% decrease, p = 0.03), so that the average dose decreased after the DSCs (from 9.7 mg to 9.4 mg, p < 0.001), a change that was not seen with eszopiclone (from 2.74 mg to 2.74 mg, p = 0.45).

CONCLUSION

The DSCs related to zolpidem-containing products shifted prescribing toward the lower-dose formulations, consistent with the recommendations in the DSCs. Copyright © 2017 John Wiley & Sons, Ltd.

Effect of Generic Competition on Atorvastatin Prescribing and Patients' Out-of-Pocket Spending

IMPORTANCE

In November 2011, the cholesterol level-lowering medication atorvastatin calcium became available in the United States as a generic drug. However, only a single generic form (from a manufacturer that qualified for market exclusivity by challenging several of Pfizer's patents) and an authorized generic form (a brand-name drug sold as a generic) were available for the first 180 days.

OBJECTIVE

To describe trends in the prescribing of generic atorvastatin after expiration of market exclusivity for the brand-name medication and the effect on patients' out-of-pocket spending.

DESIGN, SETTING, AND PARTICIPANTS

A US population-based study used commercial claims data from the Optum Clinformatics research database (UnitedHealth Group) from December 1, 2010, to May 31, 2013. Participants were 1 968 709 adults with commercial insurance who had been prescribed 1 or more statins (13 285 223 statin prescriptions). An interrupted times series model was used to examine the effect of limited and full generic competition on brand-name and generic atorvastatin prescriptions. Data were analyzed from December 1, 2010, to May 31, 2013.

EXPOSURES

Prescription of brand-name atorvastatin, generic atorvastatin, and authorized generic atorvastatin were distinguished using National Drug Codes.

MAIN OUTCOMES AND MEASURES

Total number of prescriptions dispensed per month and out-of-pocket expenditures for a typical 30-day supply of 20-mg atorvastatin during the periods of brand-name availability only, limited generic competition (lasting 180 days after market exclusivity ended), and full generic competition.

RESULTS

Of the 1 968 709 beneficiaries, 1 483 066 (58.8% male and 41.2% female; mean [SD] age, 55.6 [10.2] years) received a prescription for a single statin and were included in the analysis. The introduction of the first generic competitor was associated with a reduction in monthly brand-name atorvastatin fills by 20 896 prescriptions (level change, P = .001), an 18.1% change compared with the month preceding loss of exclusivity. Full generic competition reduced brand-name fills by 54 944 prescriptions (level change, P < .001), a 47.6% change relative to the month preceding loss of exclusivity. During the first 180 days of generic competition, no meaningful difference in monthly out-of-pocket spending was found between brand-name (median, $16.98; interquartile range [IQR], $8.76-$48.66) and generic (median, $19.98; IQR, $7.50-$54.90) atorvastatin. After full generic competition, estimated monthly out-of-pocket spending for generic atorvastatin (median $5.10; IQR, $3.36-$19.98) or authorized generic atorvastatin (median, $5.52; IQR, $3.48-$19.98) was substantially lower than that for brand-name atorvastatin (median, $30.00; IQR, $15.00-$91.38).

CONCLUSIONS AND RELEVANCE

Among patients with commercial health insurance, delays in generic uptake and high levels of out-of-pocket spending during the first 180 days after atorvastatin lost market exclusivity slowed changes in drug prescribing and decreases in patients' out-of-pocket costs.

A genome-wide association study meta-analysis of clinical fracture in 10,012 African American women

Background

Osteoporosis is a major public health problem associated with excess disability and mortality. It is estimated that 50–70% of the variation in osteoporotic fracture risk is attributable to genetic factors. The purpose of this hypothesis-generating study was to identify possible genetic determinants of fracture among African American (AA) women in a GWAS meta-analysis.

Methods

Data on clinical fractures (all fractures except fingers, toes, face, skull or sternum) were analyzed among AA female participants in the Women's Health Initiative (WHI) (N = 8155), Cardiovascular Health Study (CHS) (N = 504), BioVU (N = 704), Health ABC (N = 651), and the Johnston County Osteoarthritis Project (JoCoOA) (N = 291). Affymetrix (WHI) and Illumina (Health ABC, JoCoOA, BioVU, CHS) GWAS panels were used for genotyping, and a 1:1 ratio of YRI:CEU HapMap haplotypes was used as an imputation reference panel. We used Cox proportional hazard models or logistic regression to evaluate the association of ~ 2.5 million SNPs with fracture risk, adjusting for ancestry, age, and geographic region where applicable. We conducted a fixed-effects, inverse variance-weighted meta-analysis. Genome-wide significance was set at P < 5 × 10^− 8.

Results

One SNP, rs12775980 in an intron of SVIL on chromosome 10p11.2, reached genome-wide significance (P = 4.0 × 10^− 8). Although this SNP has a low minor allele frequency (0.03), there was no evidence for heterogeneity of effects across the studies (I² = 0). This locus was not reported in any previous osteoporosis-related GWA studies. We also interrogated previously reported GWA-significant loci associated with fracture or bone mineral density in our data. One locus (SMOC1) generalized, but overall there was not substantial evidence of generalization. Possible reasons for the lack of generalization are discussed.

Conclusion

This GWAS meta-analysis of fractures in African American women identified a potentially novel locus in the supervillin gene, which encodes a platelet-associated factor and was previously associated with platelet thrombus formation in African Americans. If validated in other populations of African descent, these findings suggest potential new mechanisms involved in fracture that may be particularly important among African Americans.

•

This was a hypothesis-generating GWAS for fracture in African Americans.

•

One potentially novel locus (SVIL) was identified at GWA-significant levels.

•

SVIL has been associated with platelet thrombus formation in African-Americans.

Idiopathic Pulmonary Fibrosis in United States Automated Claims. Incidence, Prevalence, and Algorithm Validation

RATIONALE

Estimates of idiopathic pulmonary fibrosis (IPF) incidence and prevalence from electronic databases without case validation may be inaccurate.

OBJECTIVES

Develop claims algorithms to identify IPF and assess their positive predictive value (PPV) to estimate incidence and prevalence in the United States.

METHODS

We developed three algorithms to identify IPF cases in the HealthCore Integrated Research Database. Sensitive and specific algorithms were developed based on literature review and consultation with clinical experts. PPVs were assessed using medical records. A third algorithm used logistic regression modeling to generate an IPF score and was validated using a separate set of medical records. We estimated incidence and prevalence of IPF using the sensitive algorithm corrected for the PPV.

MEASUREMENTS AND MAIN RESULTS

We identified 4,598 patients using the sensitive algorithm and 2,052 patients using the specific algorithm. After medical record review, the PPVs of these algorithms using the treating clinician's diagnosis were 44.4 and 61.7%, respectively. For the IPF score, the PPV was 76.2%. Using the clinical adjudicator's diagnosis, the PPVs were 54 and 57.6%, respectively, and for the IPF score, the PPV was 83.3%. The incidence and period prevalences of IPF, corrected for the PPV, were 14.6 per 100,000 person-years and 58.7 per 100,000 persons, respectively.

CONCLUSIONS

Sensitive algorithms without correction for false positive errors overestimated incidence and prevalence of IPF. An IPF score offered the greatest PPV, but it requires further validation.