Pituitary adenomas are a common type of intracranial tumor1 that are frequently asymptomatic2,3 and slow growing.3 They are often detected either at autopsy4 or as incidental findings in patients undergoing magnetic resonance imaging (MRI) or computed tomography scans of the head,1,3 yielding rates of 10% to 20%2–4 for all pituitary tumors (symptomatic or asymptomatic). The incidence of clinically confirmed pituitary tumors is much lower, with an age-adjusted rate of 2.8 per 100,000 person-years5 and symptoms that included hypothalamic/pituitary dysfunction or visual field compromise due to large size.1
Pharmacovigilance analyses of the US Food and Drug Administration (FDA) Adverse Event Reporting System database found a disproportionately higher number of pituitary adenoma reports for risperidone, in comparison to all other drugs and to other antipsychotic drugs in the database.6 Among antipsychotics, drugs with high potency at the dopamine type 2 (D2) receptor seemed to have a higher association, and the reporting rate was higher in women.6 Approximately half of the tumors were benign, and some were associated with symptoms such as visual field defects and surgical intervention.6 A similar data-mining analysis of the World Health Organization’s adverse drug reaction database also showed an elevated reporting of pituitary tumors with risperidone and amisulpride.7 As part of routine pharmacovigilance activities, the Janssen Research and Development Benefit Risk Management team analyzed the FDA Adverse Event Reporting System database and identified disproportional reporting for benign pituitary tumors among risperidone users in April 2005.
These findings from spontaneous adverse event reporting databases may be attributable to several factors. Of most concern is the possibility that risperidone has a causal relationship to either de novo development of a pituitary tumor, or acceleration of the clinical course of a preexisting pituitary tumor. In animal studies, an increase in pituitary adenomas has been found in rodents with chronic administration of risperidone.8 However, an increased risk of pituitary tumor has not been demonstrated in humans treated with risperidone in clinical studies, and no causal association between the drug and event could be established from postmarketing surveillance.
It is also plausible that an ascertainment and/or diagnostic bias exists that results in an association between risperidone and pituitary tumors that is not actually causal, that is, the association is confounded. The suspected confounder is an increase in imaging in patients prescribed risperidone, as risperidone elevates prolactin more than other atypical antipsychotic agents,9,10 and clinical guidelines recommend that MRI be performed in all patients with hyperprolactinemia.11 If imaging is performed more frequently in patients prescribed risperidone, more clinically “silent” pituitary adenomas would be detected in these patients, producing an apparent increase in risk. Further complicating interpretation of rates is the finding that concordance among radiologists in the diagnosis of pituitary adenomas is poor, and evidence suggests that false-positive readings are common.2,12
The specificity of the diagnosis might be improved, and have more relevance, if consideration were limited to tumors associated with symptoms. Symptoms associated with pituitary adenomas include mass effects (eg, visual field disturbance) or hormonal effects that may be due to hypersecretion of prolactin or to deficiency of other pituitary hormones.1 Symptoms due to hyperprolactinemia will be the same in patients with a prolactinoma (the most common type of pituitary adenoma13) and in those receiving antipsychotic agents that increase prolactin levels through antagonism of the D2 receptor and possibly other mechanisms.14 Thus, comparisons of the incidence of hyperprolactinemia itself will not distinguish between the etiologies of tumor or the known reversible adverse effects of some antipsychotics. However, symptoms of mass effect may be useful for making this distinction.
Janssen Research and Development designed a pharmacoepidemiology study to investigate the relationship between pituitary tumors with mass effect and risperidone. On May 14, 2007, the FDA invited Janssen Research and Development to present the study design face-to-face and agreed on the design and study end points. We report here our longitudinal study comparing the incidence of newly diagnosed pituitary tumors with mass effect among patients prescribed antipsychotics. We have used large health care administrative databases as the most practical means of addressing the rarity of the event. The primary end point was designed to reduce the expected large number of false-positive findings and to focus on symptoms that are not shared by D2 antagonists. The focus was on risperidone, with use of other atypical antipsychotics as the relevant comparison group.
MATERIALS AND METHODS
This is a cohort study with a nested case-control component to investigate the relationship between clinically significant pituitary tumors and antipsychotic medication prescriptions. The prespecified primary study end point was “pituitary tumor with mass effect,” including pituitary tumors associated with visual field disturbance, pituitary hormone deficiency, or surgical or radiological treatment of the tumor. Ascertainment of a diagnosis of pituitary tumor and associated mass effect was confirmed by a review of medical records to increase validity over simple administrative codes.
Two data sources, the Department of Veterans Affairs (VA) medical care databases and the OptumInsight Life Sciences Research Database (LSRD) (formerly the Ingenix Normative Health Information Database) were used for this study owing to their large sample sizes and the ability to validate key outcomes through access to medical records. These databases were analyzed independently by investigators from VA (VA data) and OptumInsight Life Sciences (LSRD data).
Department of Veterans Affairs
Prescription, diagnostic, and identifier data for VA patients were obtained from the electronic administrative databases in the VA medical care system and linked via a unique patient identifier. The sources were the VA national patient files maintained at the Austin Information Technology Center [inpatient admissions, demographic information, International Classification of Diseases, Ninth Edition (ICD-9) diagnosis and procedure codes and Current Procedural Terminology (CPT) codes] ,and the prescription data from the Pharmacy Benefits Management Services database (PBM version 3.0) located in Tucson, AZ. Prescription data included information on the date of drug dispensation, drug name, dose, total quantity, and days’ supply of all prescriptions. These data have been used in a number of studies of drug events.15,16
OptumInsight Life Sciences Research Database
This database is a repository of health insurance claims and related administrative data reflecting the care received by a population insured through UnitedHealthcare. The claims are submitted by providers and pharmacies to obtain payment for health care services rendered, and enrollment data allow tracking of plan membership. These data capture a longitudinal record of patients’ medical services irrespective of treatment site. The demographics of the LSRD population approximate the US population for all age categories through age 64 years. In persons 65 years or older, the LSRD underrepresents the US population. These data have been used in a number of studies of drug events.17,18
The analyses were conducted separately on the 2 databases, using a common protocol, in cohorts of patients treated with an antipsychotic. The design was an open cohort. The period of cohort entry and follow-up was June 1, 1999, to September 30, 2009, for VA and July 1, 2002 to December 31, 2007, for LSRD. Each database was used to identify a cohort of patients with exposure to an antipsychotic and to follow the cohort longitudinally for ascertainment of any new diagnoses of pituitary tumors with mass effect.
Eligible patients were required to have had at least 2 dispensations of an antipsychotic medication without a gap of more than 3 months between the first and the second dispensations and with no other dispensation of antipsychotic medication within the gap. The index date was set as the date of the second dispensation. A hierarchical method of assigning exposure was used and assigned exposure preferentially as first, risperidone; second, other atypical antipsychotics (including initiators of aripiprazole, clozapine, olanzapine, quetiapine, and ziprasidone); third, other [conventional] antipsychotics). The study focused on risperidone compared to other atypical antipsychotics.
The antipsychotic exposure was required to be “new”. That is, there could be no pharmacy claim for that drug in the 6 months before the first dispensation that defined the exposure group. Eligible patients received medical and pharmacy benefits at the health plan for at least 6 months before the first dispensation of the exposure medication.
Patients were excluded if they had at least one medical claim, at any time before the start of exposure, consistent with malignancy (other than nonmelanoma skin cancer; ICD-9 codes 140–172, 174–208, 230–231, and 233–239) or any pituitary tumor. Patients with a medical claim of pregnancy (ICD-9 codes 630–677) in the baseline period were also excluded. Patients with a medical claim suggesting pregnancy during the follow-up period were censored at 9 months before the date of first documentation of pregnancy.
The cohort follow-up period began the day after the index date and lasted until a censoring event or the end of the study period (whichever occurred earlier). Censoring events were defined as: a pituitary tumor diagnosis; last claim in database; 9 months before a pregnancy code; or the end of the study period. A therapy discontinuation date was defined as the first time that a lapse in therapy (a period of days’ supply plus 30 days without another dispensation) occurred in the regimen of the study exposure; follow-up before the therapy discontinuation date was designated as the on-treatment time. If there was no discontinuation date or if the discontinuation date occurred after a patient exited his/her exposure group for conditions set out previously, all of the follow-up time for that patient was considered on-treatment.
Main Outcome Measure
Stage 1: High-Sensitivity Screen
Potential cases of new pituitary tumors with associated mass effects were identified from a diagnosis of pituitary tumor in addition to 1 of 4 clinical scenarios from inpatient or outpatient visit claims data. This stage was designed to have high sensitivity to maximize the capture of potential cases. This was followed by medical record review to eliminate the false positives inherent in a high-sensitivity screening procedure. The codes used in the first stage to capture an event that might represent a pituitary tumor were: ICD-9 diagnosis codes for neoplasms (194.3 [malignant neoplasm of pituitary gland and craniopharyngeal duct], 227.3 [benign neoplasm of pituitary gland and craniopharyngeal duct pouch], 237.0 [neoplasm of uncertain behavior of pituitary gland and craniopharyngeal duct], 377.51 [disorders of optic chiasm associated with pituitary neoplasms and disorders]) or for disorders of the pituitary gland [253.8 [other disorders of the pituitary and other syndromes of diencephalohypophyseal origin]; 253.9 [unspecified disorder of the pituitary gland and its hypothalamic control]). The subset of these with possible mass effect was identified with additional codes for any of the following qualifying scenarios for mass effect: (1) surgical or (2) radiation therapy of the pituitary after the diagnosis date; or, within 60 days before or after the diagnosis date of the pituitary tumor, a diagnosis of either (3) visual field disturbance, not explained by other ocular conditions (eg, glaucoma) or (4) pituitary hormone deficiency (panhypopituitarism, growth hormone deficiency, thyroid-stimulating hormone deficiency leading to secondary hypothyroidism, or deficiencies in adrenocorticotropic hormone resulting in secondary hypoadrenalism, excluding secondary hypogonadism).
Stage 2: Case Confirmation
All claims-based pituitary tumors with possible mass effect from stage 1 had detailed medical record review to identify true cases. The reviewers were provided access to medical records of subjects, along with the index and potential diagnosis dates but without identification of the drug exposure. The patients’ medical records were evaluated to determine whether a physician actually diagnosed a pituitary tumor, whether it was newly diagnosed and associated with mass effect, and whether any exclusion factors (eg, prior cancer) were present.
The case-control study was originally designed to be conducted within both VA and LSRD cohorts. However, because only one case was confirmed in the LSRD (see the “Results” section), the case-control analyses were conducted with VA data only. The controls were selected from cohort patients with no diagnosis of a pituitary tumor. Four controls were individually matched to each case by age (birth year ±5 years), sex, duration of follow-up in the database after the index date (± 60 days), and length of history in the database before the first study drug exposure (± 90 days). Criteria were loosened for one female patient to increase the control pool. In addition, the index date for the controls must have occurred before the case’s diagnostic date. The diagnosis date of the patients’ pituitary tumor was assigned to their matched controls for purposes of classifying exposure history. Using the end of the most proximal days’ supply date before the diagnosis date, “current” was within 30 days, “recent” was within 31 to 180 days, and “past” was more than 180 days before the case diagnosis date. Cumulative exposure was calculated by summing the days’ supply for all fills of the index drug up until the case’s diagnostic date. Average daily dose was calculated by summing daily dose over all days and dividing by the total number of days. Once the controls were selected, one randomly sampled control for each case was subject to a medical record review.
In the cohort analyses, the incidence of medical record–confirmed pituitary tumor with mass effect was compared between risperidone and other atypical antipsychotic exposures using a Cox proportional hazards regression model for the primary prespecified analysis. Incidence rates, hazard ratios (HRs), and their 95% confidence intervals (CIs) were estimated for both on-treatment time and total cohort follow-up time. Although multiple factors were considered as potential confounders or effect modifiers, including age, sex, psychiatric diagnoses, health plan coverage, antipsychotic dosage, duration of antipsychotic course, use of other prolactin-elevating drugs, and comorbidity, the limited number of confirmed cases allowed for adjustment only by age in the statistical models.
The case-control analyses compared the exposures of risperidone and other atypical antipsychotics between the cases and the controls using a conditional logistic regression model accounting for the matching factors to generate odds ratios (ORs). Comparisons of risperidone to each individual atypical antipsychotic medication among the cases and the controls were limited because of the small number of cases. Modeling was carried out only when there were at least 10 cases across the relevant exposure groups. Statistical Analysis Software version 9.2 was used for all analyses (SAS Institute, Cary, NC).
The original protocol called for conducting independent analyses in the 2 databases and then combining the summary measures. Preliminary analyses of the databases estimated 178 potential cases. We estimated that at least 70 cases would be true positives and that this would enable us to detect a risk as small as 2.2. with more than 80% power at α = 0.05. However, because of the low yield of cases in the LSRD data, the findings will focus on the results from the VA data, with a separate description of LSRD results when sufficient numbers permit.
The study followed the Health Insurance Portability and Accountability Act guidelines for protection of patient confidentiality and operated with the oversight of institutional review boards and privacy boards at VA and OptumInsight, granting a waiver of authorization to allow the use of protected health information. The study was carried out in accordance with the Declaration of Helsinki. The sponsor, Janssen Research and Development, designed the study after agreement with FDA on the primary outcome definition (pituitary tumor with mass effect), in collaboration with VA and OptumInsight. Janssen Research and Development had no access to raw data or patients’ records. The creation of the data sets and analysis were conducted by VA and OptumInsight according to the protocol under contracts with JRD. The decisions about presentation of the data in this report were made by the primary author from VA.
VA Cohort Analysis
In the VA database, a total of 370,006 patients were classified according to the study criteria for antipsychotic prescriptions. Table 1 summarizes the characteristics of the patients. Most of the patients were men (90%), with the mean ages 55 to 60 years old. Because the pharmacy database does not include the indications, diagnoses of major psychiatric disorders within 6 months before the index date are presented. Depression was the most common, followed by posttraumatic stress disorder. Reflecting the hierarchical method in which the exposure groups were defined (prioritizing risperidone), risperidone patients were least likely (65%) to have no prior prescriptions of other antipsychotics.
Hyperprolactinemia was infrequently documented based on the ICD-9 codes (0.03% in risperidone, 0.02% in other atypical, and 0.02% in conventional antipsychotic cohorts); however, the actual incidence of hyperprolactinemia could not be reliably assessed because prolactin is not routinely measured in asymptomatic individuals. The presence of common prolactin-elevating conditions and medications19 was infrequent. Hypothyroidism was the most common of these findings (3.9%, 4.0%, and 4.3% in risperidone, other atypical, and conventional groups, respectively). Use of metoclopramide or domperidone was infrequent but higher in the conventional group (0.34%, 0.28%, and 1.58%, respectively).The median number of months between the 2 exposure-defining dispensations was 1.0, 0.9, and 1.5 for risperidone, other atypical, and conventional antipsychotics, respectively. The 339,574 patients who were prescribed atypical antipsychotics (including risperidone) were the focus of the comparative study. Comparisons within the same class are most relevant, particularly because conventional antipsychotics were used much less frequently and were associated with a somewhat different patient profile (Table 1).
Figure 1 illustrates the flow of patients through the case-finding stages. In VA, a total of 807 (risperidone, 492; other atypical antipsychotics, 315) patients were identified as possible cases in the initial step of stage 1, the highest sensitivity step selecting for any diagnosis of pituitary tumor from the administrative database (Fig. 1, left panel). The requirement for administrative codes consistent with mass effect in the second step of stage 1 reduced this number to 148 patients. Application of stage 2 criteria during medical record review eliminated 109 patients, leaving 20 true cases of patients exposed to risperidone and 19 patients exposed to other atypical antipsychotics. Of the 148 patients entering stage 2, only 95 patients (64%) actually had evidence of a pituitary tumor, of which 77 patients (81%) were incident/new and 40 (52%) of the 77 patients had mass effect. Of the 40 patients with incident tumor with mass effect, one was subsequently excluded owing to a history of cancer, leaving 39 eligible cases. This represented a confirmation rate of 26% as true new pituitary tumors with mass effect.
The median total study follow-up time was 47 months (interquartile range, 21–74 months) in the risperidone group and 39 months (interquartile range, 18–48 months) in the other atypical antipsychotic group. The event rate in the risperidone group was 20 events/786,132 person-years (2.5 per 100,000 person-years). The rate was identical in the other atypical group (19 events/752,121 person-years), resulting in a crude HR of 1.0 (95% CI, 0.5–1.9). This effect size was unchanged when adjusted for age. These findings reflect total study follow-up time, as only 10% to 20% of the tumors were diagnosed during active treatment. The pattern during active treatment time was similar.
Various subgroup analyses were conducted but were difficult to interpret owing to the small number of confirmed cases. These analyses included the following: limiting person-time to active treatment time, limiting tumors to those with surgical intervention, including only patients who were treatment-naive (in the 6-month baseline period), and stratifying rates by concurrent use of selective serotonin reuptake inhibitors. None of these HRs for risperidone risk exceeded 1.1 (data not shown).
LSRD Cohort Analysis
Case finding in LSRD data proceeded similarly and independently (Fig. 1, right panel). This yielded an incidence rate of 0 event/32,959 person-years for risperidone and 1 event/70,097 person-years (1.4 per 100,000) for other atypical antipsychotics (olanzapine). The single case involved a male, although the underlying sample involved 54% women. No further analysis was done on this single case.
VA Nested Case-Control Analysis
The cases and the controls were reasonably well matched. For the cases and the controls, respectively, the values were: available history (700 vs 698 days), follow-up time (1220 vs 1216 days), age (57 vs 52 years), and year of index prescription (2005 vs 2005). The case-control analysis revealed no association between pituitary tumor with mass effect and risperidone prescriptions (OR, 1.0 for any history of risperidone use; Table 2). We repeated the analysis comparing risperidone with individual drugs. Patients with study exposures of quetiapine (n = 97,102) and olanzapine (n = 56,743) together represented 84% of the comparison cohort. When using only quetiapine as the reference, the OR for risperidone was 0.8 (95% CI, 0.3–1.9). With olanzapine as the reference, the OR for risperidone was 1.3 (95% CI, 0.5–3.4).
We attempted to assess risk according to timing of risperidone use, but small numbers limit interpretation (Table 2). The ORs for current (within 30 days before diagnostic date) and past (>180 days before diagnostic date) risperidone exposure were 1.2 and 1.8, respectively. The CIs were very wide and included 1.0.
Because the small number of cases limits the precision of our estimates, we conducted post hoc analyses to add to the body of evidence to aid interpretation. In Table 3, we show estimates for the cases and the controls with exposure to the same drug (or drug class) in a given row. The estimates of exposure time tended to be shorter in the cases compared to the controls. Whereas average daily dose of risperidone tended to be 23% higher in the cases compared to the controls, a similar pattern was seen with quetiapine (56% higher in the cases). A search for evidence of hyperprolactinemia (ICD-9 code: 253.1) was conducted in the claims data for the 39 cases in the year before and after the tumor diagnosis. Two cases in the risperidone group had claims evidence of hyperprolactinemia in the 1 to 2 months before the tumor diagnosis (data not shown).
There is an undisputed need for methods that allow early detection of serious adverse drug events. Such methods are designed to screen very large databases in search of potential safety signals, using data mining techniques. The methods must strike a balance between sensitivity and specificity, as the goal is to identify all important safety events without excessive false positives.19 Almenoff et al20 emphasize that the methods identify “observed reporting relationships” and, as such, reporting bias is a recognized concern “that no signal detection system is likely to overcome.” They emphasize that the procedures result in hypotheses, which must then be further investigated. For example, FDA’s own follow-up study of a potential signal with statins did not confirm the pharmacovigilance findings.21 The methodology and role of data mining in pharmacovigilance was described in 2005 as a “work in progress”.20
Our study was designed in response to data mining findings in 2 databases.6,7 Overall, our study demonstrated very low incidence rates of confirmed pituitary tumors with mass effect among patients with atypical antipsychotic prescriptions, consistent with rates in unselected populations.5 However, the low rates also resulted in insufficient precision in the estimates to exclude some risk associated with risperidone compared to other drugs in the same class. To address this, we conducted additional analyses to evaluate whether there were trends consistent with elevated risk with risperidone.
The FDA’s pharmacovigilance findings would lead to the hypothesis that risk magnitude would depend on the drug, that is, patients prescribed risperidone should show the highest risk, those prescribed olanzapine would be at intermediate risk, and those prescribed quetiapine would be at lowest risk. Our analyses did not show this. The data mining study also found increased risk among women. We did not find women disproportionately represented among cases. The LSRD database comprised 54% women, yet the only case found involved a man. In VA data, the proportion of women (5%) among the cases did not exceed the representation of women overall (7%) in the VA cohort. In other analyses, we did not find that cases’ duration of exposure to risperidone exceeded that of the controls, and we found only a slightly higher dose of risperidone (and quetiapine) in the cases. In contrast, we did find evidence to support the hypothesis that detection bias could account for the pharmacovigilance findings. In Figure 1, with the high-sensitivity screen in the earliest step of case finding, there is an excess of potential cases in the risperidone cohort, resulting in an HR of 1.5 (95% CI, 1.3–1.8) at this point. However, at the completion of stage 1, as the specificity of the case finding increased, the HR diminished to 1.2 (95% CI, 0.9–1.7), with further diminution after medical record review. Because 48% of those eventually subjected to chart review were not confirmed because of either no evidence of pituitary tumor or a finding that the tumor predated the exposure, it is reasonable to expect that many of the potential events found in stage 1 would also lack medical record confirmation. Similar trends in the case-finding stages were seen in both data sets, which had been analyzed by different teams with different populations, thereby increasing confidence in the findings.
It is well known that analyses of claims databases are subject to biases.22 However, a prospective study designed to avoid these biases was not practical because of the number of subjects and duration of follow-up time that would be required. Such prerequisites would delay formal evaluation of the potential safety signal identified during pharmacovigilance. In addition, some weaknesses are unavoidable when studying actual practice, for example, patients switching medications or becoming lost to follow-up. Thus, we attempted to address the anticipated biases through design and analysis.
Because we could not assign treatment, our approach for defining comparison groups retrospectively was to classify exposure to risperidone preferentially. This ensured that the comparison group had no exposure to risperidone but with the tradeoff that some in the risperidone group had exposure to another antipsychotic agent. We compared risk in the subgroup of patients who were treatment-naive upon cohort entry and found the results were consistent with no risk elevation. We attempted to evaluate subgroups based on timing of exposure to their study drug. This analysis was underpowered but showed no marked temporal trends, although the potential latency period for tumor induction, if there were a causal relationship, is not known.
To address concerns in outcome ascertainment, we increased the probability of capturing all relevant outcomes by using codes that would increase sensitivity. We then addressed the specificity of the outcome data by supplementing with chart review. Restricting the outcome to tumors with mass effect was designed to not only make the outcome more clinically relevant but also to reduce ascertainment or referral bias. The case-control method matched on various time parameters to minimize temporal bias. Although underascertainment is quite possible if veterans seek care outside of VA, there is no reason to suspect that such underascertainment would affect one exposure group more than another.
The study has several limitations. First, it remains possible that risk will emerge with increasing follow-up and/or exposure time, as pituitary tumors are slow growing.3 Second, this study did not address the possibility that risperidone might increase the risk of microadenomas and/or prolactinomas with no mass effect symptoms. To address either event would require a very large prospective study with patient consent to enable routine screening MRIs. Such a study would also need a protocol for diagnostic workup,23,24 ideally masked to exposure. A cost-effective way to accomplish such an investigation might be to partner with an existing tumor registry and study those patients treated with antipsychotics. Finally, there was the potential for conflict of interest, given that the study was funded by Janssen Research and Development. The precautions taken were to consult with FDA on the outcome definition, to involve the VA and OptumInsight investigators in protocol development, and to explicitly state in the contracts that VA and OptumInsight would have autonomy over the conduct, interpretation, and publication of the findings.
In conclusion, we found a low rate of pituitary tumors associated with mass effect in patients prescribed atypical antipsychotic medication. Whereas these results are based on small numbers of cases and should be interpreted with caution, there was little evidence for increased risk associated with risperidone.
AUTHOR DISCLOSURE INFORMATION
At the time the work was conducted, Drs McCarren and Wang were employees at the VA and Dr Jiang was an employee of the Chicago Association for Research and Education in Science (CARES), a nonprofit organization associated with the Hines VA Hospital. These 3 authors have no other conflicts to report. At the time the work was conducted, Ms Ziyadeh and Dr McAfee were full-time employees of OptumInsight, a private contract research organization whose clients include the Food and Drug Administration (FDA) and manufacturers of drugs, vaccines, and medical devices. Dr Qui is an employee of JRD.
1. Davis JRE, Farrell WE, Clayton RN. Pituitary tumours. Reproduction
. 2001; 121: 363–371.
2. Hall WA, Luciano MG, Doppman JL, et al.. Pituitary magnetic resonance imaging in normal human volunteers: occult adenomas in the general population. Ann Intern Med
. 1994; 120: 817–820.
3. Mavrakis AN, Tritos NA. Diagnostic and therapeutic approach to pituitary incidentalomas. Endocr Pract
. 2004; 10: 438–444.
4. Molitch ME, Russell EJ. The pituitary “incidentaloma”. Ann Intern Med
. 1990; 112: 925–931.
5. CBTRUS. Statistical Report: Primary Brain and Central Nervous System Tumors Diagnosed in the United States, 2004–2006. Published by the Central Brain Tumor Registry of the United States. February 2010. Available at: http://www.cbtrus.org/reports/reports.html
. Accessed November 19, 2010.
6. Szarfman A, Tonning JM, Levine JG, et al.. Atypical antipsychotics
and pituitary tumors: a pharmacovigilance
. 2006; 26: 748–758.
7. Doraiswamy PM, Schott G, Star K, et al.. Atypical antipsychotics
and pituitary neoplasms in the WHO database. Psychopharmacol Bull
. 2007; 40: 74–76.
) [package insert]. Titusville, NJ: Janssen, 2006. Available at: http://risperdal.com/sites/default/files/shared/pi/risperdal.pdf
. Accessed February 7, 2011.
9. Montgomery J, Winterbottom E, Jessani M, et al.. Prevalence of hyperprolactinemia in schizophrenia: association with typical and atypical antipsychotic treatment. J Clin Psychiatry
. 2004; 65: 1491–1498.
10. Melkersson K. Differences in prolactin elevation and related symptoms of atypical antipsychotics
in schizophrenic patients. J Clin Psychiatry
. 2005; 66: 761–767.
11. Harrison’s Online, Chapter 333. Disorders of the anterior pituitary and hypothalamus: introduction. In: Fauci AS, Braunwald E, Kasper DL, et al., eds. Harrisons Principles of Internal Medicine, 17e
. New York, NY: McGraw-Hill; 2008. Available at: http://www.accessmedicine.com/resourceTOC.aspx?resourceID=4
. Accessed April 25, 2011.
12. Chanson P, Daujat F, Young J, et al.. Normal pituitary hypertrophy as a frequent cause of pituitary incidentaloma: a follow-up study. J Clin Endocrinol Metab
. 2001; 86: 3009–3015.
13. Mindermann T, Wilson CB. Age-related and gender-related occurrence of pituitary adenomas. Clin Endocrinol
. 1994; 41: 359–364.
14. Halbreich U, Kinon BJ, Gilmore JA, et al.. Elevated prolactin levels in patients with schizophrenia: mechanisms and related adverse effects. Psychoneuroendocrinology
. 2003; 28: 53–67.
15. Aspinall SL, Good CB, Jiang R, et al.. Severe dysglycemias with the fluoroquinolones: a class effect? Clin Infect Dis
. 2009; 49: 402–409.
16. Caplan L, Pittman CB, Zeringue AL, et al.. An observational study of musculoskeletal pain among patients receiving bisphosphonate therapy. Mayo Clin Proc
. 2010; 85: 341–348.
17. Enger C, Cali C, Walker AM. Serious ventricular arrhythmias among users of cisapride and other QT-prolonging agents in the United States. Pharmacoepidemiol Drug Saf
. 2002; 11: 477–486.
18. Johannes CB, Koro CE, Quinn SG, et al.. The risk of coronary heart disease in type 2 diabetic patients exposed to thiazolidinediones compared to metformin and sulfonylurea therapy. Pharmacoepidemiol Drug Saf
. 2007; 16: 504–512.
19. Szarfman A, Tonning JM, Doraiswamy PM. Pharmacovigilance
in the 21st century: new systematic tools for an old problem. Pharmacotherapy
. 2004; 24: 1099–1104.
20. Almenoff J, Tonning JM, Gould AL, et al.. Perspectives on the use of data mining in pharmaco-vigilance. Drug Saf
. 2005; 28: 981–1007.
21. Colman E, Szarfman A, Wyeth J, et al.. An evaluation of a data mining signal for amyotrophic lateral sclerosis and statins detected in FDA’s spontaneous adverse event reporting system. Pharmacoepidemiol Drug Saf
. 2008; 17: 1068–1076.
22. Mitchell JB, Bubolz T, Paul JE, et al.. Using Medicare claims for outcomes research. Med Care
. 1994; 32: JS38–JS51.
23. Molitch ME. Medication-induced hyperprolactinemia. Mayo Clin Proc
. 2005; 80: 1050–1057.
24. Melmed S, Casanueva FF, Hoffman AR, et al.. Diagnosis and treatment of hyperprolactinemia: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab
. 2011; 96: 273–288.