Management of glucocorticoid-induced osteoporosis in patients with rheumatoid arthritis: Rates and predictors of care in an academic rheumatology practice

Authors


Abstract

Objective

To examine rheumatologists' management practices for patients with rheumatoid arthritis (RA) taking glucocorticoids seen at one academic medical rheumatology practice.

Methods

All patients diagnosed with RA making at least 2 visits to the practice who were noted to be taking glucocorticoids on an index visit in January or February 1999 were included in a retrospective chart review.

Results

Of the 623 eligible patients identified with RA, 236 patients were taking glucocorticoids at their index visit and were included in this study. The mean daily dosage of prednisone (or its equivalent) was 8.8 mg and the mean duration of use during the 2-year chart review was 15.4 months. Twenty-three percent of the study population underwent bone densitometry, and 42% were prescribed a medication that reduces bone loss (not including calcium and/or vitamin D). Calcium and/or vitamin D were noted on the medication lists of 25% of patients. Factors associated with not undergoing bone densitometry in adjusted logistic models included male patient sex and premenopausal status in women. No physician factors were significantly related to performing bone densitometry. Variables associated with not receiving prescription treatment of osteoporosis were male patient sex, premenopausal status, and having at least 1 comorbid condition.

Conclusion

Patients with RA taking oral glucocorticoids did not routinely undergo bone densitometry and/or receive prescription medications for osteoporosis. Men and premenopausal women were less likely to undergo bone densitometry and to receive a prescription medication for osteoporosis than postmenopausal women. Interventions to improve detection and prevention of glucocorticoid-induced osteoporosis are necessary.

Glucocorticoid-induced osteoporosis (GIOP) is a frequent and disabling clinical condition for patients with inflammatory arthritis. Patients with a variety of systemic rheumatic diseases have higher rates of osteoporosis, thought to be secondary to the increased rates of bone resorption associated with inflammatory arthritis, physical disability and cachexia, and glucocorticoid-mediated bone loss (1–3). The American College of Rheumatology (ACR; formerly, the American Rheumatism Association) has published guidelines regarding screening and treatment for GIOP (4, 5), but there are indications that patients are not being managed according to these recommendations.

Studies conducted over the last 6 years have found consistently low rates of osteoporosis management for patients taking long-term oral glucocorticoids. In the UK, Peat and colleagues reported that 6% of 214 patients hospitalized in a teaching institution and taking oral glucocorticoids were receiving any osteoporosis medication (6). In a survey from the UK of ambulatory patients taking oral glucocorticoids for at least 3 months, 14% of patients had received treatment for the prevention of osteoporosis (7). Two recent reports from the US described slightly higher rates of osteoporosis treatment for patients taking oral glucocorticoids. A chart review of subspecialists, including rheumatologists, showed that 29% of patients received some medication for osteoporosis (8). Another review from a large health maintenance organization suggested that ∼30% of patients received a prescription medication for osteoporosis and a slightly higher proportion received calcium or vitamin D (9). Both of the studies from the US showed that the rate of osteoporosis treatment varied by the subspecialty of the treating physician, with rheumatologists treating GIOP more often than other types of physicians. An Australian study had similar findings (10).

These data suggest that GIOP management is suboptimal and that adherence to recommendations may differ by physician training. To further understand GIOP management by rheumatologists and gain information for planning future quality improvement interventions, we reviewed the records of patients with rheumatoid arthritis (RA) seen at an academic rheumatology practice. Specifically, we examined whether patient and/or physician factors were associated with patterns of GIOP management.

PATIENTS AND METHODS

Study setting.

Patients who were seen in January or February 1999 for a followup visit in the rheumatology practice at a large academic medical center were eligible for the study. The rheumatology practice is hospital-based in a major metropolitan center in the US. Thirty-seven rheumatologists were in active practice in the rheumatology practice in 1999, and 20,500 patient visits were accrued. The diagnosis codes of all followup visits were searched, and patients with a diagnosis code of RA (ICD-9-CM 714.0) had their medical records reviewed.

The first visit during these 2 months was defined as the index visit. Beginning with this visit, the physician's notes were reviewed to determine whether the rheumatologist had indeed given patients the diagnosis of RA. Patients not taking oral glucocorticoids at the time of his/her index visit were excluded. Finally, we excluded patients without available electronic medical records. This exclusion criterion primarily affected the patients of 3 rheumatologists, 2 who handwrite their visit notes and 1 whose notes were not included in the electronic medical record.

Data sources.

Data were collected from the patients' electronic medical record using a structured chart abstraction form (available upon request). All information available from the 12 months prior to the index visit and the 12 subsequent months was reviewed. The primary source of information was the treating rheumatologist's visit notes. However, the electronic medical record contains the visit notes from most ambulatory and inpatient services conducted at the study hospital. We reviewed the visit notes as well as laboratory and radiology studies, including bone densitometry, available on the hospital's information system. Results from studies performed outside the hospital are not available in the hospital's information system and are inconsistently described in visit notes. We included such data where available. Medication information was gleaned from the rheumatology visit notes as well as the primary care medication list. This list is available for patients with primary care doctors at the study hospital. Information regarding physician characteristics was obtained from the office of the hospital's professional staff. All aspects of this study were approved by the appropriate Institutional Review Board.

Outcomes: GIOP management.

Bone density testing and use of osteoporosis medication were all assessed as markers of GIOP management. These items have been included in the ACR's recommendations as part of the management of patients taking long-term glucocorticoids (4, 5). Specifically, we examined whether the patient had undergone bone densitometry, primarily dual x-ray absorptiometry, and the results of such testing. We also collected information about medications used for osteoporosis, such as calcium, vitamin D (supplements and prescription), hormone replacement therapy (HRT, including transdermal), alendronate, etidronate, raloxifene, pamidronate, and calcitonin (all forms). In the US, risedronate was not approved for osteoporosis until after our study period. Information on all such items was determined for the 12 months prior to the index date and for the subsequent 12 months.

Predictor variables.

Glucocorticoid exposure.

To assess the use of oral glucocorticoids during the 2-year study period, we examined the dosage during the first visit in each of the 8 3-month periods (quarters), 4 quarters during the 12 months prior to the index date and 4 during the subsequent 12 months. The dosage was converted to prednisone equivalents and then averaged over the 2 years by adding the daily dosage from each quarter and dividing by 8. We created 6 clinically relevant dosage-duration categories for oral glucocorticoid use. These consisted of patients taking <5 mg/day of prednisone or its equivalent for <6 months, <5 mg/day for at least 6 months, 5–7.4 mg/day for <6 months, 5–7.4 mg/day for at least 6 months, ≥7.5 mg/day for <6 months, and ≥7.5 mg/day for at least 6 months.

Patient characteristics.

Patients' date of birth, sex, and insurance type were all assessed from the hospital information system. Physician visit notes through 1990 are available on the electronic medical record, and they were searched to determine the year of diagnosis of RA. If the onset of RA was noted prior to 1990, but no year could be determined, then the duration of disease was described as 10 years. We attempted to obtain information on the ACR functional status (11), but this information was not recorded for most patients and could not be reliably inferred. Information on rheumatoid factor (RF) status was collected; when unavailable, it was assumed to be negative. Comorbid medical conditions were assessed based on the rheumatology visit notes and the problem list noted in the primary care section of the electronic medical record. Such illnesses included hypertension, diabetes mellitus, ischemic heart disease, cancer, depression, kidney disease, liver disease, chronic lung diseases, stroke, peptic ulcer disease, thyroid disease, and a solid organ transplant. We also recorded information about the patients' current use of tobacco and menopausal status. Finally, we determined what disease-modifying antirheumatic drugs (DMARDs) the patients had been taking during the 2-year window. No information on dosage or duration of these medications was collected.

Physician characteristics.

Patients seen at this academic medical center were seen longitudinally by the same rheumatologist. We identified each doctor with an anonymous code, and the characteristics of each rheumatologist, such as sex, years since training, and percent of week spent in patient care, were assessed. Also, we determined whether the patient had been seen by other doctors at the same center during the 2-year study period.

Statistical analysis.

We described the patient and physician populations using descriptive statistics. Mean glucocorticoid dosage and duration were calculated as described above. Then, we calculated the proportion of patients who underwent bone densitometry and/or received an osteoporosis medication in each of the oral glucocorticoid dosage-duration categories. We also examined the proportion of patients with a T score (at any site) of less than or equal to −2.5 who received an osteoporosis medication.

The associations between selected aspects of GIOP management and patient and/or physician characteristics were examined using SAS (release 7.0; SAS Institute, Cary, NC). Since we were interested in determining why some patients do not receive selected aspects of the recommended management for GIOP, i.e., bone densitometry and osteoporosis medications, the dependent variables in our regression models were not receiving bone densitometry and not receiving a prescription medication. The independent variables included patient and physician characteristics as well as the glucocorticoid dosage-duration categories, which were entered as indicator variables. First, these relationships were examined in unadjusted models. Variables with P values less than 0.2 were then considered in multivariable logistic models. Patient age, sex, and glucocorticoid dosage-duration categories were also forced into the multivariable models. We also tested glucocorticoid dosage and duration as separate variables in the multivariable model. These secondary analyses gave qualitatively similar results, and thus only the primary analyses are presented. Finally, to determine whether the adjusted odds ratios for the independent variables could be attributed to the variation in care by individual physicians, we examined mixed effects models where the physician was treated as the random effect (clustering variable) and other variables were treated as fixed effects (MLwiN software, version 2.1a; Centre for Multilevel Modelling, London, UK). None of the parameter estimates from the final models changed by more than 10%, and parameter estimates and P values were similar; thus, we concluded that there were no significant effects of individual physicians. We have only presented the fixed-effect, adjusted logistic models.

RESULTS

Six hundred twenty-three people were identified who had been seen during January or February 1999 and whose rheumatologist had given them the diagnosis of RA. Eighty-four (13%) had no clear records of their medication history and thus were excluded. From this group of 539 eligible patients, 236 (44%) were taking oral glucocorticoids at their index visit and were included in the study cohort.

The characteristics of the members of the study cohort and their rheumatologists are listed in Table 1. The mean patient age was 60 years, and 80% were women. Mean disease duration was 8.8 years, and 68% of patients with available data were RF positive. Almost all patients received a DMARD at some point during the 2-year study period, and 20% received 3 or more DMARDs. Thirty-one percent had at least 1 comorbid condition. Two-thirds of the women were postmenopausal. Twenty-nine rheumatologists treated the 236 patients with RA. Five (17%) of the rheumatologists were women, and the mean number of years since finishing rheumatology fellowship training was 19. Eighty-three percent of the included rheumatologists saw patients fewer than 5 half-days per week.

Table 1. Characteristics of patients and rheumatologists*
  • *

    Except where indicated otherwise, values are the number (%). DMARD = disease-modifying antirheumatic drug. See text for explanation of comorbid medical conditions.

  • Duration of disease truncated at 10 years for patients without records prior to 1990.

  • Information regarding rheumatoid factor status was available for 148 patients.

  • §

    DMARD use reflects mention of prescribing a DMARD at any visit during the 2-year study period, not necessarily concurrent use.

  • Postmenopausal status was calculated among the 189 women.

Patients (n = 236) 
 Age, mean ± SD years60.3 ± 14.4
 Sex, % female80.1
 Duration of disease, mean ± SD years8.8 ± 5.1
 Rheumatoid factor positive101 (68)
 DMARD use§227 (96)
  One99 (42)
  Two82 (35)
  Three or more47 (20)
 Comorbid medical conditions, mean ± SD1.0 ± 1.0
  One73 (31)
  Two or more67 (28)
 Postmenopausal126 (67)
 Current tobacco use23 (10)
Rheumatologists (n = 29) 
 Sex, % female5 (17)
 Years since fellowship, mean ± SD18.6 ± 8.7
 Time spent in patient care each week 
  Fewer than two half-day sessions18 (62)
  Three to five half-day sessions6 (21)
  Five or more half-day sessions5 (17)

The study patients' exposure to oral glucocorticoids is shown in Table 2. The mean daily dosage was 8.8 mg of prednisone (or its equivalent), and the mean duration of use was 15.4 months of the 2-year study period. We grouped patients into dosage-duration categories as listed in Table 2. Forty-five percent of patients were found to have used at least 7.5 mg/day of prednisone for at least 6 months, and 1% of patients were found to have taken <5 mg/day of prednisone for <6 months.

Table 2. Patterns of glucocorticoid use during the 2-year study period*
  • *

    Except where indicated otherwise, values are the number (%).

  • Maximum duration was 24 months. The dosage and duration categories were exclusive, such that patients were grouped into the highest dosage–longest duration category that they fulfilled. In other words, a patient who received 6 months of 5 mg/day of prednisone and 6 months of 10 mg/day of prednisone during the 24-month study period would be grouped into the ≥7.5 mg, ≥6 month category.

Dosage, mean ± SD mg/day8.8 ± 3.5
Duration, mean ± SD months15.4 ± 3.3
Dosage, duration 
 <5 mg/day, <6 months3 (1)
 <5 mg/day, ≥6 months27 (11)
 5–7.4 mg/day, <6 months21 (9)
 5–7.4 mg/day, ≥6 months44 (19)
 ≥7.5 mg/day, <6 months36 (15)
 ≥7.5 mg/day, ≥6 months106 (45)

Figure 1a illustrates the rates of bone densitometry during the 2-year study period for patients in each dosage-duration category. Overall, 23% of patients underwent bone densitometry. There did not appear to be a trend toward higher rates of bone density testing for patients in higher dosage–longer duration categories (P = 0.5). Figure 1b displays the rates of prescription medication use during the 2-year study period. Forty-two percent of all patients were prescribed a medication for osteoporosis, including HRT. After excluding HRT, 29% were prescribed a medication. Specific treatments included alendronate in 45 people (19%), etidronate in 5 (2%), calcitonin in 23 (10%), HRT in 51 (22%), and raloxifene in 7 (3%). If we included calcium and vitamin D in this group of osteoporosis treatments, the treatment rate was 61%; however, 75% had no record of receiving calcium or vitamin D during the study period. There was no trend in rates of treatment with prescription medications for osteoporosis across glucocorticoid dosage-duration categories (P = 0.7). However, the rate of treatment for osteoporosis, including HRT, was significantly higher for postmenopausal women (67%) than for premenopausal women (33%) and men (4%) (P < 0.001 for trend).

Figure 1.

Glucocorticoid-induced osteoporosis management. a, Bone densitometry use, by glucocorticoid category. b, Use of prescription medication for osteoporosis, by glucocorticoid category. P values were determined by chi-square test of trend. No patients in the lowest glucocorticoid dose-duration category had bone densitometry. Prescription medications included hormone replacement therapy, raloxifene, etidronate, alendronate, pamidronate, and calcitonin.

Next, we examined the factors associated with not undergoing bone densitometry and not receiving a prescription medication. Two separate multivariable logistic models are shown in Tables 3 and 4. All variables noted in the tables were tested in unadjusted models, and then selected factors were tested in the multivariable models. Men and premenopausal women were significantly more likely to not undergo bone densitometry and not receive prescription osteoporosis treatment. Also, having 1 or more comorbid medical conditions was associated with not receiving prescription treatment. The glucocorticoid dosage-duration categories were not associated with undergoing bone densitometry or prescription osteoporosis treatment.

Table 3. Factors associated with not undergoing bone densitometry during the 2-year study period*
FactorCrude OR (95% CI)Adjusted OR (95% CI)
  • *

    For ease of interpretation, these statements are configured so that all odds ratios (ORs) reflect the risk of not undergoing bone densitometry. The glucocorticoid dosage-duration categories are described in the text. All variables with P < 0.2, patient age, patient sex, and glucocorticoid dosage-duration categories were included in the multivariable model. The multivariable model C statistic = 0.73. 95% CI = 95% confidence interval; NT = variables that were not tested in multivariable models; DMARD = disease-modifying antirheumatic drug; NA = unable to calculate since no patients in this glucocorticoid dosage-duration category underwent bone densitometry.

  • P < 0.2.

  • P < 0.01.

  • §

    P < 0.05.

  • P < 0.001.

Patient age, <65 years0.6 (0.3–1.2)0.4 (0.2–1.0)
Patient sex, male8.5 (2.0–36.5)3.5 (1.1–16.9)§
Disease duration, per year1.0 (1.0–1.1)NT
Rheumatoid factor status, negative1.1 (0.8–1.5)NT
DMARD use, per additional drug1.0 (0.7–1.4)NT
Glucocorticoid dosage, duration  
 ≥7.5 mg/day, ≥6 months1.01.0
 ≥7.5 mg/day, <6 months1.9 (0.8–4.4)1.4 (0.6–3.5)
 5–7.4 mg/day, ≥6 months1.3 (0.6–2.9)0.9 (0.4–2.3)
 5–7.4 mg/day, <6 months1.3 (0.4–3.9)1.1 (0.3–3.5)
 <5 mg/day, ≥6 months0.7 (0.2–2.1)0.6 (0.2–2.1)
 <5 mg/day, <6 monthsNANA
Comorbid medical conditions, ≥11.6 (0.9–3.0)0.7 (0.3–1.3)
Premenopausal status3.2 (1.6–6.3)3.0 (1.3–7.0)
Thyroid disease, absence of2.1 (0.9–4.8)1.8 (0.7–4.5)
Tobacco use, absence of0.8 (0.3–2.2)NT
Physician years since training, ≥100.9 (0.5–1.8)NT
Physician sex, male1.2 (0.6–2.4)NT
Physician's clinical sessions, ≥40.9 (0.5–1.8)NT
Table 4. Factors associated with not receiving prescription medications during the 2-year study period*
FactorCrude OR (95% CI)Adjusted OR (95% CI)
  • *

    For ease of interpretation, these statements are configured so that all odds ratios (ORs) reflect the risk of not receiving a prescription medication for osteoporosis (hormone replacement therapy, raloxifene, etidronate, alendronate, pamidronate, or calcitonin). The glucocorticoid dosage-duration categories are described in the text. All variables with P < 0.2, patient age, patient sex, and glucocorticoid dosage-duration categories were included in the multivariable model. The multivariable model C statistic = 0.85. 95% CI = 95% confidence interval; NT = variables that were not tested in multivariable models; DMARD = disease-modifying antirheumatic drug.

  • P < 0.05.

  • P < 0.001.

  • §

    P < 0.2.

Patient age, <65 years1.7 (1.0–2.9)1.1 (0.5–2.4)
Patient sex, male24.2 (5.7–102.8)6.0 (1.2–29.2)
Disease duration, per year1.0 (1.0–1.1)NT
Rheumatoid factor status, negative1.1 (0.8–1.4)NT
DMARD use, per additional drug1.0 (0.7–1.2)NT
Glucocorticoid dosage, duration  
 ≥7.5 mg/day, ≥6 months1.01.0
 ≥7.5 mg/day, <6 months0.8 (0.4–1.8)0.5 (0.2–1.2)
 5–7.4 mg/day, ≥6 months1.2 (0.6–2.4)0.6 (0.2–1.5)
 5–7.4 mg/day, <6 months0.6 (0.2–1.6)0.3 (0.1–0.9)
 <5 mg/day, ≥6 months0.9 (0.4–2.1)0.5 (0.2–1.7)
 <5 mg/day, <6 months2.6 (0.2–29.7)1.9 (0.1–47.7)
Comorbid medical conditions, ≥11.7 (1.0–3.3)§2.0 (1.1–5.0)
Premenopausal status11.8 (6.2–22.4)9.0 (3.8–21.4)
Thyroid disease, absence of2.0 (0.9–4.4)§1.3 (0.5–3.4)
Tobacco use, absence of1.8 (0.7–4.5)NT
Physician years since training, ≥101.7 (0.9–3.2)§1.0 (0.4–2.7)
Physician sex, male1.1 (0.6–1.9) 
Physician's clinical sessions, ≥41.8 (1.0–3.3)§1.8 (0.7–4.7)

We further examined the rates of osteoporosis treatment based on bone densitometry (Table 5). Of the 42 patients for whom bone densitometry results were available, 19 of 20 (95%) with a T score less than or equal to −2.5 received a prescription medication. Rates of prescription medication use were lower for patients with better bone density T scores (P = 0.005, chi-square test of trend). Among the patients who did not undergo bone densitometry, the rate of prescription osteoporosis treatment was 35%.

Table 5. Osteoporosis treatment after bone densitometry, by T score*
Osteoporosis treatmentLowest BMD T score
>−1 (n = 5)−1 to >−2.5 (n = 17)≤−2.5 (n = 20)
  • *

    Except where indicated otherwise, values are the number (%). Results were available for only 42 of the 54 patients who underwent bone densitometry during the study. Prescription medications for osteoporosis included hormone replacement therapy, raloxifene, etidronate, alendronate, pamidronate, and calcitonin. Column percentages total more than 100 since each patient may have been exposed to calcium and vitamin D as well as a prescription medication. All available lumbar and femoral T scores were examined. BMD = bone mineral density.

None2 (40)4 (24)1 (5)
Calcium and vitamin D1 (20)9 (53)8 (40)
Prescription medication, any3 (60)8 (47)19 (95)

DISCUSSION

We examined the rates of bone densitometry and medication use in patients with RA taking oral glucocorticoids treated at an academic rheumatology practice. As others have reported, the rate of bone densitometry was quite low (23%). The rate of prescription medication use was higher than in most previous studies (42%), but may still be suboptimal. Because we do not have data on bone density for most patients, we cannot determine the appropriate rate of treatment in the study population. However, we did not find a trend toward screening or treatment in persons taking higher glucocorticoid doses for longer periods, and good epidemiologic data suggest that the daily and cumulative doses are strong predictors of fracture risk (12). In secondary analyses, we tested multiple glucocorticoid dosage thresholds and found no relationship between dosage and osteoporosis management. Men and premenopausal women were less likely to undergo bone densitometry or receive prescription treatment. We were somewhat surprised to find that patients with at least 1 comorbid condition were less likely to receive a prescription medication for osteoporosis. Others have noted that patients with chronic medical conditions, such as diabetes, emphysema, and psychosis, are less likely to receive medical treatments for their arthritis (13); our findings reflect a variation on this theme. Patients with multiple chronic medical conditions, or their doctors, may not want to embark on preventive care for osteoporosis.

Previous studies that have explored factors associated with patterns of management for GIOP have focused on physician specialty. Rheumatologists have consistently been more thorough in obtaining bone densitometry and/or prescribing medications than other subspecialists (8, 9). Other studies, focused on non–glucocorticoid-induced osteoporosis, have found that correlates of not using bone densitometry include male physicians, physician's belief that calcium and vitamin D were adequate treatment, and physician's belief that bone densitometry did not aid in diagnosing osteoporosis (ref. 14, and Solomon DH, et al: unpublished observations).

Our findings that men and premenopausal women were less likely to receive bone densitometry or a prescription medication are not surprising. Little attention has been paid to the risk of non–glucocorticoid-induced osteoporosis in these groups, and there is a lack of consensus on appropriate screening and treatment; the findings reported here reflect the lack of consensus. Since patients with RA taking glucocorticoids are clearly at an elevated risk for osteoporosis and fractures (1), more focus needs to be placed on these issues, even among rheumatologists.

Our results must be interpreted within the limitations of the methodology. There may have been underreporting of bone densitometry if patients underwent testing outside of the study hospital and the treating rheumatologist did not mention the test in his/her notes. Also, medications prescribed by doctors outside of the study hospital may not have been recorded completely in the electronic medical record. We did include these data where they could be found in the medical records. Also, the medication use information, including glucocorticoid dosage and duration, must be interpreted with caution. We did not review pharmacy records, nor did we interview patients for their precise medication history. Physicians' notes were examined for medication data; however, there may have been glucocorticoid courses never noted in the medical record. Also, patients may not have been taking what their doctors listed. Use of calcium and vitamin D was almost definitely underreported. All of these limitations would have contributed to misclassification of predictor variables and outcomes.

We did not validate the diagnosis using standard ACR criteria (15), so some patients may not actually have had RA. Since this was not an epidemiologic study, but an investigation of the use of health care resources for patients treated for RA, it is more important to determine the diagnosis made by the treating physician (16). We also did not attempt to determine whether a specific patient's management was in strict accordance with ACR GIOP guidelines. This decision was made since the evidence for GIOP management is still evolving and clinicians may differ on their interpretation of different guidelines. Instead, we chose to focus on specific elements mentioned in other guidelines, such as bone densitometry and medication use.

These data came from the rheumatology practice at one academic medical center. Thus, our findings may not be widely generalizable. The study period was limited to the 12 months before and after the index visit. This limited sampling window meant that we did not necessarily examine a patient's care throughout the entire glucocorticoid course, but 2 years is a substantial time frame during which there should be some evidence of GIOP management. To estimate the rate of prior bone density testing, we reviewed the entire medical record of a random sample of 50 patients who did not have bone densitometry during the 2-year study period. Three patients (6%) were noted to have had a bone density scan 4–5 years before the index dates (data not shown). Finally, some of the bone densitometry tests and osteoporosis medications may have been ordered by nonrheumatologists. Using the electronic medical record, we cannot determine which physicians ordered tests or prescribed a medication.

These data add to the existing literature suggesting suboptimal management of GIOP. We examined patients with RA seen by rheumatologists practicing at an academic center and found low rates of bone densitometry and prescription medication use for osteoporosis. Postmenopausal status and male patient sex were predictors of not receiving bone densitometry or osteoporosis medications. This suggests that doctors did not believe that RA and oral glucocorticoid therapy placed these traditionally low-risk groups at high enough risk for screening and/or prescription treatment of osteoporosis.

Further studies are needed to clarify why management of GIOP is suboptimal and to test interventions for quality improvement. We are currently testing an educational program as well as a system of computer-based reminders. Rheumatologists are accustomed to monitoring the potential toxicities of DMARDs, but it is not clear why monitoring and treating osteoporosis in patients with inflammatory arthritis who take oral glucocorticoids has not been more quickly incorporated into routine practice. It may be that the rheumatologist's focus is on immediate toxicities of complex medication regimens that include multiple DMARDs; such attention may overshadow the potential for long-term toxicities associated with medications like glucocorticoids. It also might be that since many patients have taken oral glucocorticoids for long periods without any monitoring for osteoporosis, doctors and patients are resistant to initiating a program of osteoporosis prevention and treatment. Greater understanding of these issues will improve the quality of health care and the lives of patients.

Acknowledgements

We thank Dr. Robert Yood for his comments on an earlier draft of the manuscript.

Ancillary