- Top of page
- SUBJECTS AND METHODS
Glucocorticoids are estimated to be used long-term by 0.5–1% of the general population and up to 2.5% of older adults (1, 2). Despite the established role of glucocorticoids in controlling short-term inflammation, and despite emerging evidence supporting a disease-modifying role in rheumatoid arthritis (3–6), concern for adverse events (AEs) associated with glucocorticoids often limits their use. For many individuals, potential AEs such as acne, weight gain, and sleep/mood disturbance may be mild. For others, perceived AEs such as cataracts or glucocorticoid-induced osteoporosis with subsequent fracture may be severe and result in substantial morbidity and mortality (7–15). Estimates regarding the prevalence and severity of many glucocorticoid-associated AEs are largely unknown, especially those less likely to require hospitalization or specific medications for prevention or treatment. Of note, the relationship between these common AEs and various glucocorticoid doses is also unclear, and the safety of long-term, low-dose prednisone (e.g., daily dosage ≤7.5 mg) remains controversial.
We hypothesized that the prevalence of AEs associated with even low-dose glucocorticoid use would be high and would be dose and duration dependent. We therefore conducted a population-based survey of glucocorticoid users to obtain prevalence estimates of glucocorticoid-associated AEs. We specifically focused on these endpoints among individuals who had prolonged use of an average daily dose of ≤7.5 mg of prednisone.
- Top of page
- SUBJECTS AND METHODS
A total of 2,446 (38%) managed care enrollees that met the definition for long-term glucocorticoid use returned the survey. Their characteristics are described in Table 1. Individuals were middle aged, predominantly women, and received prescriptions for an average of 16 mg/day of prednisone. A majority were new users rather than prevalent glucocorticoid users. Persons with rheumatoid arthritis were prescribed a mean ± SD dosage of 12 ± 8 mg/day of prednisone, in contrast to persons with inflammatory bowel disease who received a mean ± SD prednisone dosage of 31 ± 16 mg/day (data not shown). The mean daily prednisone doses for the other diseases, including systemic lupus erythematosus, chronic obstructive pulmonary disease, and asthma, fell between these 2 extremes. Hypertension and rheumatoid arthritis were among the most common comorbid diagnoses. More than 91% of respondents indicated that they had ever taken glucocorticoids, 7% said that they had never taken them, and 2% did not answer (data not shown). Sixty-five percent of the ever users reported current use at the time they completed the survey. Among those who reported ever using glucocorticoids, a majority (68%) reported discussing potential glucocorticoid-related AEs with their doctor prior to use.
Table 1. Characteristics of long-term glucocorticoid (GC) users responding to survey (n = 2,167)*
|Age, mean ± SD years||53 ± 14|
|Female sex||1,538 (71)|
|Length of enrollment in the health plan, mean ± SD months||27 ± 13|
| White||1,628 (79)|
| African American||266 (13)|
| Other (e.g., Asian, Hispanic)||160 (8)|
|High school education or less†||792 (38)|
|Annual household income <$20,000†||236 (12)|
|GC use|| |
| Average prednisone equivalent dose, mean ± SD mg/day||16 ± 14|
| Highest dose prednisone received, mean ± SD mg/day||28 ± 30|
| Cumulative prednisone equivalent dose, mean ± SD gm||3.6 ± 3.1|
| Duration of prednisone use, mean ± SD days‡||284 ± 177|
| New GC user (≥90 days without prior GC prescription)||1,329 (61)|
|Medical comorbidities|| |
| Number of comorbid conditions, mean ± SD||7 ± 3|
| Medical diagnoses from claims data|| |
| Diabetes||342 (16)|
| Hypertension||962 (44)|
| Chronic renal failure||171 (8)|
| Congestive heart failure||170 (8)|
| Diagnoses associated with GC use|| |
| Rheumatoid arthritis||892 (41)|
| Chronic obstructive pulmonary disease||290 (13)|
| Systemic lupus erythematosus||264 (12)|
| Asthma/reactive airway disease||266 (12)|
| Inflammatory bowel disease||167 (8)|
Compared with survey responders, the 4,071 survey nonresponders were younger (mean age 48 versus 53 years) and were more likely to be male (36% versus 30%). Nonresponders' duration of enrollment in the health plan, mean prednisone dose, and average number of comorbid conditions were similar to survey responders (data not shown).
Relationship between glucocorticoid dose and AEs.
The prevalence and self-reported severity of AEs experienced, stratified by quartile of cumulative glucocorticoid use, are shown in Figure 1. The AE with the greatest self-reported prevalence was weight gain, which was experienced by almost 80% of subjects in the highest quartile of glucocorticoid use. Skin bruising/thinning and sleep disturbance were the next most commonly reported AEs. Cataracts (15% overall) and fractures (12% overall) were reported less frequently. Only 10% of subjects reported that they had not experienced any of the AEs concurrent with glucocorticoid use. We examined claims data for fractures separately during the 30-month observation period. Ten percent of the cohort had ≥1 medical service claims for a fracture during the 2.5-year period of observation.
Figure 1. Prevalence of adverse events associated with long-term glucocorticoid use stratified by cumulative glucocorticoid dosage (n = 2,167 subjects after excluding individuals who did not confirm that they had actually taken glucocorticoids). Quartiles of cumulative prednisone-equivalent glucocorticoid dosage: Q1 = <1.7 gm (e.g., 10 mg/day of prednisone for 6 months); Q2 = 1.7–2.8 gm (e.g., 10 mg/day of prednisone for 9 months); Q3 = 2.9–4.7 gm (e.g., 10 mg/day of prednisone for 12 months); Q4 = >4.7 gm (e.g., 10 mg/day of prednisone for 18 months). * P < 0.0001 (Cochran-Armitage trend test). † P < 0.01 (Cochran-Armitage trend test). ‡ P < 0.001 (Cochran-Armitage trend test).
Download figure to PowerPoint
The adjusted relative risks of AEs associated with cumulative glucocorticoid dose are described in Table 2. A strong dose-response relationship was observed between increasing quartiles of glucocorticoid use and all of the AEs examined. To assess the validity of our self-reported fracture models, we separately modeled fractures using medical claims data. Even after multivariable adjustment, a positive and significant association with increasing cumulative glucocorticoid dose was observed (data not shown).
Table 2. Relationship between cumulative prednisone-equivalent dose and self-reported adverse events*
|Characteristics/quartile†||Adjusted OR||95% CI|
|Sleep disturbance (n = 2,146)‡|| || |
|Acne (n = 2,040)§|| || |
|Skin bruising or thinning (n = 2,144)¶|| || |
|Weight gain (n = 2,040)#|| || |
|Mood problems (n = 2,025)**|| || |
|High blood sugar (among nondiabetics; n = 1,719)††|| || |
|Cataracts (n = 1,869)‡‡|| || |
|Fracture (n = 1,899)§§|| || |
Among users of >7.5 mg/day of prednisone, glucocorticoid dose and duration of use were strongly associated with all AEs (data not shown). The adjusted odds ratios of increasing duration of glucocorticoid use among individuals prescribed ≤7.5 mg/day of prednisone are presented in Table 3. Acne, skin bruising, weight gain, and cataracts were significantly associated with longer durations of low-dose glucocorticoid use. In contrast, increasing daily dose (within the 0–7.5 mg/day range) was more strongly associated with sleep disturbance and fractures than was increased duration of use. The Spearman correlation coefficient between cumulative glucocorticoid dose and average daily dose was 0.55.
Table 3. Relationship between glucocorticoid dose and duration and adverse events among low-dose (prednisone ≤7.5 mg/day) glucocorticoid users (n = 670)*
|Characteristics||90-day increase in duration of use||95% CI||Average daily dose (per 1-mg/day increase within 0–7.5 mg/day range)||95% CI|
|Skin bruising or thinning§||1.17||1.08–1.26||NS||NS|
|High blood sugar (among nondiabetics)**||NS||NS||NS||NS|
- Top of page
- SUBJECTS AND METHODS
We found a high prevalence of self-reported AEs associated with glucocorticoid use among individuals treated long-term for a variety of conditions. Almost all respondents reported at least 1 AE that was temporally associated with concurrent glucocorticoid use, and more than half reported that at least 1 AE was very bothersome. Serious AEs potentially attributable to glucocorticoid exposure were also common, with 15% of the cohort self reporting cataracts and 12% self reporting ≥1 fractures. Increasing cumulative glucocorticoid dose had a strong positive association with serious AEs typically attributed to glucocorticoid use, even after multivariable adjustment for demographic and disease comorbidities that were also associated with these outcomes.
Long-term observational studies of patients receiving commonly prescribed doses of glucocorticoids have reported high rates of glucocorticoid-associated AEs. A study of 120 patients with giant cell arteritis (mean age at diagnosis 75 years) followed for a median of 10 years found that 86% of patients experienced ≥1 AEs (18); 58% of the patients experienced ≥2 serious AEs, the most common of which were cataracts (41%) and fractures (38%). In a similarly designed study of 232 patients with polymyalgia rheumatica (19) (mean age at diagnosis 73 years) followed for an average of 8 years, 65% of individuals treated with glucocorticoids experienced at least 1 AE, the most common of which were vertebral fracture (18%) and cataracts (36%). Cumulative glucocorticoid dose had a strong association with AEs, particularly after the dosage reached 1.8 gm, which was the approximate cutpoint between the first and second quartiles in our analyses. Our results are similar to these studies because 90% of our subjects reported at least 1 AE while receiving glucocorticoids. Rates of self-reported fracture (12%) and cataract (15%) in our study were high, although lower than these 2 reports, likely reflecting a younger population, diverse indications for glucocorticoid treatment, and lower cumulative glucocorticoid exposure.
More recent interest has focused on the clinical efficacy and potential disease-modifying properties of long-term low-dose prednisone (≤7.5 mg/day), particularly for patients with arthritis. In small clinical trials of patients with rheumatoid arthritis randomized to low glucocorticoid doses (prednisolone 7.5–10 mg daily), vertebral fractures, weight gain, and hyperglycemia were observed among the glucocorticoid-treated patients, although incidence rates of these AEs were low (3, 4). These and other trials with aggressive glucocorticoid tapering regimens (5) have demonstrated that low-dose or rapidly tapered glucocorticoids may result in minimal short-term glucocorticoid-associated toxicity. The strengths of these randomized, prospective, placebo-controlled trials are their blinded outcomes assessment; lack of confounding by disease severity; and prohibition of even short-term, high-dose glucocorticoid exposure. Factors that may limit their generalizability include selection of individuals with few comorbid illnesses and insufficient duration of followup to detect many of the AEs that develop and/or progress over time.
In contrast to these clinical trial findings, an observational study with a longer followup period compared patients with rheumatoid arthritis treated with a mean of 7 mg/day of near-continuous prednisone therapy for 4.9 years with similar patients with rheumatoid arthritis not treated with glucocorticoids (12). In the glucocorticoid-treated group, 82% experienced at least 1 AE compared with 24% in the control group. Fractures (19%) and cataracts (15%) were the most common serious AEs. We also found several AEs that were significantly associated with longer durations and small increments in glucocorticoid dose, even among ≤7.5 mg/day prednisone users (Table 3), and we demonstrated a dose-dependent increase in fracture risk in this group that is concordant with other studies of low-dose glucocorticoid users (10, 20). Moreover, although the 892 individuals with rheumatoid arthritis in our study received the lowest mean daily dosage of prednisone (12 mg/day) compared with those with other diseases that we examined, only 40% received a mean of ≤7.5 mg/day over 24 months. These data suggest that long-term use of low-dose prednisone in a large, diverse managed care population is typically administered only to a minority of individuals with rheumatoid arthritis.
Our study has several strengths. The large, population-based design coupled with the claims data and pharmacy linkage permitted us to analytically account for a variety of comorbid illnesses that were ascertained using both self-reported and administrative sources. Pharmacy data, rather than self report, were used to determine glucocorticoid use patterns and thus reduce potential recall bias. We were able to capture self-reported AEs that are often not assessed in observational studies that utilize only administrative data or even medical chart review. Our study captured the diverse diseases treated with glucocorticoids and included use of intermittent glucocorticoid prescriptions, allowing better generalizability than other studies that have focused on populations treated with glucocorticoids for only 1 condition or with only continuous therapy. Finally, we were able to examine demographic covariates such as ethnicity and education that are rarely available in medical record or claims-based data sources.
One limitation of our work is that AEs were self reported, and confirmation by a physician was not available. Although self report may be the only reasonable source of information for certain subjective AEs such as mood and sleep disturbance, acne, and skin changes associated with glucocorticoid use, it may not be optimal for other outcomes such as fracture. However, the positive predictive value of self-reported fracture has been reported to be >80% at any fracture site and often in excess of 95% for osteoporotic fracture sites (21–24). Because of uncertainty as to whether a medical claim for fracture represented a new (incident) fracture or followup for a prevalent fracture, we censored fracture data after a single event and therefore likely underestimated fracture incidence among individuals who experienced >1 new fracture. Vertebral fractures may be asymptomatic and may not be identified in as many as two-thirds of patients (25, 26), which provides an additional reason for our fracture estimates to be conservative. We were able to compare our results based on self-reported outcomes with fracture outcomes from administrative data and achieved similar results. Compared with fractures, less research has been conducted to validate self report of cataracts, but the positive predictive value of self-reported cataracts has been estimated to be ∼83% (95% confidence interval 79–88%) (27).
Another limitation of our work is that the self-reported outcome data are cross-sectional, and determining the temporal sequence of AEs to implicate a causal relationship with glucocorticoid exposure is problematic. For this reason, we intentionally excluded persons with medically recognized diabetes from our high blood sugar analysis (16% of the cohort). This conservative strategy underestimates the proportion of those who developed incident diabetes as a consequence of glucocorticoid therapy. Despite the similarity of survey respondents and nonrespondents in the distribution of comorbid conditions and mean daily prednisone dose (ascertained using administrative data), possible response bias may have influenced our findings if individuals that experienced glucocorticoid-associated AEs were more likely to respond to the survey. Additionally, glucocorticoids have been associated with a variety of other AEs (e.g., infection, hypertension) that we did not assess.
Finally, we were not able to compare our prevalence of self-reported AEs with that of a control group with similar diseases of equal severity that did not receive glucocorticoids. Confounding by disease severity is a concern in observational studies where “sicker” patients have increased exposure to a drug of interest (28). We adjusted for general comorbidity using the number of unique diseases for which the individual received medical care in the prior 30 months. However, because of the diverse nature of the diseases for which individuals were prescribed glucocorticoids, we were not able to adjust for all factors indicative of more severe disease (e.g., rheumatoid nodules and erosions for patients with rheumatoid arthritis); therefore, residual confounding by disease indication or severity is possible. However, we demonstrated a strong dose-response relationship between rates and severity of AEs and increasing doses of cumulative and daily oral glucocorticoids, even after adjusting for a variety of demographic factors and comorbidities.
The prevalence of overall and very bothersome AEs associated with long-term glucocorticoid use was high, even among users of lower doses, and this observation may be underappreciated by clinicians. Because patients may be concerned about the potential AEs of glucocorticoids prior to use (29), physicians need to be well apprised of their risks. A majority of individuals in our cohort reported discussing potential glucocorticoid-associated AEs with their physician prior to initiating therapy. This observation highlights the opportunity for physicians to discuss with patients the possible AEs of glucocorticoids in the context of their expected benefits.
Although the best way to reduce glucocorticoid-associated AEs is to limit exposure and dose, for some individuals the expected benefits will outweigh the possible risks. For example, among patients with early rheumatoid arthritis, combination therapy that includes glucocorticoids may have a sustained, disease-modifying effect (30). However, minimizing potentially avoidable AEs such as glucocorticoid-induced osteoporosis, for which there are efficacious but underutilized therapies (31–34), must be a shared responsibility among physicians and patients. Despite advances in chronic disease management, it is unlikely that the long-term use of glucocorticoids will disappear any time soon. As such, heightened attention to common glucocorticoid-associated AEs that bother patients, even among users of lower doses, is warranted.