Decline of Mean Initial Prednisone Dosage From 10.3 to 3.6 mg/day to Treat Rheumatoid Arthritis Between 1980 and 2004 in One Clinical Setting, With Long-Term Effectiveness of Dosages Less Than 5 mg/day




To analyze prednisone treatment from 1980–2004 in 308 patients with rheumatoid arthritis (RA), including 75 monitored over 4–8 years and 73 monitored over >8 years, for initial dose, long-term doses and effectiveness, and adverse events.


A database of all patients of a single rheumatologist included medications and Multidimensional Health Assessment Questionnaire (MDHAQ) scores at each visit. Proportions of patients whose initial prednisone dosages were >5, 5, or <5 mg/day were computed in 5-year periods: 1980–1984, 1985–1989, 1990–1994, 1995–1999, and 2000–2004. Mean changes in MDHAQ function, pain, and Routine Assessment of Patient Index Data 3 (RAPID3) scores were compared in patients treated with <5 versus ≥5 mg/day of prednisone; scores and adverse events were analyzed in quartiles by treatment duration of ≤1, 1.1–4, 4.1–8, and >8 years.


In the respective 5-year periods, the mean initial prednisone dosages were 10.3, 6.5, 5.1, 4.1, and 3.6 mg/day, with >5 mg/day in 49%, 16%, 7%, 7%, and 3% of patients, 5 mg/day in 51%, 80%, 70%, 26%, and 10% of patients, and <5 mg/day in 0%, 4%, 23%, 67%, and 86% of patients. Most patients received early concomitant methotrexate after 1990, and prednisone <5 mg/day was maintained indefinitely. Patients treated with prednisone ≥5 mg/day had poorer clinical status as baseline and followup. MDHAQ scores improved similarly in patients treated with <5 or ≥5 mg/day. Primary adverse events were skin thinning and bruising. New hypertension, diabetes mellitus, and cataracts occurred in <10% of all patients, and <13% of those treated longer than 8 years.


The data suggest that many patients with RA might be treated effectively with initial and long-term prednisone <5 mg/day, although further research and observational data are needed to characterize more fully effectiveness and safety.


Glucocorticoids were documented to provide dramatic clinical improvement in patients with rheumatoid arthritis (RA) in 1948 (1), as well as disease-modifying activity in slowing radiographic progression during the 1950s (2). However, severe adverse effects were seen in most patients treated with long-term glucocorticoids, generally in pharmacologic dosages of 20–60 mg/day of prednisone or prednisolone, the clinical practice at the time (3). Therefore, after the mid-1950s, systemic glucocorticoids generally were regarded as contraindicated for RA, since risks of therapy appeared to outweigh risks of disease.

A reassessment of glucocorticoids in RA was begun during the 1980s, based in part on recognition of severe long-term consequences of RA (4–8), and increased early use of methotrexate (9, 10), which appeared to allow reduced prednisone doses to be effective. An early open study (11), a nonblinded clinical trial (12), and more recent double-blind clinical trials (13–21) have recognized the clinical efficacy and relative safety of low-dose prednisone, confirmed in meta-analyses (22, 23). Nonetheless, glucocorticoids continue to be recommended in RA primarily as “bridging therapy” (12), while awaiting anticipated benefits of disease-modifying antirheumatic drugs (DMARDs) or for acute disease flares or life-threatening vasculitis, seen in recent European League Against Rheumatism recommendations (24).

Although textbooks and guidelines continue to recommend limited glucocorticoid use in RA, many RA patients are treated with prednisone or prednisolone over long periods in usual care clinical settings. For example, in the international Quantitative Standard Monitoring of Patients with Rheumatoid Arthritis (QUEST-RA) database of 4,363 RA patients seen in usual care at 48 clinical sites in 15 countries, 66% were taking glucocorticoids (25).

Prednisone >5 mg/day over long periods results in increased levels of adverse events in some patients (6–8), including increased mortality rates (5–8), although no significant differences in adverse events from placebo are reported in clinical trials involving 7.5 or 10 mg/day over 2 years or less (13, 16, 18–20). A prednisone dosage of 5 mg/day does not induce suppression of the hypothalamic–pituitary–adrenal axis in most patients, as seen with higher doses (26–29). Furthermore, 5 mg/day (or less) is much less likely to result in long-term adverse event effects, other than thinning of skin and bruising (30). The efficacy of 5 mg/day (12, 17), and even 3 mg/day (21), with few adverse events has been reported in RA, but long-term observations of patients who took these low doses over 4 years or longer are not reported.

Over a 25-year period from 1980 to 2004, the corresponding author (TP) developed a clinical practice of prescribing low initial prednisone dosages of <5 mg/day in most patients with RA, maintained over long periods, which was described in a review article (31). This therapy was accompanied by increasing use of methotrexate for most patients (10), with a goal to control disease activity as effectively as possible, expressed as “no evidence of disease” (32, 33), although not necessarily as rigorously as current treat-to-target recommendations (34). Efficacy and adverse events associated with prednisone could be assessed through completion of a Multidimensional Health Assessment Questionnaire (MDHAQ) by each patient at each visit, including 75 patients who took prednisone for 4 to 8 years and 73 patients who took prednisone for longer than 8 years, as included in this study.

Significance & Innovations

  • Prednisone at dosages of <5 mg/day (3 mg/day in most patients) is effective to treat rheumatoid arthritis (RA) over long periods with minimal adverse events.

  • It is not possible to conduct randomized controlled clinical trials over 4 or more years in patients with RA, but it is possible to monitor patients in usual care indefinitely by having each patient complete a self-report questionnaire at each visit.

  • Higher doses of prednisone in rheumatology appear based in large part on practices from many years ago of treating patients with prednisone only, before incorporating disease-modifying antirheumatic drugs or immunosuppressive or cytotoxic therapies into routine management of inflammatory rheumatic diseases.

  • Many (if not most) patients with rheumatic and other inflammatory diseases may be treated with higher doses of prednisone than needed, which may result in unnecessary adverse events, without greater short-term or long-term effectiveness.



Visits of all patients to the first author (TP) at a weekly academic rheumatology setting from 1980–2006 were recorded in a database. Approximately 800 patients with RA were seen over this period (approximately 32 new RA patients per year over 25 years). Many of these patients were taking prednisone at their first visit, already prescribed by a general physician or another rheumatologist. Prednisone was initiated in this setting by TP in 308 RA patients, almost all at their first visit. Usually, an additional DMARD was included at the first or second visit, generally methotrexate after 1990 (10).

Prednisone administration.

No formal guidelines were in place for treatment with prednisone or for the dose in this setting over the period of 1980–2004. During this period, treatment of RA patients with methotrexate in this setting was increased from 10% to 78% (10). Most patients who did not take methotrexate took other DMARDs in addition to low-dose prednisone. Completion of the MDHAQ by each patient at each visit, and maintenance in a database for analysis, were approved by the Institutional Review Board of Vanderbilt University and consented to by patients.

Clinical assessment with MDHAQ.

All patients completed a version of the MDHAQ at each visit, which was entered into the database along with laboratory test results and medications. The standard MDHAQ at this time (35) included the 3 self-report measures from the RA Core Data Set (36), i.e., physical function, pain, and patient global estimate of status, each scored from 0–10 and compiled into an index of only patient self-report measures, the Routine Assessment of Patient Index Data 3 (RAPID3) (37, 38). Physical function was scored from 0–3, as used traditionally for the HAQ, from 1980 until approximately 2006, with the introduction of RAPID3 scores, when it was changed to 0–10. The MDHAQ used in this setting also queried patients for development of comorbidities and adverse effects of medications.

The MDHAQ completed by patients between 1980 and 1996 included scores for physical function and pain, but not for patient global estimate of status, which was added in 1996. Therefore, a standard RAPID3 score could not be compiled for these patients, and an estimate of RAPID3 (RAPID3-EST) was developed that included a score for physical function (on a scale of 0–10) plus pain score (also on a scale of 0–10), multiplied by 2. The RAPID3-EST includes 2 of the 3 RAPID3 measures, and is correlated with RAPID3 scores at ρ = 0.85 in patients in this database for whom RAPID3 scores also were available (data not shown).

Statistical analyses.

The initial dose of prednisone was analyzed in each patient. The mean and median doses (and the proportion of patients whose initial dosage was >5 mg/day, 5 mg/day, and <5 mg/day) were analyzed in 5-year periods: 1980–1984, 1985–1989, 1990–1994, 1995–1999, and 2000–2004. The mean and median MDHAQ scores for physical function, pain, RAPID3, and RAPID3-EST were compiled for each 5-year period according to the 3 categories of initial prednisone dosage; patients who took 5 mg/day or >5 mg/day were pooled, since few patients after 1985 took >5 mg/day.

The efficacy of prednisone was analyzed in patient groups, classified according to initial dosages of <5 mg/day versus ≥5 mg/day. Changes in MDHAQ scores for physical function, pain, RAPID3-EST, and RAPID3, which includes patient global estimate, when available, were analyzed using 2 strategies: 1) change over 12 months after initiation of prednisone therapy in the 5-year periods, and 2) change from the initial visit to the last visit in the database, available for 290 of the 308 patients in 4 quartiles according to available data concerning the duration of prednisone usage: ≤1 year, 1.1–4 years, 4.1–8 years, and >8 years.

Possible adverse events were compiled according to 1) differences between weight at baseline and the last observed followup visit in the 4 quartiles of duration of prednisone therapy, and 2) development of patient-reported cataracts, hypertension, and/or diabetes mellitus after initiation of prednisone reported on the MDHAQ.


Initial prednisone dose over 25 years.

In the 308 patients whose prednisone was initiated by TP, the mean initial prednisone dosage was 10.3 mg/day in 1980–1984, compared to 6.5 mg/day in 1985–1989, 5.1 mg/day in 1990–1994, 4.1 mg/day in 1995–1999, and 3.6 mg/day in 2000–2004 (Figure 1A). The median initial dosage was 5 mg/day in the first three 5-year periods (from 1980–1994) and 3 mg/day in the last two 5-year periods (from 1995–2004). The proportion of patients whose initial dosage was <5 mg/day was 0% in 1980–1984, 4% in 1985–1989, 23% in 1990–1994, 67% in 1995–1999, and 86% in 2000–2004 compared to 51%, 80%, 70%, 26%, and 10% treated initially with 5 mg/day and 49%, 16%, 7%, 7%, and 3% treated initially with >5 mg/day, respectively (Table 1 and Figure 1B).

Figure 1.

A, Mean and median initial prednisone dosages in 308 patients with rheumatoid arthritis (RA) seen between 1980 and 2004, computed in 5-year periods: 1980–1984, 1985–1989, 1990–1994, 1995–1999, and 2000–2004. B, Percentage of patients with RA whose initial prednisone dosage was <5, 5, or >5 mg/day in the respective 5-year periods. C, Percentage of patients with RA who were treated with methotrexate in the respective 5-year periods (data from ref.10).

Table 1. Initial prednisone dosage in 308 patients with rheumatoid arthritis seen from 1980 through 2004
Year first seenNInitial dosage, mean (median) mg/dayPercentage of patients taking initial dosage
<5 mg/day5 mg/day>5 mg/day
1980–19843710.3 (5)05149
1985–1989746.5 (5)48016
1990–1994775.1 (5)23707
1995–1999614.1 (3)67267
2000–2004593.6 (3)86103

Prednisone usually was prescribed at the first visit; another DMARD was initiated in most patients at the first or second visit, primarily gold salts prior to 1985 and methotrexate thereafter, consistent with a policy to minimize inflammation (32). The proportion of patients treated with methotrexate was 10% in 1980–1984, 26% in 1985–1989, 57% in 1990–1994, 71% in 1995–1999, and 78% in 2000–2004 (data from ref.10) (Figure 1C). Patients who were already taking prednisone (initiated by a family practitioner, internist, or other rheumatologist) at the first visit to TP are not included in this study.

MDHAQ scores for function, pain, and RAPID3 at baseline and 1 year later, according to prednisone dose.

Patients who were treated initially with <5 mg/day of prednisone had baseline scores for physical function of 2.4 on a scale of 0–10 (0.8 on a 0–3 scale) compared to 3.5 (1.2 on a 0–3 scale) in patients treated initially with ≥5 mg/day of prednisone (Figure 2). Mean baseline pain score was 5.2 (on a visual analog scale of 0–10) in the <5 mg/day group versus 6.3 in the ≥5 mg/day group (Figure 2). Mean RAPID3 scores were 13.2 versus 17.3 (on a scale of 0–30), and mean RAPID3-EST scores (on a scale of 0–30) were 12.4 versus 15.9 (Figure 2). Similar patterns were seen over each 5-year period (31).

Figure 2.

Scores at baseline and 12-month followup for A, function, B, pain, C, estimate of Routine Assessment of Patient Index Data 3 (RAPID3-EST; see Methods), and D, RAPID3, in 290 patients with rheumatoid arthritis, according to initial prednisone dosages of <5 versus ≥5 mg/day. MDHAQ = Multidimensional Health Assessment Questionnaire.

These data reflect clinical practice to treat patients who had greater disease severity with higher doses of prednisone, consistent with “confounding by indication,” as seen in most observational clinical research studies. Improvement in MDHAQ function, pain, RAPID3-EST, and RAPID3 scores 12 months after prednisone initiation was 34%, 37%, 37%, and 37%, respectively, in patients whose initial dosage was <5 mg/day versus 40%, 37%, 38%, and 31%, respectively, in patients treated with ≥5 mg/day (Figure 2).

Changes in MDHAQ scores from 1980–2004 in quartiles according to duration of prednisone treatment.

Data were available in 290 of 308 patients (18 patients had only a single initial visit) from the first to last visit in the database, and were analyzed in quartiles according to length of followup (Table 2). Mean prednisone doses were largely unchanged in patients monitored over ≤1, 1.1–4, 4.1–8, and >8 years, although patients were given an option to taper their dose by 1 mg every month or two, and some patients reduced the initial dosage from 3 mg/day to 2 or 1 mg/day, or raised the dosage to 4 mg/day. Mean scores recorded at each visit for physical function and pain were improved from baseline, other than identical scores for physical function (for which scores tend to rise with aging) after 8 years (Table 2).

Table 2. Baseline and last available MDHAQ scores for function, pain, and RAPID3-EST, and new adverse events reported by patients, according to length of prednisone treatment, in 290 patients with rheumatoid arthritis*
 Duration of prednisone use, yearsTotal (n = 290)
0.1–1.0 (n = 72)1.1–4.0 (n = 70)4.1–8.0 (n = 75)>8 (n = 73)
  • *

    Values are the mean ± SD unless otherwise indicated. MDHAQ = Multidimensional Health Assessment Questionnaire; RAPID3-EST = estimate of Routine Assessment of Patient Index Data 3.

Effectiveness data     
 Baseline function (0–10 scale)3.2 ± 2.02.8 ± 1.63.0 ± 2.13.3 ± 1.93.1 ± 2.0
 Last visit function (0–10 scale)2.7 ± 2.32.3 ± 1.82.7 ± 2.33.3 ± 2.42.8 ± 2.2
 Baseline pain (0–10 scale)5.9 ± 2.55.8 ± 2.66.1 ± 2.45.9 ± 2.65.9 ± 2.6
 Last visit pain (0–10 scale)4.0 ± 3.04.0 ± 2.74.1 ± 3.04.5 ± 3.24.2 ± 3.0
 Baseline RAPID3-EST (0–30 scale)15.1 ± 6.414.4 ± 5.915.0 ± 6.614.9 ± 6.614.8 ± 6.3
 Last visit RAPID3-EST (0–30 scale)10.6 ± 7.810.3 ± 6.710.9 ± 7.612.3 ± 7.911.0 ± 7.5
Possible adverse events     
 Baseline weight, lbs163.3 ± 39.1167.3 ± 40.0161.5 ± 36.7163.5 ± 38.8165.7 ± 40.7
 Last visit weight, lbs169.8 ± 36.4165.6 ± 38.6160.0 ± 42.4161.4 ± 41.2164.2 ± 39.7
 Hypertension, no. (%)1 (1.4)4 (5.7)5 (6.7)9 (12.3)19 (6.6)
 Cataracts, no. (%)01 (1.4)08 (11.0)9 (3.1)
 Diabetes mellitus, no. (%)003 (4.0)5 (6.9)8 (2.8)

Adverse events from 1980–2004 in quartiles according to duration of prednisone treatment.

Mean weight gain was 6 pounds (2.7 kg) over 1 year in patients monitored for ≤1 year, but weight was similar to baseline in patients monitored for longer than 1 year (Table 2). Among patients monitored for >8 years, new incidence of hypertension was seen in 12%, cataracts in 11%, and diabetes mellitus in 7%. Overall, <7% of patients developed hypertension during observation while receiving low-dose prednisone, <4% developed cataracts, and <3% developed diabetes mellitus (Table 2).


The data indicate a decline in mean initial prednisone dosage in 308 patients with RA treated by a single rheumatologist between 1980 and 2004 from 10.3 mg/day in 1980–1984 to 3.6 mg/day in 2000–2004 (Table 1). The proportion of patients who were taking methotrexate was increased from 10% in 1985 to 78% in 2000 (39) (Figure 1), which likely explains in part the reduction of the prednisone dose over this period. Prednisone at initial and long-term dosages of <5 mg/day with indefinite continuation appeared effective for most patients with RA, with few adverse effects.

The reported findings are consistent with the report of Da Silva et al that “adverse effects associated with [low-dose prednisone] are modest, and often not statistically different from those of placebo” (40), based on randomized trial data. Clinical trials provide optimal information concerning the efficacy of a therapy compared to a placebo or control treatment. However, it is pragmatically impossible to conduct a clinical trial of low-dose prednisone over 4 years or longer (41). Analysis of patient care with quantitative data concerning clinical status and adverse events provides “evidence-based medicine” not available from randomized clinical trial data (42, 43), including data on 75 patients over 4–8 years and on 73 patients over >8 years in this study, with information concerning adverse events that may develop over years, rather than months (3, 6–8).

Adverse effects were primarily bruising and skin thinning. Fewer than 10% of patients developed diabetes mellitus, hypertension, and/or cataracts, including fewer than 13% of 73 patients treated longer than 8 years (Table 2). The findings are consistent with those in a German database of self-report of RA patients indicating few adverse events (30). It is not possible to identify precisely expected levels of comorbidities in RA in the total absence of glucocorticoids over 5 years or longer, and the incidence and prevalence of complications of glucocorticoid therapy were not ascertained systematically as in a randomized trial. However, patients generally report development of hypertension, diabetes mellitus, and cataracts on the MDHAQ, and it is not likely that the incidence of these adverse events was substantially higher than what was found and reported. Even if somewhat higher than expected, the risk/benefit ratio appears justified to control inflammation, prevent joint damage, maintain physical function, and control pain.

Patients treated with ≥5 versus <5 mg/day of prednisone had more severe clinical status, as might be expected in clinical practice, consistent with “confounding by indication.” These patients also had poorer status at 12-month followup, consistent with evidence that dosages ≥5 mg/day led to no greater improvement than <5 mg/day. No formal criteria were used to determine dose. Most patients attempted to taper the dose to levels as low as possible, often on the advice of other physicians, friends, relatives, or pharmacists, many several times. Although discontinuation was seen in a few patients, most were titrated to the lowest dose at which the patient detected a treatment effect that she/he wished to continue. The acceptable dosage was always ≤5 mg/day (the usual dosage in >80% was 3 mg/day), and was continued indefinitely. Three patients appeared to “require” long-term dosages of >5 mg/day; these patients had been treated previously at other sites for long periods with >10 mg/day.

The patients analyzed in this report had a general improvement in status over the 25 years (10, 39), with significantly lower swollen joint counts, MDHAQ function, and radiographic scores in 2000 versus 1985 (39). Improved patient status may have resulted from many changes over the study period, including earlier treatment with a goal of “no evidence of disease” (32), increase in concomitant use of methotrexate (10), and milder natural history of RA (44). Nonetheless, long-term low-dose prednisone in most patients may have contributed to better patient status, with rates of adverse events acceptable to the patients and their physician.

Several important limitations are seen in this study. First, as emphasized, the study is not a randomized controlled trial, which would isolate prednisone dose as the only variable for which there was variation in the study patients, while other therapies remain constant. It would be optimal to compare results with 3 mg/day of prednisone versus 5 or 10 mg/day, as well as no prednisone. However, a dosage of 10 mg/day is known to be associated with long-term adverse events (3, 6–8), whereas adverse events other than skin thinning and bruising were not greater than expected among patients treated with <5 mg/day, as reported herein and elsewhere (26). Furthermore, it is not possible to maintain randomization in a symptomatic condition such as RA over 4 years. Although improved patient status may be explained in part or entirely by variables other than prednisone, most patients elected to continue low dosages of 1–4 mg/day when given opportunities to taper and discontinue prednisone. Patients with RA generally discontinue therapies perceived to be of no value over 6–12 months, and do not continue over >4 years. For example, 50% of 1,077 courses of injectable gold salts, penicillamine, hydroxychloroquine, and azathioprine were discontinued after 2 years in 532 patients with RA (45); more recently, 50% of 975 courses of biologic agents were discontinued after 2 years in a cohort of 770 patients (46). Continuation of prednisone <5 mg/day over >4 years suggests perceived value to patients, particularly in light of extensive advice to discontinue this medication.

A second limitation is that the data involve only a single practitioner. Definitive evidence concerning effectiveness and safety of low-dose prednisone <5 mg/day (or any therapy) cannot be derived from observational data at a single site, and is best documented in a multicenter randomized controlled clinical trial or observational study, as noted above. It is hoped that this report might stimulate such research.

Third, the single practitioner undoubtedly has biases in estimating patient improvement. However, all estimates of efficacy reported here were by MDHAQ patient self-report scores, which are considerably less likely than joint counts to improve with placebo treatment in clinical trials (47).

Fourth, the data concerning long-term effectiveness are derived from an MDHAQ, a patient self-report questionnaire. However, self-report allowed long-term data to be collected at low cost in the infrastructure of usual care (48), since the patient does most of the work. Self-report RAPID3 scores are as efficient as the Disease Activity Score with 28-joint count (DAS28) and the Clinical Disease Activity Index (CDAI) to distinguish active from control treatments in clinical trials involving methotrexate, leflunomide, anakinra, adalimumab, and abatacept, and are correlated significantly with DAS28 and CDAI scores in clinical trials and usual clinical care (37, 38, 49). Therefore, self-report appears to provide reasonable measures of effectiveness over long periods.

Fifth, ascertainment of adverse events was also only through patient self-report, without systematic query concerning hypertension, diabetes mellitus, and cataracts, as in clinical trials. However, other reports indicate similar frequencies through self-report (30), and MDHAQ self-report has been quite effective to ascertain major conditions such as hypertension, diabetes mellitus, and cataracts. Even if adverse events may be underestimated through self-report, the reported levels of hypertension, diabetes mellitus, and cataracts may present an acceptable risk/benefit ratio for enhanced control of RA, as noted above, based on an informed decision between the doctor and the individual patient.

The data included here suggest that current recommendations to regard glucocorticoids as temporary “bridging” therapy in RA, as well as the current practice of initiating prednisone at dosages of 5 mg/day or higher for treatment of RA, may be reassessed. Dosages of 10 mg/day and higher may be associated acutely with polyphagia, sleeplessness, and mental status changes in some patients (3, 33), and with recognized long-term physiologic adverse effects (3, 6–8), including premature mortality (5–8). High doses used in the 1950s reflect glucocorticoids as the only available therapy for inflammatory rheumatic diseases at the time (3). At present, methotrexate and other DMARDs, cytotoxic drugs, or biologic agents are used early in most patients, which may allow efficacy of lower glucocorticoid doses than in previous decades. Further clinical trials and further complementary observational studies over long periods (such as that presented here) appear desirable to advance knowledge toward optimal use of low-dose prednisone for patients with RA.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Pincus had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Pincus, Sokka, Castrejón, Cutolo.

Acquisition of data. Pincus.

Analysis and interpretation of data. Pincus, Sokka, Castrejón, Cutolo.