Glucocorticoids have been in use for almost 60 years. They are powerful and cost-effective drugs with strong anti-inflammatory and immunomodulatory effects that are used to treat rheumatic and other diseases.1^–3 Their therapeutic use has increased continuously. In 2002, it was reported that up to approximately 10 million new prescriptions were written just for oral glucocorticoids each year in the USA and that the total market size has reached approximately US$10 billion per year.4 5 The introduction of more targeted therapies may have weakened this development in recent years, and there are arguments against the use of glucocorticoids based mainly on the concern of toxicity. However, this concern sometimes originates in observations of adverse effects in patients using high doses of glucocorticoids.6 Current evidence on the actual risk–benefit ratio of glucocorticoids is incomplete and/or inconsistent because only limited data defining the “threshold” dose for particular adverse events are available, and toxicity reports often cover a heterogeneous group of glucocorticoid-treated diseases.6^–8 One reason for this situation is that glucocorticoids were introduced into clinical medicine long ago. Therefore, we do not have any data comparable with the modern US Food and Drug Administration/European Medicines Agency approval data.5 8 Moreover, there are several different glucocorticoid drugs with different relative drug potencies on the market, making the collection of homogenous data and direct comparisons complicated.

With regard to adverse effects per se, it is sometimes difficult to differentiate unfavourable outcomes attributable to glucocorticoids from those occurring as a result of the underlying disease or other comorbidities. There is also a strong selection bias for glucocorticoid use because physicians are inclined to treat patients with more active and severe diseases with glucocorticoids. Finally, it should be noted that frequent but less serious adverse effects (eg, skin thinning, cushingoid appearance) may be of great concern to patients, whereas more debilitating toxicities (eg, osteoporosis, cataracts and glucocorticoid-induced hypertension) may initially go unrecognised or be asymptomatic.

As a consequence, robust data on the long-term toxicities of properly used glucocorticoids are lacking.7 9 A question of special clinical importance is whether or not patterns relating the frequency of adverse effects to glucocorticoid dosage and/or length of glucocorticoid treatment exist. Is there a threshold effect for certain adverse events if we treat with glucocorticoids above a certain dose? Alternatively, is there rather a linear relationship such that more glucocorticoids produce more intense adverse effects? In this article, therefore, we analysed dose-related patterns of patient-reported symptoms from 1066 patients with rheumatoid arthritis (RA) with ongoing long-term (>6 months) glucocorticoid therapy. We also verified these patterns using data from 12 289 RA patients enrolled in the national database of the German Collaborative Arthritis Centres (NDB) in the year 2003.

PATIENTS AND METHODS

The NDB has been the most important source of information regarding health services in German rheumatology since 1993. Collecting clinical and patient-derived data from all outpatients with inflammatory rheumatic diseases in the participating rheumatological units once a year, it allows for the investigation of treatment patterns, trends in treatment and outcomes of rheumatic diseases. In 2003, 88 rheumatological units (university departments, departments of rheumatology at general hospitals and rheumatologists in private practice) recorded a total of 12 289 patients with RA according to the American College of Rheumatology (ACR) criteria.10 The database has been described in greater detail elsewhere.12^–14

From 1 September to 31 December 2003, an add-on module was implemented in nine rheumatological units with the aim of gathering more detailed information on the prevalence, diagnosis and treatment of osteoporosis among patients with RA. These nine units in 2003 enrolled a total of 3792 RA patients, of which 1066 consecutive cases were seen within the respective time frame and reported with data from the osteoporosis module (OM).

The OM comprised among other things detailed information on the prescription of glucocorticoids. This permitted the classification of cases by their duration and dose of glucocorticoid intake. The patient questionnaire used for all cases in the national database in 2003 contained a comprehensive list of 51 subjective health disorders divided into seven organ-related groups. The patients were asked which of these disorders they had had during the past 6 months. Patients could classify them as “light”, “moderate”, “severe”, or “I did not have such health disorders”. Health disorders commonly attributed to glucocorticoid exposure were analysed for possible relationship patterns to the dose of an at least 6-month glucocorticoid intake based on frequencies. Only health problems indicated as “moderate” or “severe” were evaluated in the analysis. Furthermore, relations were verified in multivariate, stepwise forward logistic regression models incorporating potential impact parameters age, sex, disease duration, disease severity and 20 comorbidities. SPSS version 14.0 was used for data analysis.

RESULTS

Representativeness of the sample

The nine participating outpatient clinics and practices form a subsample of all units participating in the NDB in 2003, and the OM was only applied for 4 months in that year. For these reasons, we investigated the representativeness of: (1) the selected units for the whole dataset of the NDB; (2) the OM sample for the total of patients in the respective units and (3) the OM subgroup for the whole dataset of the NDB in 2003. Table 1 shows the comparison of the respective demographic and disease-related data. We found that: (1) the cases enrolled by the nine selected units were representative of the whole dataset of the NDB with regard to demographic, clinical and therapy-related data; (2) the OM subsample was comparable with the total group of patients with RA seen in the respective units and (3) it was also comparable with those patients enrolled in the NDB in 2003 with regard to demographic and most of the clinical data. Only the percentages of patients with glucocorticoid therapy of 7.5 mg/day or less and the proportion of patients with severe pain, poor general health and low functional capacity were slightly higher for the OM group.

Dose-related frequency patterns detected in the OM

Of the 1066 patients reported with the OM, 779 could be assigned to one out of four distinct groups with regard to their current and past glucocorticoid intake. Group 1 comprised patients who had no glucocorticoid for at least 12 months, and groups 2 to 4 comprised those with glucocorticoid therapy for more than 6 months at a current dose of less than 5, 5–7.5 or more than 7.5 mg prednisone equivalent per day, respectively (table 2). Patients with short-term glucocorticoid use during the past 12 months were excluded. A total of 307 out of 1066 patients (29%) had not undergone any glucocorticoid therapy during the past 12 months and 472 (44%) had been on glucocorticoids for more than 6 months. Of these, 60% were treated with dosages between 5 and 7.5 mg/day, whereas dosages of less than 5 mg/day were used for maintenance therapy in only 22%.

Table 2 shows the prevalence of self-reported, glucocorticoid therapy-related health problems for these four groups. In general and as expected, the frequency was lowest in the group without recent glucocorticoid exposure and increased with dosage. We observed two distinct patterns: a continuous, approximately linear rising and a threshold pattern (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Examples of “linear” (ecchymoses) and threshold (glaucoma) pattern of adverse event rates. GC, glucocorticoid.

For cushingoid phenotype, ecchymosis, leg oedema, mycosis, parchment-like skin, shortness of breath and sleep disturbance, a “linear” increase in the frequency with increasing dose was identified. The term “threshold pattern” describes an elevation in the frequency of health problems beyond a certain threshold value. The threshold for the increase in glaucoma, as well as depression/listlessness and an increase in blood pressure, was found in the OM group at a dosage of more than 7.5 mg/day. Dosages of 5 mg/day or more turned out to be relevant for epistaxis and weight gain. A very low threshold was seen for eye cataract (<5 mg/day).

We could confirm the influence of the glucocorticoid dose on the incidence of most of the respective adverse events in multivariate logistic regression models (table 3). For mycosis, shortness of breath and eye cataract we did not find a significant impact of glucocorticoid dose as other factors were more determinant for the occurrence of the considered health problems. For adverse events with a threshold pattern greater than 7.5 mg/day we can not make a reliable statement on the effect of the glucocorticoid dose, because case numbers were too small in the highest and thus relevant dose group for multivariate regression models.

There was no increase in the use of other drugs with increasing glucocorticoid dose, with the exception of leflunomide and analgesics (data not shown). As these substances are not commonly associated with the adverse events considered, it is very unlikely that other drugs generated the “linear” incidence pattern. With regard to threshold patterns, the following medications showed a simultaneous increase in prescription rates: parenteral gold, azathioprine and bisphosphonates at more than 7.5 mg/day prednisone, infliximab at 5–7.5 mg/day and calcium/vitamin D at less than 5 mg/day. None of the events observed, however, is known to be associated with either of these drugs.

Verification of the patterns in the national database

We then investigated whether we could confirm these patterns using the much larger dataset of the NDB. The limitation of this approach is that the case report form for the database only distinguishes between glucocorticoid doses of 7.5 mg/day or less and above and, therefore, only these two groups can be compared. Crucial parameters describing the cohort of RA patients summarised in the NDB are given in table 4.

There is no striking difference when comparing these data with those of the OM sample given in table 2, although the percentage of women in the dosage category more than 7.5 mg/day was higher in the OM sample. As there is no distinction between less than 5 and 5–7.5 mg/day in the NDB, a threshold pattern at 5–7.5 mg/day was not identifiable. We therefore expected a “linear” trend for those adverse events in which we had observed a threshold pattern at 5–7.5 mg/day. Except for sleep disturbance and eye cataract, all patterns were confirmed under these conditions (table 4).

DISCUSSION

We used data from the NDB. The participating rheumatologists can be considered representative of healthcare in German rheumatology as a whole.

In addition to the data from the NDB, we administered an add-on module examining osteoporosis diagnostics and therapy in a selection of units for one third of the year 2003. Comparing patient characteristics of these three samples, we found that patients in the selected units did not differ from those seen in the other units for any of the parameters under study. Patients included in the OM did not differ in clinical aspects, but were more often treated with glucocorticoids and had slightly more pain and disability. This minor selection bias can be explained by the fact that more severely ill patients are seen more often by a rheumatologist; therefore, these patients have a higher a priori chance of being included if there is only a limited time of enrolment. However, this difference is small enough to be of no concern.

The strength of our approach is that we had specialist-derived diagnosis and treatment data as well as patient-derived data on general symptoms and quality of life indicators. Taking advantage of a subsample of patients for whom we had more detailed data on exposure to glucocorticoids, we could identify two distinct patterns of dose-related prevalence of adverse symptoms. We then confirmed these patterns in the larger dataset. We are aware that adverse symptoms cannot be attributed to one or several singular mechanisms of glucocorticoid action,15 because our patients were not under glucocorticoid therapy alone. More than 80% of them were also treated with disease-modifying anti-rheumatic drugs and more than 90% were treated with other drugs. Another limitation is that the health problems were self-reported by patients and could not be post-validated for this analysis. Therefore, we restrained our exploration to health disorders that are known to be associated with glucocorticoid use. As we found many of our results in accordance with known mechanisms and confirmed effects from the literature, we assumed that the patient reports were not severely biased. Not all patient-reported adverse events will undergo evaluation by a physician anyway. The application of our reference group with “no glucocorticoid exposure in the past 12 months” might abate observed glucocorticoid effects, but with a lack of complete medication history we were unable to select a preferable control group of completely glucocorticoid-naive patients.

The most clearly attributable adverse drug reaction to glucocorticoids, Cushing’s syndrome, becomes evident after at least one month of treatment, with an incidence that is dose dependent,16 as we could see in our data. In one study, “moon faces” developed in 13% of patients receiving 4–12 mg of triamcinolone for 60 days or less.17 We observed a cushingoid phenotype in 16% of our patients exposed to 5–7.5 mg/day prednisone equivalent for more than 6 months.

Centripetal fat accumulation is a characteristic feature of patients on long-term high-dose glucocorticoids, but it is even seen with low doses. In three out of four prospective trials,18^–21 weight significantly increased under glucocorticoid treatment within 1–2 years; only in two studies, however, was this weight gain different from that in the control group.20 21 We observed an elevation of reported recent weight gain at dosages of 5 mg/day or greater. Along with weight gain, there was an increase in patient reports of leg oedema and shortness of breath, both following a “linear” pattern.

Even at low doses, skin thinning and ecchymosis are among the most common glucocorticoid adverse events.22 For ecchymosis and parchment-like skin, our data demonstrate a remarkable increase already at less than 5 mg/day and this increase grows with higher dosages. Catabolic skin effects during systemic glucocorticoid therapy include cutaneous atrophy. Moreover, decreased vascular structural integrity is very likely another key determinant of purpura and easy bruisability in glucocorticoid-treated patients.22 23 These effects were reported to affect over 5% of those patients exposed to 5 mg or more of prednisone equivalent for one year or more.24 25 Even in patients with less than 5 mg/day of glucocorticoids over a course of over 6 months, we observed an increased frequency of skin bruising compared with the group with no recent glucocorticoid exposure. For epistaxis, we saw a threshold effect at dosages higher than 5 mg/day; this is certainly also an implication of decreased vascular structural integrity.

Hypertension is a well-established adverse effect of glucocorticoids, and it is observed in approximately 20% of patients exposed to exogenous glucocorticoids.26 It needs to be stressed that hypertension is considered a dose-related adverse effect and is less likely with lower dosages.22 We can confirm this, because we observed a dose-related increase of the parameter “increased blood pressure” only at dosages greater than 7.5 mg/day. Toxicity data from four extensively reviewed trials on low-dose glucocorticoids in RA were analysed with regard to hypertension. In neither of the studies was prednis(ol)one at 5,27 7,19 7.518 28 or even 10 mg/day21 found to have significant effects on blood pressure. It should be noted, however, that patients with severe hypertension were excluded from most of the trials. These results may support the idea that the threshold for glucocorticoids to induce hypertension is indeed above 7.5 mg/day. One retrospective study of 195 patients with RA or asthma treated with less than 20 mg/day prednisone for more than one year did not show any correlation between dose or duration of glucocorticoid treatment and a rise in blood pressure.29 However, that study may not have been adequately powered to identify the differences we were able to find.

The exact incidence of minor mood disturbances, such as depressed mood, in patients exposed to common doses of glucocorticoids cannot be drawn from the literature.22 Many patients with RA report a slight increase in their overall sense of wellbeing when beginning low-dose glucocorticoid therapy; this appears to be independent of an improvement in disease activity. Symptoms of akathisia, insomnia and depression are also occasionally observed in patients taking low-dose glucocorticoid therapy.30 We observed that the frequency of depression/listlessness rose only at a glucocorticoid dose greater than 7.5 mg/day. Daily split-dose therapy, in particular, tends to be troublesome, because the evening dose disrupts normal diurnal variation in endogenous glucocorticoid levels and promotes sleep disturbances.30 In our data, sleep disturbance seemed to accompany increasing doses.

Posterior subcapsular cataracts are a well-described complication of prolonged glucocorticoid use. Limited to long-term glucocorticoid users, we saw a remarkable escalation in the frequency of eye cataracts at low doses of glucocorticoids. In a matched-pairs analysis of RA patients, 15% of eye cataracts were found in patients taking an average of 6 mg/day prednisone compared with 4.5% in controls.31 This is consistent with our findings.

Medium to high-dose glucocorticoid therapy, particularly when administered for prolonged periods, may lead to an increased risk of serious infections. With regard to lower glucocorticoid dosages, a meta-analysis showed that the rate of infection was not significantly increased in patients given a mean dose of less than 10 mg/day of prednisone or a cumulative dose of less than 700 mg.32 Of the infections we examined, a slight dose relation appeared for mycosis; however, an impact of the glucocorticoid dose on mycosis was not confirmed in the multivariate analysis.

To summarise our findings: at intake for more than 6 months, glucocorticoid treatment at low dosages (<5 mg/day) could avoid elevated incidences of cushingoid phenotype, epistaxis, leg oedema and weight gain. From our data, glaucoma will only become an issue at dosages above 7.5 mg/day. Ecchymosis and parchment-like skin become more frequent with increasing doses.

Although we relied on patient reports alone for the establishment of adverse events, the associations found are in agreement with biological mechanisms and clinical observations. As there is a paucity of real-life data on the adverse effects of glucocorticoids prescribed to unselected groups of patients, our data may help the clinician to adapt therapy with glucocorticoids accordingly and improve the benefit–risk ratio.