Corticosteroids Do Not Alter the Threshold for Vertebral Fracture


  • Presented in part at the second joint meeting of the IBMS and ASBMR, San Francisco, California, U.S.A., December 1998.


Corticosteroid use is one of the most important secondary causes of osteoporosis. Generally, it has been believed that in addition to its effect on bone mineral density (BMD), it also causes an alteration in bone quality that means that fractures occur at a lower BMD than might be expected. To establish if this is the case, we have compared the relationship between BMD and vertebral fracture in patients receiving corticosteroids with that in patients who had never received such therapy. Information was gathered on those patients who had been referred to the participating centers and had both BMD measurements and lateral thoracolumbar radiographs. In all, 452 patients (391 female) were identified; of these 82 (63 female) were receiving corticosteroids. There was no significant difference in BMD between the patients on corticosteroids and those with other suspected causes of osteoporosis. Vertebral fractures were present in 53% of patients on steroids compared with 35% of those who had no such treatment (p = 0.0035). The fractures were more likely to be multiple in patients on corticosteroids (p = 0.0042). However, if the relationship between bone density and fracture is investigated by plotting the cumulative prevalence of fracture against the bone density, measured by T score, the median BMD for fractures actually was marginally lower in patients on steroids, −2.74 (95% confidence interval [CI], −2.77 to −2.70) compared with −2.65 (95% CI, −2.66 to −2.65) in those who had not received steroids. Our results fail to support the notion that the fracture threshold is altered in patients on long-term steroids and suggest that the same diagnostic criteria should be used for osteoporosis in patients whether or not they are taking corticosteroid therapy. (J Bone Miner Res 2000;15:952–956)


Corticosteroid treatment has been one of the major advances of the latter half of the 20th century. The use of these drugs has improved the outcome in patients with a variety of diseases drawn from many different areas of medicine. Unfortunately, this therapeutic gain has been obtained at the cost of significant adverse effects.

One of the most important of these is the increased risk of bone loss and fracture in the patient receiving long-term corticosteroid therapy. Recent studies have suggested that osteoporotic fractures may occur with twice the frequency in patients receiving corticosteroid treatment than in their nonsteroid-treated peers.(1,2) Although it is well known that patients receiving corticosteroids may suffer bone loss as a result of this some authors have suggested that corticosteroid treatment also may increase the fracture risk by altering bone quality. In particular Luengo et al. showed in a small group of patients receiving corticosteroids for asthma that the bone mass associated with fracture was substantially higher than in those patients who had vertebral fracture as a result of idiopathic osteoporosis.(3) In a study of patients with rheumatoid arthritis who were receiving corticosteroid therapy Peel et al. suggested that the disturbance in the expected relationship between bone mass and fracture might indicate alteration of bone quality.(4) In this case it is not possible to dissect out separate effects of corticosteroids themselves and the underlying rheumatoid disease on bone mass and fracture.

If the use of corticosteroids does has an effect on fracture risk independently of the effects on bone density this would have important implications for the identification of patients receiving corticosteroids who might benefit from bone sparing therapy. In particular, the currently accepted bone mass thresholds for intervention with bone sparing agents would not apply to patients on corticosteroids.(4,5) This would require the development of different treatment strategies for patients on corticosteroids compared with those suffering from idiopathic osteoporosis.(6) Because the bisphosphonates etidronate and alendronate have been shown to minimize bone loss in patients receiving corticosteroid therapy, it is important that these adjuvant therapies be targeted at appropriate individuals.(7,8) To determine whether or not the effect of corticosteroids to induce fracture is mediated through mechanisms other than bone density, we have examined the bone densities associated with fracture in patients receiving corticosteroids for a variety of different reasons.


Information was retrieved from the radiological databases of six centers offering bone mass measurement in northwestern England. Successive patients in whom both bone mass and lateral thoracolumbar radiographs were available were eligible for inclusion. At that stage patients were unselected with regard to age, gender, corticosteroid use, underlying disease, or reason for bone density measurement. Patients who had previously received corticosteroids and those currently taking them who had received them for less than 6 months were subsequently considered ineligible for study as were patients who were already receiving bisphosphonates or hormone replacement therapy.

Bone mineral density (BMD) was measured in the lumbar spine and proximal femur by dual energy X-ray absorptiometry (DXA) in all subjects. In three of the centers the equipment employed was supplied by the Lunar Corp. (Madison, WI, U.S.A.) and in two of the centers the equipment was supplied by Hologic (Waltham, MA, U.S.A.); the remaining center had equipment from both manufacturers.

The presence of vertebral fractures was based on the reports of the lateral spinal radiographs present in the case notes. The criteria for diagnosis of fracture were, therefore, not predefined but did represent those which generally are applied in clinical practice in the United Kingdom. If the presence of a vertebral fracture was recorded, it also was noted whether this was a single fracture or if multiple deformities were present.

Statistical analysis

Because of the differences in normal ranges between these machines, standardized BMD was calculated and used for subsequent analysis.(9) In addition, the T score was derived for lumbar spine BMD using the respective manufacturer's normal range.

Differences in the proportions of patients with fracture between the two groups were examined using Fisher's exact test. Two-way analysis of variance was used to examine the effects of gender and steroid use on BMD.

Patients within both the steroid-treated groups and the nontreated groups were ranked according to lumbar BMD. The cumulative prevalence of vertebral fracture was then determined and plotted against bone density. A sigmoid curve was matched to each distribution using the nonlinear regression facility within Prism. The curve was constrained to have a minimum and a maximum of 0 and 1, respectively. From this curve the bone mass at which 50% of the fractures will have occurred (together with its 95% confidence interval [CI]) was estimated.

Statistical analysis was undertaken using GraphPad Prism (GraphPad, Inc., San Diego, CA, U.S.A.).


In all, 452 eligible patients were identified and further information was obtained from the case notes. Three hundred ninety-one of the 452 (87%) were female, 82 (18%) of the patients were currently receiving in excess of 5 mg of corticosteroids daily and had done so for at least the preceding 6 months. Of the 82 patients receiving corticosteroids 63 (77%) were female.

The characteristics of the patients who had received corticosteroids are compared with those who had not in Table 1. The patients who have not received corticosteroids had a mean age of 62 years (SD 10). This was not significantly different from those who had received steroids who were aged 65 years (SD 8). Corticosteroids were given for respiratory diseases in 20% of the cases, rheumatic diseases in 44%, polymalgia rheumatica in 20%, neurological conditions in 9%, and skin diseases in 7% of subjects. The mean dose of corticosteroids was 7.5 mg of prednisolone equivalent per day and the median duration of use was 5 years (range, 0.5–34 years).

Fractures were present in 43 of the 82 patients on steroids compared with 130 of the 370 who had not received them (53% versus 35%; p = 0.0035 Fisher's exact test). Of the 43 patients on steroids who had suffered fractures 16 (37%) were recorded as having multiple fractures. These were present in only 20 of the 130 patients who had not received steroids and who had suffered fracture (15%; p = 0.0042 Fisher's exact test).

Table Table 1.. Characteristicsofthe Corticosteroid- and Nonsteroid-Treated Populations
 Corticosteroids (n = 82)No steroids (n = 370)p
  1. Values are given as percent or mean (SD).

Percent female77880.001
Age (years)65 (8)62 (10)NS
Machine used (percent lunar)72860.03
Mean BMD   
 Hologic (g/cm2)0.867 (0.135)0.856 (0.192)NS
 Lunar (g/cm2)0.900 (0.158)0.890 (0.164)NS
 Standardized (mg/cm2)878 (152)857 (166)NS
Men with fracture (%)63.251.8NS
 Women with fracture (%)46.829.50.001
Table Table 2.. Standardized Lumbar Spine BMD (mg/cm2) in Menand Women Who Are Takingor Have Never Received Corticosteroids
On steroids862 ± 19956 ± 34
No steroid use856 ± 10874 ± 16
Figure Fig. 1..

Relationship between the cumulative frequency of patients with vertebral fracture and bone mass, expressed as lumbar T score, in patients on corticosteroids (•) and those who have never received steroid therapy (○).

The standardized BMD measurements are given in Table 2. Two-way analysis of variance indicated that bone mass was significantly greater in men than women (F = 5.03 and p = 0.025). There was no significant effect of corticosteroid usage on bone mass (F = 3.11 and p = 0.079) and no interaction between the effect of gender and steroid use (F = 2.32 and p = 0.13).

When the bone masses were expressed as T scores, the difference between the genders disappeared. For women the mean T score was −2.93 (1.43) in those not taking steroids and −2.4 (1.2) in those taking steroids. For men the corresponding values were −2.49 (1.04) and −2.0 (1.3). Applying two-way analysis of variance to these results revealed no significant main effect of steroid use or gender on bone mass and no interaction between steroid use and gender. To allow the data from the men and women to be combined, bone mass has been expressed as T scores in subsequent analyses.

Figure 1 shows that the relationship between bone density, expressed as T score, and the cumulative proportion of fractures is similar in those patients who have received steroids to those who have not. Furthermore, the shape of the sigmoid curve suggests that both groups of patients have a normal distribution of bone mass in patients with fracture. Thus, the point at which 50% of fractures will have occurred represents the median bone density associated with fracture. This occurs at a T score of −2.65 (95% CI, −2.66 to −2.65) in patients who have not receive steroids compared with −2.74 (95% CI, −2.77 to −2.70) in those who have. Because these CIs do not overlap it can be seen that the median bone density associated with fracture is statistically significantly different between those who received steroids and those who have not. On the other hand, the magnitude of the difference of less than one-tenth of an SD unit is unlikely to be associated with any clinically relevant difference. In any case the difference in bone density would suggest that, if anything, patients on corticosteroids do not suffer fractures until they have lost more bone than those not receiving such therapy.

It might be argued that differences in gender, steroid dose, and underlying disease might all have influenced the relationship between BMD and fracture. This could have attenuated any difference between the two study groups. Although it is difficult to control for all the differences, there were sufficient data obtained for women alone to allow a separate analysis to be undertaken. The results of this are shown in Fig. 2. From this it can be seen that the results obtained with women do not differ significantly from those obtained in the population at large. The median bone density associated with fracture was −2.68 (–2.69 to −2.67) for those women who had not received steroids and −2.84 (–2.87 to −2.80) in women on steroid treatment. Although these differences are statistically significant, the magnitude of the difference is such that there is unlikely to be any clinical relevance.

From Table 1 it can be seen that there was a difference between the proportion of patients who had received corticosteroid therapy measured by each of the two machines. To ensure that this did not introduce bias in the results, they were analyzed separately for each machine. This did not significantly alter the results, with the median T score for fracture in patients receiving corticosteroids being −2.80 (–2.84 to −2.77) on Lunar machines and −2.51 (–2.73 to −2.29) on Hologic machines. This compares with values in the control group of −2.47 (–2.48 to −2.45) with Lunar densitometers and −2.57 (–2.62 to −2.53) on Hologic machines. Although the difference on Lunar machines is statistically significant and larger than that seen in the group as a whole, it is in the opposite direction from what would be accepted as current dogma; in other words it would tend to suggest that patients on corticosteroids do not fracture until they reach a lower bone density.

Figure Fig. 2..

Relationship between cumulative frequency of patients with vertebral fracture and bone mass in women on corticosteroids compared with those who have never received them. Symbols are the same as in Fig. 1.


These results suggest that the distribution of bone density in patients with fracture is very similar whether or not they have been receiving corticosteroid therapy. These results therefore differ from those previously reported as suggesting a change in fracture threshold in patients on corticoste-roids.(3) Our data would suggest that patients receiving corticosteroids have a similar BMD threshold for fracture as those with osteoporosis for other reasons.

Our results do have several limitations. Perhaps the most important of these is that because patients were drawn from several different hospital sources, there is little to ensure consistency of patient selection or, more importantly, diagnosis of fracture. This is an inherent problem with our study design. However, because there is no reason to believe that different criteria for the diagnosis of fracture would apply in patients on steroids, none of these factors is likely to have introduced a systematic bias. It is more likely that they would have led to imprecision in our estimates and reduced the apparent significance of our results.

On the other hand, the very fact that patients represent a wide clinical and geographical range means that there is more likely to be general applicability of our results. Indeed, the distribution of reasons for prescription of corticosteroids is very similar to that described in a population survey of a long-term corticosteroid use.(10) In addition, the dose and duration of steroids given to our patients were similar to those seen in the U.K.-based population survey. All of this suggests that our population group is likely to be reasonably representative of patients receiving corticosteroids in the United Kingdom.

No study that focuses on looking at the effect of corticosteroids on bone is able to truly take into account the effect of the underlying disease. Because corticosteroids generally are given to people with more severe forms of the underlying disease, there is a significant risk that some of the effects attributed to steroid usage might be the result of the underlying disease. In this respect, our study is no different from other studies of corticosteroid-induced osteoporosis. It is virtually impossible to make allowances for this particular confounding factor apart from a few particular instances in which there is a genuine clinical controversy regarding the use of corticosteroids in a specific disease state.

Also, be it must remembered that in attempting to define the bone mass associated with fracture, we are making longitudinal inferences from a cross-sectional study. This can only be considered a hypothesis-generating process. To show a true difference in fracture threshold, it would be necessary to undertake a prospective study to determine what level of BMD is associated with fracture risk in patients on and off corticosteroid therapy.

Measurement of BMD by the DXA in the lumbar spine is prone to several sources of error. The most important of these would be the influence of extraskeletal calcification.(11,12) However, we had little option but to use this site. Although BMD was measured at the proximal femur in all patients the standardization of this measurement is based on the total hip measurement.(13) Many of our patients who were measured on a Lunar machine did not have this measurement available. Furthermore those units using Hologic machines have not universally applied the National Health and Nutrition Examination Survey (NHANES) data set and so the T scores derived from these machines probably were inaccurate and it would have been inappropriate to combine these with those derived from Lunar instruments.(14) Analysis of the results from the two types of equipment separately does not lead to any different conclusions.

The fact that there was no discernible effect of corticosteroid use on BMD in this study should not be taken as an indication that corticosteroids are having no effect on bone mass. It must be remembered that all the patients referred to our units for assessment of bone density had been seen by a clinician who felt that there was reason to suspect osteoporosis. All that our observation is telling us is that patients who are referred for assessment of possible osteoporosis based on corticosteroid therapy are neither more, nor less, osteoporotic than those patients referred for assessment of possible osteoporosis because of other causes.

Despite the similar bone mass seen in patients on and off corticosteroids and the similar relationship between bone mass and fracture, in the two groups there were more fractures apparent in patients with corticosteroids. Furthermore, the fracture diathesis was worse in patients on corticosteroid in as much as there was a greater proportion of patients with multiple fractures. At first sight this seems to present a paradox because if the corticosteroids were causing increased fracture risk by an effect independent of BMD, we might expect fractures to be seen at higher BMD on steroids just as reported by Luengo.(3) Because this is not the case it is difficult to attribute these fractures to the changes in bone quality that have been described on corticosteroid therapy.(15,16) However, it must be remembered that our study was carried out in patients who had received corticosteroids for a long period of time. It must be presumed that these patients have come through the initial rapid phase of bone loss associated with the introduction of such therapy and are in a relatively stable state even though they may well be losing bone at a greater rate than normal.(17,18) There is now increasing evidence to suggest that high bone turnover, particularly that associated with secondary hyperparathyroidism, is associated with an increased risk of fracture.(19) At the time of introduction of corticosteroid therapy there is such a period of increased bone turnover, which may be associated with secondary hyperparathyroidism. The increased number of fractures seen in patients on corticosteroids, therefore, may not reflect their current state of bone mass but the severity of the secondary hyperparathyroidism and increased bone loss immediately after the institution of therapy. This proposition would be capable of testing as part of a prospective study the effect of corticosteroid therapy on fracture risk as proposed above. We have shown that there does not appear to be a difference in the relationship between BMD and vertebral fractures in patients receiving corticosteroids compared with those not taking that medication. Therefore, we do not believe that corticosteroids are influencing fracture risk by a mechanism independent of bone mass and we do not believe that there is any need to apply different criteria for the diagnosis of osteoporosis in patients on corticosteroid therapy.


We are grateful to Procter and Gamble Pharmaceuticals for assistance with the collation of the data used in this study.