Recent corticosteroid use and recent disease activity: Independent determinants of coronary heart disease risk factors in systemic lupus erythematosus?




Systemic lupus erythematosus (SLE) is characterized by a markedly elevated risk for coronary heart disease (CHD), the exact pathogenesis of which is unknown. In particular, the causal roles of corticosteroid therapy and SLE disease activity, and whether their putative effects are mediated through conventional risk factors, remain unclear.


Data abstracted retrospectively from the charts at 11,359 clinic visits for 310 patients with SLE to the Montreal General Hospital were used to investigate the associations of recent corticosteroid dose and recent Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) score with 8 CHD risk factors (total serum cholesterol, high-density lipoprotein [HDL] cholesterol, low-density lipoprotein cholesterol, apolipoprotein B [Apo B], triglycerides, systolic blood pressure [BP], body mass index, and blood glucose) and the aggregate estimate of 2-year CHD risk. Separate multivariable linear regression models estimated the mutually-adjusted effects of average daily corticosteroid dose and average SLEDAI score within the past year on the current level of each risk factor while adjusting for age, sex, cumulative damage score, disease duration, and, where appropriate, use of relevant medications.


Higher past-year corticosteroid dose was independently associated with significantly higher overall 2-year CHD risk and with higher levels of all 8 individual risk factors. Higher past-year lupus disease activity was independently associated with higher overall 2-year CHD risk, lower HDL cholesterol, and higher values of systolic BP, Apo B, triglycerides, and blood glucose.


In SLE, both recent use of corticosteroids and recent lupus activity are independently associated with higher values of several well-recognized CHD risk factors and overall 2-year CHD risk.


The dramatic increase in coronary heart disease (CHD) in patients with systemic lupus erythematosus (SLE) has been widely recognized and is an important concern of current clinical research (1, 2). However, the pathogenesis of CHD is unknown. Because baseline values of conventional CHD risk factors alone do not explain the risk increase (3), an investigation of the role of SLE-specific risk factors is a priority. However, most patients with SLE are young or middle-age women, for whom the background rate of CHD outcomes is very low. In addition, most SLE cohorts are small; therefore, the actual number of observed CHD events may not provide sufficient statistical power for testing their associations with putative SLE-specific risk factors. To avoid such limitations, recent studies of CHD risk in SLE have considered surrogate cardiovascular outcomes such as the presence of carotid plaques (1), coronary artery calcification (2), vascular stiffness (4), and/or endothelial function (5). However, SLE-specific risk factors may increase CHD risk by increasing the values of conventional risk factors, which are highly predictive of actual CHD events (6).

Among putative SLE-specific risk factors for CHD, corticosteroid medication use is central. Petri et al reported that prednisone dose is associated with increases in total serum cholesterol, mean arterial blood pressure (BP), and body weight (7). Others reported similar findings (8–14). However, methodologic limitations can affect the estimated associations between corticosteroid use and CHD risk factors. Reported analyses often fail to account for the use of other medications and lupus disease activity. Disease activity may be an important potential confounder of the apparent associations between corticosteroid use and CHD risk factors. First, recent SLE activity is associated with corticosteroid use and higher doses of corticosteroids, which induces a risk of confounding by indication. Second, SLE activity may independently lead to hypercholesterolemia, dyslipoproteinemia, hypertension, weight gain, and impaired glucose metabolism (15–18). Other limitations of most published studies of the associations between corticosteroid use and conventional CHD risk factors include low statistical power and lack of a detailed description of how the corticosteroid dose was calculated.

To explore the complex relations of SLE-specific factors with CHD risk while attempting to overcome the above limitations, in this study we used a longitudinal database with 11,359 visits on 310 lupus patients; provided an explicit algorithm for calculation of corticosteroid dose; adjusted for and investigated the independent associations with SLE activity; accounted for temporal aspects of the investigated associations; and quantified the overall impact of corticosteroid use and SLE activity on CHD risk by using an aggregate score, which is a powerful predictor of subsequent clinical CHD events (6).

We have investigated the independent associations of recent corticosteroid dose and lupus activity with total serum cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, apolipoprotein B (Apo B), triglyceride, systolic BP, body mass index (BMI), and blood glucose level, as well as with the overall level of CHD risk (19).


Data source.

We used data from the Montreal General Hospital Lupus Clinic from 1971–2003. Details of the data collection and followup procedures are described elsewhere (20). All patients who met the American College of Rheumatology (ACR) classification criteria for diagnosis of SLE (21) were included.

Each patient was followed from the diagnosis of SLE or the first visit to the clinic until the development of a cardiovascular or cerebrovascular event (i.e., angina pectoris, myocardial infarction, transient ischemic attack, or cerebrovascular accident), peripheral vascular disease, December 31, 2003, or loss to followup, whichever came first. Subjects with a cardiovascular or cerebrovascular event or peripheral vascular disease prior to their first visit were excluded. A trained research nurse reviewed patients' records and abstracted data on medication use, disease activity, and physical and laboratory measurements for each visit. The cardiovascular, cerebrovascular, and peripheral vascular outcomes were independently adjudicated by an investigator (LP) blinded to other patient characteristics.

Measurement of corticosteroid dose.

To assess recent exposure to corticosteroids, we had to retrospectively calculate corticosteroid doses over the relevant time window. All corticosteroid use with a mode of administration other than intraarticular, topical, or inhalation was documented. The doses for all corticosteroid types were converted into equivalent prednisone doses (22). First the corticosteroid dose was reconstructed for each day of the followup. If a tapering algorithm was explicitly provided in the patient's record, the dose for each day was calculated accordingly. Otherwise, the following algorithm was adopted. For corticosteroid medications administered orally, if a prednisone-equivalent dosage prescribed at a given visit was <100 mg/day, then the dosage was tapered at a rate of 10% per week until either the next dose was recorded or the medication was stopped. If the dosage was ≥100 mg/day, it was tapered at a rate of 20% per week until it reached 100 mg/day, and at a rate of 10% per week thereafter. These tapering algorithms were commonly practiced in the Montreal General Hospital Lupus Clinic during the study period (John Esdaile, personal communication). If the time period between 2 consecutive visits exceeded 6 months, the daily corticosteroid dose was reconstructed for the first 6 months only, and thereafter was set to missing until the next visit. After the dose was reconstructed for each day, the average daily prednisone-equivalent dose for 1 year prior to a given visit was calculated using only days with a nonmissing reconstructed dose. If the reconstructed dose was available for less than one-third of the preceding year, the corresponding observation was not used in the analysis.

Measurement of disease activity.

Lupus disease activity was measured using the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) score (23), retrospectively scored (24) at each visit, based on abstracted data. The SLEDAI score for each day between visits was imputed as the mean of the 2 scores. If only the earlier SLEDAI score was available, it was carried forward for all subsequent visits. Finally, the average SLEDAI score during the year prior to each visit was calculated as the mean of all imputed or observed daily scores.


Cumulative organ damage, based on the Systemic Lupus International Collaborative Clinics/ACR Damage Index (SDI) score (25), was recorded once a year (26). Current disease duration was computed as the difference between the date of the given visit and the date of the lupus diagnosis (20). At each visit, the proportions of days during the past year in which a lipid-lowering, antihypertensive, and blood-glucose–lowering medication was taken were computed.

Conventional CHD risk factors and overall CHD risk.

The following CHD risk factors served as outcome variables: total serum cholesterol, HDL cholesterol, LDL cholesterol, Apo B, triglycerides, systolic BP, blood glucose, and BMI. Only visits with a given risk factor measurement available were used in the analysis.

To assess the overall impact of changes in different risk factors, we computed the 2-year coronary heart disease risk score for each patient at each visit as an aggregate measure of total serum cholesterol, systolic BP, BMI, and glucose intolerance (nonfasting blood glucose level of ≥6.7 mmoles/liter and/or a prescription for a glucose-lowering medication). First, a multiple logistic regression model was estimated from the Framingham Heart Study population (19). The binary dependent variable was the occurrence of CHD (myocardial infarction, angina pectoris, coronary insufficiency, or CHD death) in the next 2 years (27). The independent variables were sex, current age, total serum cholesterol, systolic BP, BMI, number of cigarettes smoked, and glucose intolerance (the estimated regression coefficients are shown in Appendix A). In calculations, we used the mean values of age, the number of cigarettes smoked, and the proportion of men in the sample because it was very unlikely that corticosteroid use and/or lupus activity could affect these variables. We also used sample mean BMI because BMI information was missing for 82% of the visits at which other risk factors were measured. Then, the estimated probability of a CHD event in the 2 years after a given visit for a given subject was calculated by multiplying the regression coefficients from the multiple logistic model by, respectively, mean values of age, sex, number of cigarettes smoked, and BMI, and by actual individual values of total serum cholesterol, systolic BP, and glucose intolerance observed at a given visit. Thus, variation in SLE-specific risk factors was associated with variation in overall CHD risk due to total serum cholesterol, systolic BP, and/or glucose intolerance.

Statistical analysis.

As the distributions of blood glucose and of the estimated 2-year coronary risk were highly positively skewed, these variables were log-transformed. Primary analyses relied on linear mixed model generalization of multiple linear regression for longitudinal data (28) with repeated measures of a specific CHD risk factor as the dependent variable. These models incorporated a random intercept to account for the correlation among repeated measures from individual subjects, assuming the variance components covariance structure (29). Nine separate multivariable mixed models were fitted to estimate the mutually-adjusted effects of the average daily corticosteroid dose and the average SLEDAI score in the year before a given visit on the current level of the 8 risk factors (total serum cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, Apo B, systolic BP, BMI, and blood glucose), and on the overall coronary risk in the next 2 years. All of the models were adjusted for age, sex, lupus disease duration, past-year use of cyclophosphamide, and past-year use of other immunosuppressive medications (i.e., azathioprine, methotrexate, cyclosporine, leflunomide, or mycophenolate mofetil). In addition, the models for total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, and Apo B adjusted for past-year use of lipid-lowering medications, while the models for systolic BP and blood glucose adjusted for past-year use of antihypertensive and blood-glucose–lowering medications, respectively. The model for the overall 2-year CHD risk adjusted, in addition, for past-year use of lipid-lowering and antihypertensive medications. Finally, to test whether the impact of increased prednisone dose varied depending on recent SLE activity, we added to each model the interaction between the average daily corticosteroid dose and the average SLEDAI score for the past year.

All the models were repeated in a sensitivity analysis to check the robustness of the results with respect to reconstruction of corticosteroid doses and SLEDAI scores. In the sensitivity analysis, we reconstructed corticosteroid doses for each day during followup, assuming that no tapering took place; i.e., the initially-prescribed dose was kept constant until the next recorded dose change. Similarly, the previous visit's SLEDAI score was imputed for all days until the next SLEDAI score was available. As the results did not change materially, only the results from the primary analyses are reported.


Overall, 310 patients contributed 11,359 observations during the total followup of 4,094 person-years. Patient characteristics at baseline and across the followup are summarized in Table 1. The median followup was 12.4 years (interquartile range [IQR] 6.7–18.4), and the median number of visits was 29 (IQR 10–62).

Table 1. Characteristics of study base, Montreal General Hospital Lupus Clinic cohort*
CharacteristicBaselineAcross followup visits
  • *

    Values are the mean ± SD unless indicated otherwise. SLEDAI = Systemic Lupus Erythematosus Disease Activity Index; LDL = low-density lipoprotein; HDL = high-density lipoprotein; BP = blood pressure; BMI = body mass index; CHD = coronary heart disease.

Age, years38.5 ± 13.841.4 ± 13.6
Male sex, %1110
Prednisone-equivalent dose, mg17.0 ± 68.417.3 ± 71.0
SLEDAI score7.6 ± 3.64.9 ± 4.9
Total serum cholesterol, mmoles/liter5.2 ± 1.65.1 ± 1.4
LDL serum cholesterol, mmoles/liter3.0 ± 1.12.9 ± 1.0
HDL serum cholesterol, mmoles/liter1.4 ± 0.41.4 ± 0.5
Triglycerides, mmoles/liter1.9 ± 1.41.6 ± 1.0
Apolipoprotein B, mmoles/liter1.1 ± 0.51.0 ± 0.3
Systolic BP, mm Hg123.4 ± 19.3122.8 ± 17.8
BMI, kg/m223.1 ± 4.624.1 ± 5.2
Blood glucose, mmoles/liter5.2 ± 1.35.2 ± 1.9
2-year CHD risk0.005 ± 0.0040.005 ± 0.004

Preliminary analyses indicated a significant positive Pearson correlation of 0.32 (P < 0.0001) between mean corticosteroid dose and mean SLEDAI score in the same year, pooled across all patient visits. This confirmed that SLE disease activity may be an important confounder for the corticosteroid dose, and may underscore the importance of adjusting each of the 2 variables for the other.

Table 2 summarizes the results of the mixed model analyses, with each row corresponding to a separate outcome. Because not all the variables were available at each visit for each patient, the numbers of observations and the numbers of patients that contributed the observations differed for different models. Each association of primary interest was assessed in 2 different multivariable models. The partially-adjusted estimate for prednisone-equivalent dose is adjusted for other covariates but not for SLEDAI score, and vice versa. In contrast, the fully-adjusted estimates are obtained from the same model, i.e., they are adjusted for each other.

Table 2. Estimated regression coefficients from linear regression models*
OutcomeNo. persons/ person- momentsAverage prednisone-equivalent dose in past year (increase by 10 mg/day)Average SLEDAI score in past year (increase by 6 points)
Partially adjusted (95% CI)Fully adjusted (95% CI)Partially adjusted (95% CI)Fully adjusted (95% CI)
  • *

    SLEDAI = Systemic Lupus Erythematosus Disease Activity Index; 95% CI = 95% confidence interval; HDL = high-density lipoprotein; LDL = low-density lipoprotein; Apo B = apolipoprotein B; BP = blood pressure; BMI = body mass index; CHD = coronary heart disease.

  • Adjusted for age, sex, use of cyclophosphamide in the past year, use of immunosuppressive medications other than cyclophosphamide in the past year, and lupus duration. See Patients and Methods for description of additional adjustments made to specific models.

  • Adjusted for the same covariates as the partially-adjusted model and for average prednisone-equivalent dose and average SLEDAI score in the past year.

  • §

    P < 0.01 (significance was set at ≤ 0.05).

  • P < 0.001 (significance was set at ≤ 0.01).

  • #

    P ≤ 0.001.

Total cholesterol, mmoles/liter199, 2,0190.47 (0.40, 0.54)#0.41 (0.34, 0.48)#0.17 (0.07, 0.27)0.06 (−0.03, 0.16)
HDL cholesterol, mmoles/liter149, 8210.08 (0.04, 0.13)#0.08 (0.04, 0.12)#−0.04 (−0.10, 0.01)−0.06 (−0.12, −0.01)§
LDL cholesterol, mmoles/liter148, 8000.23 (0.15, 0.31)#0.22 (0.14, 0.31)#0.07 (−0.03, 0.18)0.03 (−0.07, 0.14)
Apo B, mmoles/liter123, 4150.09 (0.04, 0.13)0.06 (0.01, 0.11)§0.14 (0.08, 0.20)#0.12 (0.06, 0.18)#
Ln-triglycerides, mmoles/liter173, 1,2170.16 (0.11, 0.20)#0.15 (0.11, 0.19)#0.14 (0.08, 0.20)#0.13 (0.07, 0.18)#
Systolic BP, mm Hg214, 4,9092.3 (1.6, 3.0)#1.6 (0.9, 2.3)#3.9 (3.0, 4.7)#3.4 (2.5, 4.3)#
BMI, kg/m2155, 1,1030.4 (0.1, 0.7)0.4 (0.1, 0.7)−0.1 (−0.4, 0.2)−0.2 (−0.5, 0.2)
Ln-blood glucose, mmoles/liter192, 3,0690.03 (0.01, 0.04)0.02 (0.01, 0.03)0.04 (0.02, 0.06)#0.04 (0.02, 0.05)#
Ln–2-year CHD risk, %177, 1,4350.16 (0.13, 0.19)#0.15 (0.12, 0.18)#0.09 (0.04, 0.13)#0.05 (0.01, 0.09)§

The estimate for prednisone-equivalent dose represents the estimated change in the outcome associated with a 10-mg increase in the mean daily dose. Even after adjustment for SLE activity and other potential confounders, a 10-mg increase in the average daily prednisone-equivalent dose in the preceding year was associated with statistically significant increases of 0.41 mmoles/liter in total serum cholesterol, 0.08 mmoles/liter in HDL cholesterol, 0.22 mmoles/liter in LDL cholesterol, 0.06 mmoles/liter in Apo B, 0.15 mmoles/liter in ln-triglycerides, 1.6 mm Hg in systolic BP, 0.4 kg/m2 in BMI, and 0.02 mmoles/liter in ln-blood glucose level, as well as a 16% increase in the estimated 2-year CHD risk.

The estimate for disease activity corresponds to the estimated change in the outcome associated with an increase in the average SLEDAI score of 6 points in the preceding year, considered a minimum clinically meaningful increase (30). A 6-point increase in average SLEDAI score in the past year was associated with statistically significant increases of 0.13 mmoles/liter in ln-triglycerides, 3.4 mm Hg in systolic BP, 0.12 mmoles/liter in Apo B, and 0.04 mmoles/liter in ln-blood glucose level, and with a decrease of 0.06 mmoles/liter in HDL cholesterol, even when adjusted for corticosteroid use and the other covariates. The fully-adjusted effect of SLE activity on total serum cholesterol was statistically nonsignificant (Table 2), and SLE activity did not seem to affect LDL cholesterol or BMI (P > 0.3 for both in Table 2). Still, a 6-point increase in average past-year SLEDAI score was associated with a statistically significant 5% increase in the probability of CHD in the next 2 years. Additional adjustment for the updated SDI score of damage did not materially change any of the results (data not shown).

Additional analyses revealed statistically very significant interactions between the average daily corticosteroid dose and the average SLEDAI score for 3 outcome measures: total cholesterol (P < 0.0001), systolic BP (P < 0.0001), and 2-year CHD risk (P = 0.0015). In each case, the estimated interaction coefficient indicated that the impact of increasing the average corticosteroid dose on increasing the level of a respective risk factor was stronger for patients with higher SLE disease activity in the past year (data not shown).

In the sensitivity analyses with alternative approaches to corticosteroid dose and SLEDAI score reconstruction (see Patients and Methods), regression coefficients, estimated in fully-adjusted models, were generally similar to those in Table 2 (data not shown). With respect to statistical significance, in the sensitivity analyses the effect of the average SLEDAI score in the past year became significant (P = 0.033), while its effects on HDL cholesterol and ln-blood glucose, and the effect of a prednisone-equivalent dose on BMI, all became nonsignificant (P = 0.1801, 0.0595, and 0.0905, respectively).


We have shown that even after adjusting for lupus disease activity, disease duration, history of cardiovascular medication use, age, and sex, higher corticosteroid dose in the past year was associated with significantly higher levels of total serum cholesterol, systolic BP, HDL cholesterol, LDL cholesterol, Apo B, triglycerides, BMI, and blood glucose in lupus patients. Similarly, even after adjustment for corticosteroid dose and the other potential confounders, higher recent lupus disease activity was independently associated with higher levels of systolic BP, triglycerides, blood glucose, and Apo B. Overall, a 10-mg increase in daily corticosteroid dose was associated with an ∼16% increase in the estimated risk of a CHD event in the next 2 years, whereas a 6-point increase in SLEDAI score was associated with a 5% risk increase. Additional analyses suggested that the impact of increased corticosteroid dose on increasing total serum cholesterol and systolic BP levels becomes significantly stronger among patients with higher recent SLE disease activity. While these interactions may reflect true biologic synergistic effects between SLE activity and corticosteroid use, further research is necessary to both confirm such interactions and explore the underlying reasons for them.

Some previous studies reported positive associations between corticosteroid use and traditional coronary risk factors, such as total plasma cholesterol (7–9, 12, 13, 17, 31), systolic BP (7), triglycerides (12–14, 17, 32, 33), LDL cholesterol (13, 14, 31), Apo B (10), and hyperinsulinemia (33). However, most of these studies did not adjust for disease activity, which in our cohort was correlated both with corticosteroid dose (r = 0.32) and with most CHD risk factors, and therefore may act as a confounder. In addition, in our longitudinal data, we adjusted for prior use of antihypertensive, blood-glucose–lowering, lipid-lowering, and immunosuppressive agents, which could affect the observed risk factor values. Therefore, our results are less subject to potential confounding bias.

It has been suggested that corticosteroids might decrease BMI through their promotion of protein catabolism, resulting in a reduction in fat-free mass, whereas corticosteroid-induced redistribution of fat would not be accompanied by changes in body fat mass (34). Kipen et al (35) reported negative associations between cumulative corticosteroid dose and both fat mass and fat-free mass in women with lupus. In contrast, our results are consistent with Petri et al (7), who documented a positive association between prednisone dose and BMI.

Another important finding of our study was that of the positive independent association between corticosteroid dose and HDL cholesterol in patients with lupus, which is consistent with a finding by Ilowite et al (17) in pediatric lupus patients and with previous reports in nonlupus populations (8, 12, 36). However, as demonstrated by our finding of a statistically significant increase in the aggregate CHD risk, the overall effect of corticosteroid medication is mostly atherogenic. Based on our results, the estimated 2-year coronary risk for a subject who has received an average prednisone-equivalent dosage of 30 mg/day for 1 year, corresponding to the typical daily dose for management of moderately active lupus (37), is approximately 60% higher than it would be for a patient with the same levels of SLE activity and the same characteristics for other covariates who received no corticosteroids.

There are several pathophysiologic pathways through which lupus disease activity could lead to changes in the traditional coronary risk factors. Systemic inflammation and anti-insulin antibodies could promote the development of insulin resistance (38, 39), with the ensuing dyslipidemia, hypertension, and increase in body weight. Also, endothelial damage and dysfunction, induced by lupus systemic inflammation, can be accompanied by overproduction of the vasoconstrictors thromboxane and endothelin 1, leading to elevation of BP (40). The endothelium may also be the target of specific autoantibodies or may be affected by the action of antiphospholipid antibodies, leading to acute thrombosis in an already abnormal vessel or to a proatherogenic profile. In addition, blood lipid profiles in lupus patients could be rendered atherogenic due to suppression of lipoprotein lipase activity, which is possibly reduced in people with SLE (16), as a result of increased levels of interleukin-1 and interferon-γ (41, 42) and/or presence of antilipoprotein lipase antibodies (43). Finally, tumor necrosis factor α, which has been proposed as a biomarker for SLE activity (44), is a possible promoter of hyperlipidemia (45), insulin resistance (46), hypertriglyceridemia, low levels of HDL cholesterol, and increased central adiposity (47).

A recent review suggested that “… aggressive treatment of the inflammatory process may reduce the risk of cardiovascular disease in patients with SLE” (48). However, we have also found that increased prednisone dose is associated with statistically significant increases in several CHD risk factors, independent of SLE activity. These results should urge us to carefully design alternative antiinflammatory steroid-sparing regimens (37). However, although corticosteroid treatment seems to increase the values of several CHD risk factors, it also obviously helps control lupus disease activity, which in turn may reduce risk factor values. Therefore, future research should attempt to quantify the net balance of these 2 effects, which could require estimating the impact of corticosteroid dose on the subsequent change in SLE activity. Our study provides the estimates of the 2 other pieces of this complex puzzle, i.e., the independent effects of corticosteroid dose and SLE activity on changes in specific risk factors.

Our results suggest potentially important atherogenic roles for both corticosteroid medications and lupus disease activity. Thus, our results are consistent with Doria et al, who studied the effect of several traditional and nontraditional risk factors for atherosclerosis in a prospective cohort of 78 patients with SLE, finding a positive association between cumulative prednisone dose and the presence of carotid plaque (49). Cumulative prednisone dose was also positively associated with the presence of thickened intima, but this association was not statistically significant, possibly due to the low statistical power. In contrast, Roman et al (1) reported associations of the presence of carotid plaque with a lower 5-year daily dose of prednisone and a lower SLEDAI score. In another recent report (2), no associations were found between either SLEDAI score or cumulative prednisone dose and coronary atherosclerosis, as indicated by computed tomography. However, in the former study (1), no adjustment was performed for any important covariates, while in the latter (2), only age and sex were adjusted for. Therefore, both findings could be affected by confounding bias.

Our study also has limitations. First, we relied on pre-existing medical records in which many relevant values were not documented. However, our algorithm for imputing missing corticosteroid doses was based on the steroid tapering regimens commonly used in the Montreal General Hospital Lupus Clinic during the study period. Furthermore, in the sensitivity analyses, in which we changed the underlying assumptions, the conclusions regarding associations between corticosteroid dose and particular risk factors did not change materially. Future research would be facilitated if all dose changes of relevant medications were systematically recorded. Similarly, retrospective SLEDAI scoring could also be subject to some measurement error, but random measurement errors in corticosteroid dose and/or SLEDAI score could only dilute the putative associations and make them less significant (50). By relying on a longitudinal database with 4,094 person-years of followup and on appropriate statistical methods, we were able to detect many statistically significant associations.

Second, we were not able to adjust for dietary factors, which could affect the levels of the coronary risk factors of interest. However, we do not believe that the dietary patterns would correlate with corticosteroid medication doses and/or lupus disease activity levels to the extent that could have caused any substantial confounding. Furthermore, our database did not reflect potential changes in patient management that may have occurred over the study period. However, because we focused on rather short-term associations of recent corticosteroid dose and SLE activity with coronary risk factors, any long-term practice changes were unlikely to materially affect our results.

Finally, with respect to measurement of disease activity, the SLEDAI instrument, similar to other available SLE activity measures, aggregates recent disease activity across different organs and systems (23). Future research should investigate whether different subcomponents of SLEDAI have differential impact on CHD risk factors, but such studies will require very detailed clinical data and a large sample size.

In conclusion, our study provides evidence that in patients with SLE, both recent corticosteroid therapy and recent lupus disease activity play a role in CHD development by affecting several traditional CHD risk factors. Future studies should investigate direct associations between SLE-specific factors and actual CHD events, as well as other possible pathways leading to CHD in lupus patients (51, 52).


Dr. Abrahamowicz 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 design. Karp, Abrahamowicz, Fortin, Esdaile.

Acquisition of data. Karp, Fortin, Neville, Esdaile.

Analysis and interpretation of data. Karp, Abrahamowicz, Pilote, Pineau, Esdaile.

Manuscript preparation. Karp, Abrahamowicz, Pilote, Pineau, Esdaile.

Statistical analysis. Karp.


VariableRegression coefficient
  • *

    From the Framingham Heart Study population (1948–1978).

Age, years0.0540
Serum cholesterol, mmoles/liter0.2431
Systolic blood pressure, mm Hg0.0149
Current smoking, cigarettes/day0.0142
Glucose intolerance, 1 if yes, 0 if no0.3944
Body mass index, kg/m20.0374
Sex, 1 if male, 0 if female0.8069