Patients were participants in the University of California San Francisco (UCSF) Lupus Outcomes Study (LOS). The LOS is an ongoing longitudinal survey of 1180 patients with physician-diagnosed SLE recruited between 2002 and 2009. Subjects eligible for this study were enrolled between 2002 and 2004 and followed through 2006. The LOS introduced the Systemic Lupus Erythematosus Activity Questionnaire (SLAQ) in the second interview year (2003). Therefore, we drew data for this study from the second through fifth consecutive years of the LOS (waves 2–5). Eight hundred and eighty seven individuals were interviewed at least once during this period. On average, 94% of eligible subjects from each wave were re-interviewed in the subsequent wave. Forty six observations were excluded because they were missing key variables.
Patients were originally recruited from academic rheumatology offices (23%), community rheumatology offices (11%), and other community-based sources (66%), such as SLE support groups, the Internet, and media advertisements. Two-thirds of the participants were residents of California while the remainder resided in 40 other US states. SLE diagnoses for all LOS participants were confirmed to meet ACR criteria by formal chart review by a rheumatologist, or trained nurse or research assistant supervised by a rheumatologist prior to entry into the cohort. All participants provided informed consent prior to the interviews. The UCSF Committee on Human Research approved the study protocol.
Data were collected via an annual, structured, 1-hour telephone survey conducted by trained interviewers. The survey included validated items covering the following domains: demographics and socioeconomic status, status of SLE, disability, general health and social functioning, employment, psychological and cognitive status, health care utilization, medications, and health insurance coverage.
Socioeconomic and demographic characteristics included age at interview (in years), sex, self-reported race/ethnicity (Caucasian, Hispanic, Asian, African-American, or other/multiple), education level (less than high school, high school graduate, some college/no degree, associate degree/trade or vocational school, college graduate, advanced or professional degree), and income (annual household income at or below vs. above 125% of the federal poverty threshold for the year prior to interview).
SLE-specific variables included duration of disease (in years), self-reported SLE flare within the 3 months prior to the interview, and disease activity over the 3 months prior to the interview as measured by the SLAQ, a validated self-report measure of SLE activity.17,18
Patients were also queried regarding any episodes of organ involvement since the time of the last interview, including specific questions about incident proteinuria or other kidney problems, hemoptysis or other lung problems, deep vein thrombosis or pulmonary embolism, and stroke or myocardial infarction.
Health care utilization
The health care utilization section of the questionnaire asked participants about their medical care over the past 12 months. It included an enumeration of all health care practitioner visits by specialty. Participants were asked directly about who serves as their “main SLE doctor.” Internal medicine and family physicians were considered generalists. We created a measure of pill burden by summing the number of reported daily medications for each patient (range 0–15). In the logistic regression model, HCQ use was subtracted from this total.
The insurance section of the questionnaire, derived from the Medical Expenditures Panel Survey, included items regarding the type of health plan (health maintenance organization versus fee-for-service) and source of coverage, if any (employment based, individually purchased plan, Medicare, or Medicaid).19
Hydroxychloroquine and other medication use
Subjects were asked about “ever” use of multiple drugs upon entry into the study (wave 1). In subsequent interviews, they were asked about use of these drugs (waves 2–5) since the last interview, including corticosteroids (oral and intravenous), mycophenolate mofetil and azathioprine, cyclophosphamide, methotrexate, and biologic agents. Biologic agents were defined as any of the following: etanercept, infliximab, adalimumab, or kineret. Subjects were also asked if they used antihypertensives or anticoagulants (heparin or warfarin).
The main outcome of interest was prevalent HCQ use since the time of the last interview. Respondents were defined as users if they answered “yes” to the question, “Since the time of your last interview, have you taken plaquenil, hydroxychloroquine, or chloroquine?”
Patients were followed for between 1 and 4 interview cycles and could therefore contribute up to 4 observations to the analysis. All analyses were performed by comparing the HCQ user person-years to the HCQ non-user person-years. In order to account for the multiple observations contributed by individuals in this analysis, demographic, clinical, and health system utilization characteristics of the cohort were compared using the Taylor series method for variance estimation in all bivariate statistical tests, as implemented in the survey analysis procedures in SAS 9.2.
We used multiple logistic regression to assess the association between HCQ use and various characteristics, several of which do not vary over time and were measured at study entry (e.g., date of birth, date of diagnosis, race/ethnicity) and several of which can vary year-to-year and were measured annually (e.g., education status, poverty status, physician type, and insurance status). Covariates in the model included age (by decade), ethnicity (Caucasian vs. non-Caucasian), income (below poverty vs. not below poverty), education level, insurance type, disease duration, disease activity score as measured by the SLAQ, total pill burden (excluding plaquenil), and main SLE physician (rheumatologist, nephrologist, generalist, or other). To account for multiple observations contributed by a single individual, we used generalized estimating equations (GEE) to determine predictors of HCQ use. SAS 9.2 (Cary, NC) was used for all analyses.
Eight hundred and eighty one individuals contributed 3095 person-years of data over 4 interview cycles in this analysis. describes the flow of patients through the 4 waves of interviews with regard to HCQ use. Wave 2 began with 804 patients, and 452 (56%) were HCQ users. Between waves 2 and 3, 59 subjects left the cohort and 68 additional subjects entered. Wave 3 contained 813 individuals, 448 (55%) of whom were HCQ users. Fortysix subjects left the cohort between waves 3 and 4, leaving 767 patients in wave 4. In wave 4, 410 (53%) of individuals used HCQ. Forty six subjects left the cohort between waves 4 and 5; 711 patients participated in wave 5. Overall, 55% of patients reported use of HCQ each year. Each year, approximately 7% of patients transitioned from one category of HCQ use to the other.
Hydroxychloroquineuse between 2004 and 2007 in the UCSF Lupus Outcomes Study
describes demographic, socioeconomic, and clinical variables by person-years of observation. The cohort was mostly female (92%) and Caucasian (69%). HCQ users were younger compared with non-users (47.9 vs. 51.1, p < 0.0001), had shorter disease duration (13.7 vs. 16.8, p < 0.0001), and were more likely to carry employer-based insurance (60.3% vs. 51.8%, p = 0.0006). HCQ users and non-users had similar SLAQ scores (12.9 vs. 12.2, p = 0.18) although a higher proportion of HCQ users reported a recent SLE flare (51.5% vs. 41.7%, p = 0.0002). Users and non-users had similar rates of organ involvement. Users were much more likely to report having a rheumatologist as their main SLE doctor and more likely to have seen a rheumatologist at least once in the past 12 months (88% vs. 66%, p<0.0001).
Demographic, socioeconomic, and clinical characteristics of 3095 person-years of observation stratified by hydroxychloroquine use in the UCSF Lupus Outcomes Study
In order to explore whether HCQ non-users were more likely to use corticosteroids or other medications, we analyzed medication use by person-year (). Sixty percent of HCQ users also used corticosteroids versus 52% of non-users (p = 0.02). HCQ users were more likely to take methotrexate compared with non-users (10.4% vs.5.4%, p = 0.0003). Other DMARD use, including the use of biologic agents, was similar between the two groups. Out of the entire cohort (3095 person-years), 69% were taking at least one DMARD (inclusive of HCQ); 13% reported using corticosteroids and no HCQ or any other DMARD. Forty percent of HCQ non-users were not taking any SLE-specific treatment (that is, they were not taking a corticosteroid or a DMARD, n = 562 person-years). This group did not differ significantly from the rest of HCQ non-users in demographics or SES. However, there was a trend towards slightly lower SLAQ scores for this group (mean, SD of 10.7, 7.6) though 36% (n = 205) reported an SLE flare within the past 3 months.
Medication use among 3095person-years of observation in the UCSF Lupus Outcomes Study
We examined predictors of HCQ use. Univariate analyses revealed that patients over 60 years old, those with longer disease duration, and those without a rheumatologist as their main SLE physician were less likely to use HCQ (). Increasing SLAQ score and increasing pill burden were associated with increased likelihood of HCQ use. In a multivariate logistic regression model adjusting for all listed covariates, HCQ use was associated with shorter disease duration. For each additional 10 years of disease duration, there was a 27% decrease in odds of using HCQ. HCQ use was less frequent among those who obtained SLE care from non-rheumatologists compared to rheumatologists with odd ratios ranging from 0.51 (95% CI 0.31–0.84) for nephrologists to 0.68 (0.56–0.83) for those reporting no SLE physician. Sensitivity analyses using (1) number of visits to a rheumatologist as a measure of subspecialty care instead of main SLE doctor and (2) SLE flare within the past 3 months instead of SLAQ score yielded nearly identical results (data not shown).
Multiple logistic regression model predicting hydroxychloroquine use among 3095person-years of observation in the UCSF Lupus Outcomes Study.