Abstract
Objective. To assess factors associated with the ability to achieve and maintain target serum urate (SU) with allopurinol in patients with gout.
Methods. We used US Veterans Affairs (VA) databases from 2002–2012. Eligible patients had ≥ 1 inpatient or ≥ 2 outpatient visits with a diagnostic code for gout, filled a new index allopurinol prescription, had at least 1 posttreatment SU level measured, and met the 12-month observability rule. Treatment successes were defined as the achievement of postindex SU < 6 mg/dl (success 1) and postindex SU < 6 mg/dl that was sustained (success 2).
Results. Of the 198,839 unique patients with allopurinol use, 41,153 unique patients (with 47,072 episodes) and 17,402 unique patients (with 18,323 episodes) were eligible for analyses for success 1 and success 2; 42% each achieved (success 1) or achieved and maintained postindex SU < 6 mg/dl (success 2). In multivariable-adjusted models, factors associated with significantly higher odds of both outcomes were older age, normal body mass index, Deyo-Charlson index score of 0, rheumatologist as the main provider rather than non-rheumatologist, midwestern US location for the healthcare facility, a lower hospital bed size, military service connection for medical conditions of 50% or more (a measure of healthcare access priority), longer distance to the nearest VA facility, and lower preindex SU.
Conclusion. We identified novel factors associated with maintaining SU < 6 mg/dl based on a theoretical model. Several potentially modifiable factors can be targeted by individual/provider/systems interventions for improving successful achievement and maintenance of target SU in patients with gout.
Failure to achieve and maintain target serum urate (SU) is a critical shortcoming of current gout management1,2,3. Fewer than 50% of patients treated with allopurinol, an effective and inexpensive urate-lowering therapy (ULT), achieve target SU < 6 mg/dl1,4. Maintenance of target SU is associated with lower risk of gout flares and tophi and lower healthcare costs4–10 and is recommended by every treatment guideline11–17.
A single US Veterans Affairs (VA) center study (n = 643; 253 with SU; 39%) reported that a lower medical comorbidity load was associated with higher odds of reaching the target SU < 6 mg/dl1. In claims database studies, factors associated with higher likelihood of achieving target SU < 6 mg/dl were older age, female sex, higher allopurinol dose, and the absence of kidney disease (n = 3363; 2059 had SU; 61%)18; higher allopurinol adherence (n = 18,243; 4277 with SU; 23%)19; and female sex, older age, white race, rheumatologist care, and higher allopurinol start dose and adherence (n = 9581 incident users)20. These studies had important limitations. All studies except 21,20, used prevalent user design, which tends to bias estimates and overestimate adherence. None examined factors associated with maintaining target SU, the main goal recommended by gout guidelines11,21,22. Most studies examined demographic and clinical characteristics (i.e., predisposing factors), but none evaluated gout severity (i.e., need factors), or important system-level or healthcare access factors, such as region, rural location, distance to the medical center, etc. (i.e., enabling factors)1,18,19,20. Therefore, significant knowledge gaps remain. No conceptual model was invoked in any of these studies. Andersen’s Behavioral Model of need, enabling, and predisposing factors23,24 offers a potential solution to improve our understanding of associated factors.
We examined the data from the VA healthcare system25,26, the largest integrated healthcare system in the United States. It provides care to 6 million participants annually26. We hypothesized that in patients with gout who are taking allopurinol, the needs, enabling, and predisposing factors based on Andersen’s model23,24 would be associated with a patient’s ability to achieve and maintain target SU < 6 mg/dl; and the initial allopurinol dose and previous allopurinol use in the baseline year would also be associated independently with these outcomes.
MATERIALS AND METHODS
Study cohort and eligibility and data sources
We used the VA national databases from 2002 to 201227,28,29, which are reliable for demographics and most common diagnoses30 and valid for specific diagnoses31. The Institutional Review Board (IRB) at the University of Alabama at Birmingham (X120928002) and the Birmingham VA Medical Center (01487) approved the study. The IRB waived the need for informed consent for this database study. We followed the Strengthening of Reporting in Observational studies in Epidemiology (STROBE) guidelines32.
Patients were eligible if they had ≥ 1 inpatient or ≥ 2 outpatient visits with an International Classification of Diseases, 9th ed (ICD-9) code 274.x for gout, were treated with allopurinol, had a postindex SU level taken, and met the 12-month observability period (i.e., for each 12 months, there must be an ICD-9 code 274.x recorded in the system). Index allopurinol prescription was defined as no allopurinol exposure in the previous 121 days. This included a 91-day clearance period and a 30-day grace period between prescriptions, because patients often have a small stock of medication, especially with the 90-day prescriptions, the most common day supply at the VA. A gap of > 30 days between any 2 allopurinol prescriptions was considered the end of an episode and led to the beginning of another drug exposure period.
Patient demographic and comorbidity data were obtained from the VA Patient Treatment File and Outpatient Clinic tables. Results of SU tests were obtained from each Veterans Health Information Systems and Technology Architecture system accessed, using VA Informatics and Computing Infrastructure (VINCI)33. Medication data were obtained through the Decision Support System’s Pharmacy National Data Extract, which contains records for all inpatient and outpatient prescriptions, including every medication filled and refilled from all VA facilities (number of days’ supply, dose, number of pills, and the start and end date for medication filling and refills). Provider factors were obtained from the MedSAS Outpatient provider data. Systems factors [VA location, bed size, community-based outpatient clinic (CBOC) vs VA clinic, etc.] were obtained from the VA planning systems support group office and VINCI.
Outcome
We examined 2 key outcomes of success: (1) achieving target SU < 6 mg/dl postindex allopurinol prescription [(i.e., during the followup at any time 14 days or after the index allopurinol prescription (success 1)]; and (2) maintaining target SU < 6 mg/dl [i.e., meeting the previous definition and having all subsequent (≥ 2) SU levels < 6 mg/dl postindex prescription during the followup, with at least 1 day between laboratory assessments (success 2)].
Associated factors: covariates and potential confounders
We examined the following factors, as they mapped to the Andersen’s Model23,24.
Predisposing factors included these patient characteristics: age (in yrs), sex, race/ethnicity (white, Hispanic, black or African American, other, and unknown), body mass index, and marital status (single, married, divorced, widowed, and unknown). Comorbidity was assessed using the Deyo-Charlson index34, a validated measure consisting of 17 comorbidities, examined as summation score and categorized as 0, 1, or 2 or more comorbidities.
Enabling factors were provider factors [i.e., the main provider of gout care categorized by provider specialty as rheumatologists versus other (non-rheumatologist; including primary care)].
System factors were the location of the VA facility (rural vs urban, affiliation with a teaching hospital, yes or no), outpatient clinic type (CBOC vs VA medical center vs both vs other), VA facility bed size (categorized into ≤ 50, 51–100, 101–200, and > 200), and region (mid-Atlantic, Midwest, Northeast, South, and West).
Healthcare access factors included the distance to the nearest VA medical center as a measure of accessibility, an important predictor of outcomes35,36. It was calculated as the straight-line miles from the centroid of the patient’s residential postal code to the nearest VA site, as previously. Military service connection and means test were also included, because both were significant predictors of SU monitoring in our previous VA study37. Military service connection is an indicator of access to care. It ranges from 0% to 100% and is awarded for conditions that begin during or result from active military duty38. Veterans with ≥ 50% service connection do not have co-payments for medical care or prescriptions, and get priority in VA healthcare access. Means test measures household income and assets and is completed yearly by most veterans39. Categories are AN (most needy but not service-connected), AS (most needy and service connected), and C (not most needy).
Regarding medication factors: allopurinol start dose was calculated as unit dose × quantity divided by days’ supply based on the first and last filled prescription and categorized as ≤ 100, 101–200, 201–300, and > 300 mg/day. We also examined any previous allopurinol use in the 1-year baseline.
Need factors were disease severity factors: duration of gout, assessed as the time from meeting the definition (1 inpatient ICD-9 code or 2 or more outpatient ICD-9 codes) to the beginning of the index allopurinol prescription; and preindex SU level, categorized as < 6, 6 to < 8, 8 to < 10, 10 to < 12, and ≥ 12 mg/dl.
Statistical analyses
We compared the characteristics of patients who did or did not receive a post-allopurinol SU test as well as did or did not achieve success 1 or 2, using chi-square/comparison of proportions test or t test as applicable. We used multivariable-adjusted logistic regression models to assess whether needs, enabling, and predisposing factors were associated with the ability to achieve target SU < 6 mg/dl (Model 1a) and to achieve and maintain SU < 6 mg/dl in gout patients taking allopurinol (Model 1b). We reported OR and 95% CI. We performed sensitivity analyses by (1) replacing Deyo-Charlson individual comorbidities in the main model with a score (0, 1, ≥ 2; model 2a, 2b); and (2) additionally adjusting the main model for allopurinol use in the baseline 1-year and the starting allopurinol dose (model 3a, 3b). We performed exploratory analyses by additionally adjusting the main model for allopurinol variables including the start and the end dose, use in the baseline 1-year, dose escalation (normal, fast, slow, none), and medication possession ratio (MPR; model 4a, 4b), calculated as the medication supply actually received by the patient divided by medication supply that could have been received.
RESULTS
Cohort characteristics
Of the 627,693 patients with gout, 198,839 patients (310,695 episodes) had a new allopurinol prescription and at least 12 months of observability (Figure 1). Of these, 41,153 patients (47,072 episodes) and 17,402 patients (18,323 episodes) contributed to the analyses for achieving target SU (success 1) or maintaining target SU (success 2), respectively.
Cohort selection flowchart. The flowchart shows the selection of a sample for the study, based on eligibility criteria. We examined 41,153 people with gout for achievement of target serum urate (success 1) and 17,402 people with gout for achievement and maintenance of target serum urate (success 2).
The study cohort for achieving target SU (success 1) had a mean age of 66.8 years, body mass index (BMI) of 33.6 kg/m2, were male (99%), and white (61%; Table 1). Comorbidities were common and were higher compared to patients who did not get a postindex prescription SU test (Table 2). Characteristics were similar for the 2 study cohorts for achieving target SU (success 1; Table 3) or maintaining target SU (success 2; Table 4).
Selected characteristics of overall sample and the study cohort.
Comorbidity characteristics of overall sample and the study cohort.
Unadjusted and adjusted analyses of patients with gout for achievement of target SU of < 6 mg/dl.
Unadjusted and adjusted analyses of patients with gout for achievement and maintenance of target SU of < 6 mg/dl.
Unadjusted characteristics of patients achieving or maintaining target SU
For the success 1 cohort, the mean followup duration was 784.8 days (∼26 mos; SD 810.3 days). The mean time to achieving target SU was 273.3 days (∼9 mos; SD 303.5 days) and mean allopurinol dose was 193.5 mg/day (SD 104 mg/day; Table 1). Only 42% patients each achieved target SU (success 1: 17,284/41,153 patients; 19,535 episodes) or achieved and maintained target SU (success 2: 7,309/17,402 patients; 18,323 episodes). Unadjusted characteristics are shown in Tables 3 and 4. Mean (SD) preindex SU was lower in patients reaching versus not reaching target SU during followup: 7.8 (2.2) versus 8.9 (2.0) mg/dl for achieving target SU (success 1); and 8.1 (2.2) mg/dl versus 9.2 (2.1) mg/dl for achieving and maintaining target SU (success 2).
Multivariable-adjusted correlates of achieving or maintaining target SU
Factors associated with significantly higher odds of achieving target SU < 6 mg/dl (success 1) were older age, male sex, white race, rheumatologist as the main provider of gout care, a lower hospital bed size of ≤ 50 or 101–200 (compared to > 200), Midwest location for the healthcare facility, and the presence of comorbidities (rheumatologic disease, peptic ulcer disease, diabetes with complications and severe liver disease; Table 3). Medical comorbidities (heart disease, diabetes, and renal disease), preindex SU higher than 8 mg/dl (vs 6 to < 8), and longer gout duration were significantly associated with lower odds of achieving target SU (Table 3).
Factors associated with significantly higher odds of maintaining target SU < 6 mg/dl (success 2; at least 2 SU levels at target postindex prescription) were white race, rheumatologist as the main provider, a lower hospital bed size such as 101–200 (ref > 200 beds), and a normal BMI (Table 4). These were associated with lower odds: medical comorbidities (heart disease, mild liver disease, diabetes, renal disease, malignancy and malignant neoplasm without specification of site), being single, a southern US location for the healthcare facility (vs Midwest), and a preindex SU level higher than 8 mg/dl (vs 6 to < 8).
In sensitivity analysis (model 2), a higher Deyo-Charlson comorbidity index score ≥ 2 was associated with lower likelihood of achieving or maintaining target SU (Figure 2).
Association of Deyo-Charlson comorbidity index score with achieving or maintaining target serum urate in a multivariable-adjusted model. A. Association of Deyo-Charlson comorbidity index score with the odds of achieving target serum urate. B. Association of Deyo-Charlson comorbidity index score with the odds of achieving and maintaining of target serum urate. Reference category is Deyo-Charlson comorbidity index score of zero (no comorbidity). OR estimates are shown as diamonds and 95% CI are shown as whiskers. Estimates with 95% CI overlapping the OR of 1 (solid line) are statistically not significant. For example, compared to a score of 0, a Deyo-Charlson comorbidity index score of 2 or more was significantly independently associated with lower odds of achieving or maintaining target serum urate with OR (95% CI) of 0.82 (0.79–0.86) and 0.63 (0.59–0.68), respectively. The results are from model 2, in which the individual Deyo-Charlson comorbidities in the main model (model 1) were replaced with a Deyo-Charlson index score of 0, 1, or ≥ 2.
Analyses of effect modification including allopurinol dose and use variables, overall and in explaining racial differences in achieving or maintaining target SU
In multivariable-adjusted models adjusted additionally for previous allopurinol use and start dose, compared to start dose of ≤ 100 mg/day, higher allopurinol doses were associated with higher odds and allopurinol use in the baseline 1-year with lower odds of achieving or maintaining target SU (Supplementary Tables 1 and 2, available with the online version of this article). In exploratory analyses, normal or fast allopurinol escalation (compared to no escalation), higher allopurinol end dose, and higher allopurinol medication possession ratio were associated with higher odds of both outcomes, with higher allopurinol start dose being only borderline significant (Supplementary Table 3, available with the online version of this article).
Although in the main analysis African Americans had significantly lower adjusted odds of achieving target SU (OR 0.94, 95% CI 0.89–0.99) and similar odds of maintaining target SU (OR 1.02, 95% CI 0.93–1.11), after adjusting for the rate of allopurinol dose escalation and MPR they had significantly higher odds of achieving target SU (OR 1.16, 95% CI 1.09–1.23) and maintaining target SU (OR 1.22, 95% CI 1.11–1.34; Supplementary Table 3, available with the online version of this article). Further analyses revealed that this effect modification was due to allopurinol MPR.
DISCUSSION
To our knowledge, no study to date has examined factors associated with maintaining target SU. Our comprehensive, national cohort study performed robust analyses that controlled for patient, provider, systems, medication, and disease severity factors and advances the understanding of factors associated with target SU. Patients who saw rheumatologists (< 3% of patients) as the main providers for gout care were more likely to achieve target SU and to maintain target SU. A better quality of gout care with a rheumatology provider1 and higher odds of target SU achievement with a rheumatologist provider20 explain our findings. This may be due to the prioritization of gout management during a rheumatology visit. This finding has potential policy implications for the VA.
We recognize that a multifaceted approach with several policy initiatives is required to address this quality gap. Expanded rheumatology care teams (nurse practitioners, physician assistants, community health workers), and technology-based solutions (tele-health, e-consults, and virtual health communities) may address this problem. This would require provision of more resources for the VA rheumatology workforce. Nurse- or pharmacist-led interventions are effective in improving gout care and outcomes40,41,42. Maintaining target SU < 6 mg/dl is associated with reduction in gout flares, resolution of gouty tophi, and improvement in quality of life and function5,10 and is key to optimal management11,21,22.
It took a mean of 9 months to achieve target SU. This is longer than might be expected based on the American College of Rheumatology treatment guideline with frequent ULT dose titration11. Only 42% of the patients maintained target SU, a key treatment goal recommended by gout guidelines11–17. Allopurinol is well-tolerated with few adverse events, which does not explain this low rate of success. Patients prioritized lowering of SU (to target) in gout as an important goal43, challenging the recent American College of Physicians position to consider treat to symptom control in gout with no SU monitoring as an alternative to treat to target SU17. Thus, improvement in rates of target SU achievement and maintenance in gout are needed.
African Americans had lower odds of achieving target SU in models not accounting for allopurinol adherence and gout severity factors, but higher odds of achieving or maintaining target SU, in exploratory analyses. This is a novel insight. It indicates that the lower chance of target SU achievement in African Americans can be explained largely by lower allopurinol adherence, and to a lesser degree by worse baseline disease, lower allopurinol start dose, and improper dose escalation. After accounting for these confounding factors, African Americans with gout have a better SU outcome than whites (Supplementary Table 3, available with the online version of this article). Addressing these modifiable factors has the potential of improving gout outcomes in African Americans and reducing health disparities in gout.
Prior studies lacked a theoretical model and were limited because of inclusion of a select few variables or univariate analysis1,18,19,20, or they were single-center1 or a regional sample20, lacked gout disease severity variables1,18,19,20, or used a prevalent allopurinol user design (and not incident user design)18,19. Using the Andersen model23,24, we specified a priori that system, healthcare access, and disease severity factors (i.e., enabling and predisposing factors, not included in previous studies) in addition to patient and provider factors (need factors) will be associated with target SU outcomes. We found higher odds of target SU achievement in males compared to females in contrast to previous studies18,20. Several things may explain these differences: differences in setting (national vs regional sample), population (veterans vs US population), design (incident vs prevalent user design)18, control for disease severity (disease duration and SU vs SU20, or neither18), and additional covariate adjustment (system and healthcare access vs neither18,20).
Our study highlights the role of comorbidities in the achievement of target SU with allopurinol. The presence of 2 or more comorbidities was independently associated with 20% lower odds of maintaining target SU < 6 mg/dl compared to no comorbidities. A linear dose gradient was seen with increasing comorbidity load. This novel finding adds to the current knowledge. We found that heart disease, renal disease, and diabetes were associated with lower odds of achieving or maintaining target SU < 6 mg/dl. Our national sample, controlled using the Deyo-Charlson index, a validated comorbidity index, confirmed previous findings from regional/single-center studies controlled for selected conditions1,18,19,20. Although polypharmacy might be associated with higher allopurinol adherence20, damage from higher comorbidity or specific medications on SU levels might make it more challenging to achieve target SU in people with specific comorbidities or an overall high comorbidity load, as our study and other studies show1,18,20,44,45. Future studies should examine whether optimization of comorbidity management can improve gout outcomes.
Geographic region was associated with odds of achieving and maintaining target SU. Compared to the Midwest, patients living in the South were less likely to achieve or to achieve and maintain target SU. Differences in resources, regional economies, and/or patient populations may underlie these differences. Policy makers may need to provide additional VA resources in these US regions to improve gout outcomes. This finding indicates that findings from even well-designed studies limited to 1 US region20 are unlikely to be generalizable to the entire United States. We assessed allopurinol start dose and previous allopurinol exposure as potential confounders of achieving or maintaining target SU rather than mediators, because they meet the classic definition of confounders by being associated both with the covariates and the outcome. For example, while adjusting for allopurinol MPR clearly changed the apparent relationship between race and achieving SU targets, it seems more reasonable to infer that adjusting for MPR accounted for the confounding between race and medication adherence than the mediation interpretation that the effect of race on SU adherence “acted through” medication adherence.
Higher allopurinol start doses > 100 mg/day increased the odds of achieving or maintaining target SU by 1.8–3.5–fold, slightly higher than the 1.7–fold found in a previous study20. Higher allopurinol end dose, higher allopurinol MPR, and normal or fast allopurinol escalation were also associated with very high odds of achieving or maintaining target SU. This is not surprising because allopurinol is an effective ULT. Most treatment guidelines recommend a low allopurinol starting dose of 100 mg/day and a gradual allopurinol dose escalation, to avoid gout flares and rare allopurinol hypersensitivity. Thus, the pros and the cons of high versus low allopurinol start dose and fast vs normal/slow allopurinol dose escalation must be considered and discussed with an individual patient in a shared decision-making approach. A high allopurinol MPR and higher end dose (i.e., to adequately lower SU) are noncontroversial approaches in gout care.
The study findings must be interpreted considering limitations. The use of ICD-9 codes for the diagnosis of gout and other comorbidities using the VA databases is subject to misclassification bias. However, ICD-9 codes for gout in the VA had high accuracy with sensitivity of 90% and specificity of 100%37 and VA databases are valid and reliable for several diagnoses30,31, although other database studies have reported lower accuracy for gout code46. Findings from this predominantly male veteran population may not be generalizable to women and non-veterans. However, no previous studies have found sex or veteran status as confounders of these associations and the male-predominant VA population made it an excellent clinical laboratory to study gout, also a male-predominant condition. As in any other US healthcare system (Medicare, Medicaid, etc.), care provided outside the system could not be accounted for in these analyses and may have affected precision. The effect of this missingness on our results is unclear. Our findings are likely not generalizable to other healthcare systems, because systems-level factors can vary. However, identified factors can now be targeted to improve these outcomes nationally for the large VA healthcare system, at a minimum.
Study strengths include the use of an integrated national VA database, a large sample size, an incident allopurinol user design, the use of a theoretical model, inclusion of several important covariates previously not included, examination of several models to test the robustness of findings, and examination of factors associated with maintaining target SU that were not reported previously.
We conducted a comprehensive national study of factors associated with achieving and maintaining target SU < 6 mg/dl in patients with gout, using data from one of the largest integrated national US healthcare systems. We identified key factors independently associated with achieving and maintaining target SU < 6 mg/dl: those of the patient, comorbidity, physician, system, healthcare access, allopurinol dose/adherence, and disease severity. Several characteristics may be amenable to patient/physician/systems–targeted interventions to improve the chances of maintaining target SU < 6 mg/dl in gout in this large national healthcare system. This, in turn, can improve gout outcomes and reduce the patient morbidity and the societal effect of gout.
Acknowledgment
We thank Michael Conner, BS, at Birmingham VA Medical Center for assisting in obtaining and programming the VA data for these analyses, and Jeffrey Foster at the University of Alabama at Birmingham for assistance in drafting tables and figures and proofing this paper for errors. We thank Dr. Joshua Richman at Birmingham VA Medical Center for his advice regarding the statistical analyses.
Footnotes
This study was funded by the US National Institute of Arthritis, Musculoskeletal and Skin Diseases P50 AR060772 grant that also support efforts for Dr. Singh (Project PI) and Dr. Saag (main PI). The authors are ready to share the data with colleagues, after obtaining appropriate permissions from institutional review boards at the Birmingham VA Medical Center and the University of Alabama at Birmingham, related to their privacy and data-sharing policies. JAS has received consultant fees from Crealta/Horizon. JAS owns stock options in Amarin and Viking Pharmaceuticals Inc. JAS is a member of the Veterans Affairs Rheumatology Field Advisory Committee. JAS is the editor and the Director of the UAB Cochrane Musculoskeletal Group Satellite Center on Network Meta-analysis. JAS is a member of the executive of OMERACT, an organization that develops outcome measures in rheumatology and receives arms-length funding from 36 companies. He is also supported by the resources and the use of facilities at the VA Medical Center at Birmingham, Alabama, USA. KGS serves as a consultant for Amgen, Merck, and Radius and receives funding from Amgen and Merck.
- Accepted for publication August 2, 2019.
REFERENCES
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
In this issue
