Abstract
Aims
To evaluate the effects of comorbidities and patient characteristics on treatment continuation among patients starting their first course of lipid-lowering drug (LLD) therapy.
Methods
Within the UK General Practice Research Database (GPRD), we identified 22 408 patients who started LLD therapy due to coronary heart disease, hyperlipidaemia, or other atherosclerotic diseases, and who received ≥ two prescriptions for LLD between January 1 1990 and December 31 1997. Differences in potential predictors of treatment continuation between patients who continued, and patients who discontinued/switched lipid-lowering therapy within 1 year after treatment initiation were compared by fitting multivariate logistic regression models. The effects of baseline characteristics on treatment continuation after switching of LLDs were also analysed.
Results
Discontinuation/switching of lipid-lowering therapy was common during the study period, especially among patients who received nonstatin, nonfibrate LLDs (log-rank test P = 0.0001). Statin use, more frequent physician visits, more concurrent cardiovascular medications, diabetes, and fewer noncardiovascular medications were associated with treatment continuation of LLDs. Among patients who switched therapy, prescribing of a statin as the substituted LLD, more concurrent cardiovascular medications, and later treatment switching were related to a higher probability of treatment continuation after switching LLDs.
Conclusions
Treatment continuation after initiation or switching of lipid-lowering therapy largely increased with concomitant cardiovascular comorbidities, and more health care utilization, and is more common for statins than for other LLDs. Practice guidelines, patient education, and quality of care assessment for lipid-lowering therapy should emphasize factors that predispose patients to discontinuation/switching, in an effort to optimize the choice of therapeutic regimens and to improve patient adherence.
Keywords: comorbidity, General Practice Research Database, hydroxymethylglutaryl-CoA reductase inhibitors, lipid-lowering drugs, treatment refusal
Introduction
Hyperlipidaemia is a major risk factor for coronary heart disease (CHD), and numerous clinical trials have shown the effectiveness of lipid-lowering drugs (LLDs) in reducing the risk of CHD [1–6]. As a result of these trial findings, the use of LLDs, especially 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins), has rapidly increased in recent years. Although the beneficial effects of LLDs are well-documented, these results may not be applicable to the primary care settings where quality of care may be less optimal, and where noncompliance and comorbidities both may play a role.
Several studies have examined the rate of, or factors related to, the use of LLDs in everyday practice [7–14]. All studies disclosed that the rate of discontinuation/noncompliance of LLDs in daily practice was much higher than that in the clinical trials, especially among users of nonstatin LLDs. For example, Andrade et al.[7] found that the 1-year probability of treatment discontinuation was 15% among patients treated with lovastatin, and 46% for patients treated with niacin.
Discontinuation/noncompliance of LLDs is likely to be a complex phenomenon in which the physician, the patient, various comorbidities, and the characteristics of the prescribed medications all play a role. Although previous studies have evaluated the relationship between patient characteristics and treatment discontinuation/noncompliance of lipid-lowering therapy [8–10, 13], the results are inconclusive. In addition, most studies only examined the effects of comorbidities or medication use that existed before the initiation of LLDs, which might be less relevant to treatment discontinuation compared with newly diagnosed comorbidities or concurrent medications. Finally, none of the published studies looked at factors associated with therapy switching. Because the predictors of treatment discontinuation/switching of LLDs remain poorly understood, and because these determinants may be important in the evaluation of the benefits or risks of LLDs, we decided to study these determinants in a general population-based research database.
Methods
The study was based on information derived from the General Practice Research Database (GPRD). The GPRD has previously been described in detail elsewhere [15, 16]. Briefly, the GPRD contains computerized records for some 3 million residents in the UK enrolled with selected general practitioners (GPs), who use office computers to record anonymous information on patient demographics, diagnoses, referrals, hospitalizations, and drug prescriptions. Previous studies have documented the high quality and completeness of the data recorded in the GPRD [15, 16].
Within the GPRD, we identified subjects who initiated lipid-lowering therapy by receiving a first-ever recorded prescription for a statin, a fibrate, or a nonstatin, nonfibrate LLD between 1 January 1990 and 31 December 1997. For all subjects, we started follow-up from the date on which the first-ever recorded LLD prescription was received until the patient discontinued the LLD, switched to another LLD (within or between classes of LLDs), left the practice, died, or until 1 year after initiation of LLD (whichever came first). The date on which the follow-up ended for a given subject was defined as their index date. For subjects who switched to another LLD, we followed them for the full 1 year (when possible) to determine whether they continued the substituted LLD. Discontinuation of the original LLD was identified when a patient did not receive any further LLD prescription within 3 months after the expected last date of any previous LLD prescription. Patients who continued receiving the original LLD but had a follow-up of less than 1 year due to transferral out of the practice were classified as lost to follow-up. Patients who died ≥ 1 month after the initiation of LLD were retained in the study and were categorized as treatment continuation if they did not change or discontinue the treatment before death. All subjects were required to have at least 1 year of information in the GPRD before the index date.
For patients with treatment discontinuation/switching, we collected the following information on or before the index date to evaluate the effects of baseline characteristics: patient characteristics (age, sex, body mass index [BMI], smoking status, and the year of treatment initiation); class of originally assigned LLDs (statins, fibrates, or other LLDs); number of risk factors other than serum cholesterol for CHD [17]; concurrent medications; abnormal liver or renal function within 3 months before the index date; number of GP visits (excluding visits for prescription renewal only), diagnosis of selected comorbidities (mainly cardiovascular diseases), and overall comorbidity index (modified Charlson comorbidity index [18, 19]) in the previous year; and any history of alcohol/drug abuse or certain psychiatric disorders. For patients who continued the originally assigned LLD, we collected the above information on or before the date 6 months after treatment initiation, instead of using data before the end of follow-up. For patients who switched to another LLD during the follow-up period, we further obtained data on the substituted LLD, and the timing when he/she changed the treatment. In an attempt to compare the predictive ability of using different baseline characteristics in the analysis, we also collected relevant information before starting lipid-lowering therapy.
Certain data, such as the reason for treatment discontinuation/switching could not be directly obtained through automated computer search. To derive relevant information for guiding further analyses, we manually reviewed the computer records of a random sample of 3000 patients. In the review process, it was found that 366 patients who discontinued LLD therapy had only one prescription record for any LLD. As the majority of these people (329 patients, 89.9%) did not have any recorded reason for discontinuing the treatment and it was unclear whether they really took the prescribed LLDs, we excluded all patients who received only one LLD prescription. We also did not evaluate the reason for treatment discontinuation/switching of LLDs in further analyses because the reason was frequently missing in the GPRD.
Statistical analysis
Time to treatment discontinuation/switching was plotted using Kaplan–Meier estimates for patients receiving different classes of LLDs. Log-rank test statistics were used to examine the inter-group difference in the 1-year probability of treatment continuation.
The relationships between baseline characteristics and therapeutic changes among patients who continued and patients who discontinued/switched therapy were analysed using the Pearson chi-squared test and logistic regression modelling. A stepwise multivariate logistic regression model was fitted to assess the relative risk (RR) of continuing LLD therapy that simultaneously adjusted for other baseline characteristics. Because of the large sample size and the possibility of chance variation, we excluded variables (other than patient characteristics) that had a prevalence of less than 1% or a RR ranging between 0.70 and 1.40 from the final model. We further stratified all analyses by the year of initiating lipid-lowering therapy (before/after 1995 and before/after 1996). To evaluate the goodness-of-fit of various multivariate logistic regression models, we used the Hosmer–Lemeshow statistic and considered the area under the receiver operating characteristic (ROC) curve (i.e. c statistic) [20].
All analyses were performed using SAS software (version 6.12; SAS Institute Inc, Cary, NC, USA). Relative risks are presented with point estimates and relevant 95% confidence intervals (CIs), and P values are two-tailed.
Results
We identified 22 408 patients (including 15 488 statin users, 5829 fibrate users, and 1091 other LLD users) who started receiving LLDs due to CHD, hyperlipidaemia, or other atherosclerotic diseases, and who received ≥ two LLD prescriptions during the study period. Among them, 11 922 patients continued the treatment, 3740 patients discontinued LLDs, 2835 patients switched to another LLD, and 3911 patients were lost to follow-up. Most changes in lipid-lowering therapy (1898 patients; 66.9%) entailed switching to a statin (Table 1). Treatment discontinuation/switching occurred shortly after initiating lipid-lowering therapy (Figure 1). Furthermore, the risk of discontinuation/switching differed for various classes of LLDs (P = 0.0001). By the end of 1 year, 69.8% of patients with a statin, 56.4% of patients with a fibrate, and 38.4% of patients with a nonstatin, nonfibrate LLD still used the original LLD.
Table 1.
Pattern of change in treatment among 2835 patients who switched to another lipid-lowering drug after starting the first course of lipid-lowering therapy
| Class of initial lipid-lowering drug | |||
|---|---|---|---|
| Class of present lipid-lowering drug | Statins (%) (n = 1351) | Fibrates (%) (n = 1228) | Other lipid-lowering drugs (%) (n = 256) |
| Statins (n = 1898) | 867 (64.2) | 901 (73.4) | 130 (50.8) |
| Fibrates (n = 806) | 436 (34.0) | 259 (21.1) | 111 (43.4) |
| Other lipid-lowering drugs (n = 131) | 48 (3.6) | 68 (5.5) | 15 (5.9) |
Figure 1.
Probability of discontinuation or switching of lipid-lowering therapy within 1 year of treatment initiation. *Log-rank test P = 0.0001 (for overall and all pairwise comparisons)
There were many differences in measured baseline characteristics between patients who continued and patients who discontinued LLDs. Using patients who discontinued LLDs as the reference group, patients with treatment continuation were older, had more GP visits in the previous year, more concurrent cardiovascular medications, more diabetes, and more pre-existing or recently diagnosed cardiovascular diseases. They also received statins as the initially prescribed LLD more frequently and had fewer concurrent noncardiovascular medications (Table 2, c statistic = 0.68). In addition, patients who continued LLDs had a much lower probability of recent liver (RR 0.20, 95% CI 0.12, 0.36) or renal abnormality (RR 0.43, 95% CI 0.24, 0.77; prevalence of both conditions was less than 1%).
Table 2.
Distribution of baseline characteristics and the relevant relative risk for treatment continuation among patients receiving the first course of lipid-lowering therapy.
| Number of patients with treatment continuation (%) (n = 11 922) | Number of patients with treatment discontinuation (%) (n = 3740) | Number of patients with treatment switching (%) (n = 2835) | Adjustment RR for treatment continuation* (95 % CI) Continuation vs discontinuation | Continuation vs switching | |
|---|---|---|---|---|---|
| Age (years) | |||||
| < 50 | 1760 (14.8) | 809 (21.6) | 495 (17.5) | 1.0 | 1.0 |
| 50–59 | 3564 (29.9) | 1112 (29.7) | 850 (30.0) | 1.32 (1.18, 1.48) | 1.14 (1.00, 1.29) |
| 60–69 | 4728 (39.7) | 1253 (33.5) | 1113 (39.3) | 1.51 (1.35, 1.69) | 1.13 (0.99, 1.28) |
| ≥ 70 | 1870 (15.7) | 566 (15.1) | 377 (13.3) | 1.26 (1.09, 1.45) | 1.21 (1.03, 1.42) |
| Sex | |||||
| Male | 6491 (54.5) | 1842 (49.3) | 1440 (50.8) | 1.0 | 1.0 |
| Female | 5431 (45.6) | 1898 (50.8) | 1395 (49.2) | 0.83 (0.76, 0.89) | 0.88 (0.80, 0.96) |
| Year of treatment initiation | |||||
| 1990–91 | 1128 (9.5) | 510 (13.6) | 330 (11.6) | N/A* | N/A* |
| 1992–93 | 2048 (17.2) | 883 (23.6) | 526 (18.6) | N/A | N/A |
| 1994–95 | 3413 (21.8) | 960 (25.7) | 720 (25.4) | N/A | N/A |
| 1996–97 | 5333 (44.7) | 1387 (37.1) | 1259 (44.4) | N/A | N/A |
| Body mass index (kg m−2) | |||||
| < 25 | 3091 (25.9) | 1027 (27.5) | 704 (24.8) | 1.0 | 1.0 |
| 25–29.9 | 4371 (36.7) | 1255 (33.6) | 1070 (37.7) | 1.07 (0.97, 1.18) | 0.93 (0.83, 1.04) |
| ≥ 30.0 | 2019 (16.9) | 602 (16.1) | 475 (16.8) | 0.99 (0.88, 1.12) | 0.97 (0.85, 1.11) |
| Unknown | 2441 (20.5) | 856 (22.9) | 586 (20.7) | 1.07 (0.97, 1.18) | 0.98 (0.85, 1.13) |
| Smoking status | |||||
| None | 6169 (51.7) | 1848 (49.4) | 1419 (50.1) | 1.0 | 1.0 |
| Current | 2102 (17.6) | 775 (20.7) | 541 (19.1) | 0.85 (0.76, 0.94) | 0.90 (0.80, 1.01) |
| Past | 2182 (18.3) | 556 (14.9) | 542 (19.1) | 1.06 (0.95, 1.18) | 0.88 (0.79, 0.99) |
| Unknown | 1469 (12.3) | 561 (15.0) | 333 (11.8) | 0.96 (0.83, 1.10) | 1.23 (1.04, 1.45) |
| Class of initial lipid-lowering drug (LLD) | |||||
| Other LLDs | 396 (3.3) | 359 (9.6) | 256 (9.0) | 1.0 | 1.0 |
| Statins | 8513 (71.4) | 2303 (61.6) | 1351 (47.7) | 2.56 (2.19, 3.00) | 4.04 (3.40, 4.79) |
| Fibrates | 3013 (25.3) | 1078 (28.8) | 1228 (43.3) | 2.10 (1.78, 2.48) | 1.57 (1.32, 1.87) |
| Number of general practitioner visits in the previous year | |||||
| 0–4 | 1654 (13.9) | 897 (24.0) | 577 (20.4) | 1.0 | 1.0 |
| 5–6 | 1525 (12.8) | 567 (15.2) | 421 (14.9) | 1.46 (1.28, 1.66) | 1.28 (1.11, 1.49) |
| 7–8 | 1615 (13.6) | 545 (14.6) | 394 (13.9) | 1.61 (1.42, 1.84) | 1.41 (1.22, 1.64) |
| 9–11 | 2168 (18.2) | 604 (16.1) | 485 (17.1) | 2.01 (1.77, 2.28) | 1.61 (1.40, 1.86) |
| 12–15 | 2145 (18.0) | 507 (13.6) | 426 (15.0) | 2.48 (2.17, 2.83) | 1.84 (1.59, 2.14) |
| ≥ 16 | 2815 (23.6) | 620 (16.6) | 532 (18.8) | 2.80 (2.46, 3.19) | 1.98 (1.72, 2.29) |
| Prior diagnosis of diabetes | |||||
| No | 10 159 (85.2) | 3316 (88.7) | 2457 (86.7) | 1.0 | 1.0 |
| Yes | 1763 (14.8) | 424 (11.3) | 378 (13.3) | 1.66 (1.47, 1.88) | 1.43 (1.26, 1.63) |
| Number of noncardiovascular medications | |||||
| 0 | 4755 (39.9) | 1335 (35.7) | 1044 (36.8) | 1.0 | 1.0 |
| 1 | 2715 (22.8) | 771 (20.6) | 612 (21.6) | 0.75 (0.67, 0.83) | 0.84 (0.75, 0.95) |
| 2–3 | 2711 (57.5) | 899 (24.0) | 684 (24.1) | 0.57 (0.51, 0.63) | 0.70 (0.62, 0.78) |
| ≥ 4 | 1741 (14.6) | 735 (19.7) | 495 (17.5) | 0.36 (0.31, 0.40) | 0.55 (0.48, 0.63) |
| Classes of cardiovascular medications | |||||
| 0 | 3339 (28.0) | 1480 (39.6) | 908 (32.0) | 1.0 | 1.0 |
| 1 | 2762 (23.2) | 903 (24.1) | 648 (22.9) | 1.22 (1.10, 1.35) | N/A |
| 2 | 2312 (19.4) | 578 (15.5) | 531 (18.7) | 1.58 (1.41, 1.78) | N/A |
| ≥ 3 | 3509 (29.4) | 779 (20.8) | 748 (26.4) | 1.78 (1.59, 2.00) | N/A |
| Any pre-existing cardiovascular diseases† | |||||
| No | 2312 (19.4) | 1238 (33.1) | 666 (23.5) | N/A | N/A |
| Yes | 9610 (80.6) | 2502 (66.9) | 2169 (76.5) | N/A | N/A |
| New cardiovascular diseases after initiation of lipid-lowering treatment† | |||||
| No | 11 391 (95.5) | 3629 (97.0) | 2726 (96.2) | N/A | N/A |
| Yes | 531 (4.5) | 111 (3.0) | 109 (3.8) | N/A | N/A |
| Number of risk factors for coronary heart disease (CHD) or a history of CHD‡ | |||||
| 0–1 | 2818 (23.6) | 1452 (38.8) | 804 (28.4) | N/A | N/A |
| ≥ 2 | 2582 (21.7) | 788 (21.1) | 590 (20.8) | N/A | N/A |
| Prior CHD | 6522 (54.7) | 1500 (40.1) | 1441 (50.8) | N/A | N/A |
RR denotes relative risk and CI denotes confidence interval; N/A indicates that the RR estimate for the variable was not available because the variable was not included in the final model. All adjusted RRs were derived from a model that simultaneously controlled for the other variables;
Including hypertension, congestive heart failure, coronary heart disease, peripheral vascular disease, and cerebrovascular disease;
The following risk factors other than serum cholesterol for coronary heart disease were included: (1) male ≥ 45 years or female ≥ 55 years or premature menopause without oestrogen replacement therapy; (2) current cigarette smoking; (3) hypertension; and (4) diabetes.
Compared with patients who switched therapy, patients with treatment continuation also had more GP visits and more diabetes, and received statins as the initial LLD more frequently, yet had fewer concurrent noncardiovascular medications (Table 2, c statistic = 0.68). In addition, they were less likely to have recent renal abnormality (RR 0.41, 95% CI 0.22, 0.78). Nevertheless, they had slightly more cardiovascular diseases and received slightly more concurrent cardiovascular medications (data not shown).
When we compared people's patterns of LLD use after switching, we found that patients who continued receiving LLDs after treatment switching were more likely to receive concurrent cardiovascular medications, to have pre-existing cardiovascular diseases, and to switch to another LLD 6 or more months after initiation of treatment (Table 3, c statistic = 0.72). Furthermore, patients with treatment continuation after switching therapy frequently received statins as the substituted LLD. They also tended to have more newly diagnosed cardiovascular diseases (data not shown).
Table 3.
Distribution of baseline characteristics and the relevant relative risk for treatment continuation among patients who switched lipid-lowering therapy.
| Number of patients who continued LLD after switching (%) (n = 1970) | Number of patients who discontinued LLD after switching (%) (n = 566) | Crude RR for treatment continuation* (95% CI) | Adjusted RR for treatment continuation* (95% CI) | |
|---|---|---|---|---|
| Age (years) | ||||
| < 50 | 346 (17.6) | 102 (18.0) | 1.0 | 1.0 |
| 50–59 | 600 (30.5) | 162 (28.6) | 1.09 (0.82, 1.45) | 1.08 (0.80, 1.47) |
| 60–69 | 769 (39.0) | 225 (39.8) | 1.01 (0.77, 1.32) | 0.98 (0.73, 1.32) |
| ≥ 70 | 255 (12.9) | 77 (13.6) | 0.98 (0.70, 1.37) | 0.91 (0.62, 1.34) |
| Sex | ||||
| Male | 1048 (53.2) | 244 (43.1) | 1.0 | 1.0 |
| Female | 922 (46.8) | 322 (56.9) | 0.67 (0.55, 0.81) | 0.80 (0.65, 0.99) |
| Year of treatment initiation | ||||
| 1990–91 | 244 (12.4) | 86 (15.2) | 1.0 | N/A* |
| 1992–93 | 392 (19.9) | 129 (22.8) | 1.07 (0.78, 1.47) | N/A |
| 1994–95 | 504 (25.6) | 151 (26.7) | 1.18 (0.87, 1.60) | N/A |
| 1996–97 | 830 (42.1) | 200 (35.3) | 1.46 (1.09, 2.00) | N/A |
| Body mass index (kg m−2) | ||||
| < 25.0 | 458 (23.3) | 161 (28.5) | 1.0 | 1.0 |
| 25.0–29.9 | 770 (39.1) | 199 (35.2) | 1.36 (1.07, 1.73) | 1.24 (0.96, 1.60) |
| ≥ 30.0 | 337 (17.1) | 82 (14.5) | 1.45 (1.07, 1.95) | 1.28 (0.93, 1.76) |
| Unknown | 405 (20.1) | 124 (21.9) | 1.15 (0.88, 1.50) | 1.22 (0.86, 1.72) |
| Smoking status | ||||
| None | 989 (50.2) | 286 (50.5) | 1.0 | 1.0 |
| Current | 354 (18.0) | 115 (20.3) | 0.89 (0.70, 1.14) | 0.87 (0.66, 1.14) |
| Past | 395 (20.1) | 84 (14.8) | 1.36 (1.04, 1.78) | 1.18 (0.89, 1.57) |
| Unknown | 232 (11.8) | 81 (14.3) | 0.83 (0.62, 1.10) | 0.82 (0.56, 1.20) |
| Class of lipid-lowering drug (LLDs) after switching of therapy | ||||
| Other LLDs | 72 (3.7) | 51 (9.0) | 1.0 | 1.0 |
| Statins | 1361 (69.1) | 313 (55.3) | 3.08 (2.11, 4.50) | 2.24 (1.49, 3.36) |
| Fibrates | 537 (27.3) | 202 (35.7) | 1.88 (1.27, 2.79) | 1.74 (1.14, 2.65) |
| Interval between initiation and switching of therapy | ||||
| < 3 months | 383 (19.4) | 246 (43.5) | 1.0 | 1.0 |
| 3 to < 6 months | 661 (33.6) | 220 (38.9) | 1.93 (1.55, 2.41) | 1.88 (1.50, 2.37) |
| ≥ 6 months | 926 (47.0) | 100 (17.7) | 5.95 (4.58, 7.72) | 5.56 (4.26, 7.27) |
| Classes of cardiovascular medications | ||||
| 0 | 598 (30.4) | 226 (39.9) | 1.0 | 1.0 |
| 1 | 435 (22.1) | 140 (24.7) | 1.17 (0.92, 1.50) | 1.18 (0.91, 1.53) |
| 2 | 372 (18.9) | 99 (17.5) | 1.42 (1.09, 1.86) | 1.52 (1.14, 2.04) |
| ≥ 3 | 565 (28.7) | 101 (17.8) | 2.11 (1.63, 2.75) | 2.02 (1.52, 2.69) |
| Any pre-existing cardiovascular diseases† | ||||
| No | 430 (21.8) | 178 (31.5) | 1.0 | N/A |
| Yes | 1540 (78.2) | 388 (68.6) | 1.64 (1.34, 2.02) | N/A |
| Number of risk factors for coronary heart disease (CHD) or a history of CHD‡ | ||||
| 0–1 | 517 (26.2) | 212 (37.5) | 1.0 | N/A |
| 2–4 | 419 (21.3) | 107 (18.9) | 1.61 (1.23, 2.10) | N/A |
| Prior CHD | 1034 (52.5) | 247 (43.6) | 1.72 (1.39, 2.12) | N/A |
RR denotes relative risk and CI denotes confidence interval; N/A indicates that the RR estimate for the variable was not available because the variable was not included in the final model. All adjusted RRs were derived from a model that simultaneously controlled for the other variables;
Including hypertension, congestive heart failure, coronary heart disease, peripheral vascular disease, and cerebrovascular disease;
The following risk factors other than serum cholesterol for coronary heart disease were included: (1) male ≥ 45 years or female ≥ 55 years or premature menopause without oestrogen replacement therapy; (2) current cigarette smoking; (3) hypertension; and (4) diabetes.
We further stratified the analyses by calendar year and found no significant effect modification for all baseline characteristics or for rate of treatment continuation. We further found that the baseline characteristics collected before treatment initiation were generally less predictive of pattern of LLD utilization than the information collected before the index date (data not shown).
Discussion
We found that discontinuation/switching of lipid-lowering therapy was frequent during the study period. Our findings also suggest that treatment continuation after starting LLDs was more frequent among patients who initially received a statin, and in patients who had more frequent GP visits, more concurrent cardiovascular medications, and more diabetes. Conversely, patients using more noncardiovascular medications were less likely to continue lipid-lowering therapy. Further, treatment continuation after switching therapy was more likely among patients who changed their treatment to statins, in patients who switched to another LLD 6 or more months after treatment initiation, and in patients who received concurrent cardiovascular medications.
Our findings on treatment discontinuation/switching of various LLDs were consistent with previous observational studies, which have found a higher rate of discontinuation/noncompliance than that in the clinical trials, especially among users of nonstatin LLDs [7, 8, 9, 10, 11, 12, 13, 14]. Although differential prescribing of LLDs may be sought as an explanation for the observations, the difference between various LLDs persisted after adjusting for measured baseline characteristics. The safety profile, the potency, and the dosing formulation/frequency have been noted to vary between different classes of LLDs [21, 22]. Therefore, despite the fact that all LLDs reduce the risk of CHD in clinical trials [1, 2, 3, 4, 5, 6], statin therapy was associated with more treatment continuations in the present study.
The relationship between other predictors and treatment continuation of LLDs has also been studied [8, 9, 13, 14]. For example, Avorn et al. reported an association between risk factors for CHD, statin use, and long-term LLD use, and better compliance of LLDs [8, 13]. Simons et al.[9] found that older age, and use of concurrent cardiovascular drugs or analgesics were associated with a higher rate of treatment continuation, while gemfibrozil and antidepressants were predictive of treatment discontinuations. No adjustment for smoking status, BMI, and comorbidities or health care utilization that occurred after treatment initiation was performed in previous studies. Our study, however, confirms that even after adjusting for important characteristics that occurred after initiation of lipid-lowering therapy, most of the aforementioned findings still hold.
Similar association between treatment continuation and comorbidities, medication use, and health-care utilization has also been noted among users of other preventive medications, such as antihypertensive agents [23–26]. According to the Health Belief Model [27], patients who have a perceived need to treat their conditions effectively are more likely to take the recommended medications. Therefore, patients with higher risk for CHD are likely to achieve better compliance because the beneficial effects of LLDs on CHD are well-documented. General practitioners also focus the treatment of hyperlipidaemia more among patients with higher risk of developing CHD [28]. Conversely, patients with multiple unrelated medications are less likely to be compliant because both GPs and patients focus on treating noncardiovascular comorbidities [29, 30]. Multiple physician visits may have allowed for better education of patients, which may explain the observed relationship between treatment continuation and health-care utilization [23]. Further, frequent GP visits may indicate more patient-initiated visits, which are important in determining compliance [27, 29].
One of the strengths of this study is the ability to include information on new diagnoses and treatment that occurred after initiation of lipid-lowering therapy. The use of computerized records in the GPRD to assess factors potentially associated with discontinuation/switching of lipid-lowering therapy further offers major advantages in the study of LLD utilization. Both drug exposure and the baseline characteristics are unlikely to be misclassified, as previous studies have shown the high quality of information recorded in the GPRD [15, 16], and LLDs are usually prescribed on an outpatient basis and are filled continually. Although misclassification remains possible in the study, the magnitude of such misclassification is likely to be small.
Some limitations of this study should be acknowledged. Firstly, we did not have adequate information on the reason for treatment discontinuation. Therefore, it was not possible to evaluate the differences in patients with different reasons for treatment discontinuation. Nevertheless, after looking into many baseline characteristics, we are not aware of any major difference between these patients. Secondly, we only studied patients who were on the first course of LLD therapy. As long-term LLD users were likely to be different from new users [8], interpretation of the study findings should be cautious. Thirdly, this study only included patients who initiated lipid-lowering therapy before 1998. As the national guidelines on use of LLDs [31, 32] and the actual prescribing behaviour of LLDs [33, 34] in the UK changed significantly after the publication of the Scandinavian Simvastatin Survival Study in November 1994 [35], it might be questioned whether the inclusion of more recent data would change the study findings. In analyses stratified by year of treatment initiation (by 1995 or by 1996), we did not find any significant effect modification by calendar year, although the effect of statin prescribing on treatment continuation was somewhat stronger among patients with more recent treatment initiation. Two studies that included patients who initiated statin therapy up to December 1998 have reported similar unsatisfactory adherence to statin therapy in recent years [13, 14]. Therefore, our findings should be generalizable to patients who initiated LLD therapy after 1997.
In conclusion, we found that the choice of LLDs, patient characteristics, comorbidities, and health care utilization were associated with treatment continuation in patients newly treated with LLDs. Although it is not possible to achieve a full understanding of the cause of treatment discontinuation/switching, our findings provide important information about the target population that could potentially benefit from more patient education and monitoring. Practice guidelines and quality-of-care assessment for lipid-lowering therapy should emphasize factors that predispose patients to treatment discontinuation/switching, in an effort to optimize the choice of therapeutic regimens and to improve patient adherence.
Acknowledgments
We are grateful to K. Arnold Chan and Kimberly M. Thompson for their helpful comments on the manuscript. We also thank the participating general practitioners for their excellent cooperation and generous help. The Boston Collaborative Drug Surveillance Program is supported in part by grants from Berlex Laboratories, GlaxoSmithKline, Ingenix Pharmaceutical Services, Johnson & Johnson Pharmaceutical Research & Development LLC, Hoffmann-La Roche, Pharmacia Corporation and Novartis Farmacéutica. Chen-Chang Yang receives support from the Harvard Pharmacoepidemiology Teaching and Research Fund, the Yin Xun-Ruo Educational Foundation, and the Foundation for Poison Control in Taiwan. This study was not directly funded.
References
- 1.LaRosa JC, He J, Vupputuri S. Effects of statins on risk of coronary heart disease: a meta-analysis of randomized controlled trials. JAMA. 1999;282:2340–2346. doi: 10.1001/jama.282.24.2340. [DOI] [PubMed] [Google Scholar]
- 2.Committee of Principal Investigators. A cooperative trial in the primary prevention of ischaemic heart disease using clofibrate. Br Heart J. 1978;40:1069–1118. doi: 10.1136/hrt.40.10.1069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Frick MH, Elo O, Haapa K, et al. Helsinki Heart Study: primary-prevention trial with gemfibrozil in middle-aged men with dyslipidemia. N Engl J Med. 1987;317:1237–1245. doi: 10.1056/NEJM198711123172001. [DOI] [PubMed] [Google Scholar]
- 4.Dorr AE, Gundersen K, Schneider JC, Spencer TW, Martin WB. Colestipol hydrochloride in hypercholesterolaemic patients – effect on serum cholesterol and mortality. J Chronic Dis. 1978;31:5–14. doi: 10.1016/0021-9681(78)90076-0. [DOI] [PubMed] [Google Scholar]
- 5.Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. I. Reduction in incidence of coronary heart disease. JAMA. 1984;251:351–364. doi: 10.1001/jama.1984.03340270029025. [DOI] [PubMed] [Google Scholar]
- 6.The Coronary Drug Project Research Group. Clofibrate and niacin in coronary heart disease. JAMA. 1975;231:360–381. [PubMed] [Google Scholar]
- 7.Andrade SE, Walker AM, Gottlieb LK, et al. Discontinuation of antihyperlipidemic drugs – do rates reported in clinical trials reflect rates in primary care settings? N Engl J Med. 1995;332:1125–1131. doi: 10.1056/NEJM199504273321703. [DOI] [PubMed] [Google Scholar]
- 8.Avorn J, Monette J, Lacour A, et al. Persistence of use of lipid-lowering medications: a cross–national study. JAMA. 1998;279:1458–1462. doi: 10.1001/jama.279.18.1458. [DOI] [PubMed] [Google Scholar]
- 9.Simons LA, Levis G, Simons J. Apparent discontinuation rates in patients prescribed lipid- lowering drugs. Med J Austral. 1996;164:208–211. doi: 10.5694/j.1326-5377.1996.tb94138.x. [DOI] [PubMed] [Google Scholar]
- 10.Bruckert E, Simonetta C, Giral P and the CREOLE study team. Compliance with fluvastatin treatment: Characterization of the noncompliant population within a population of 3,845 patients with hyperlipidemia. J Clin Epidemiol. 1999;52:589–594. doi: 10.1016/s0895-4356(99)00019-0. [DOI] [PubMed] [Google Scholar]
- 11.Oster G, Borok GM, Menzin J, et al. Cholesterol-Reduction Intervention Study (CRIS) Arch Intern Med. 1996;156:731–739. doi: 10.1001/archinte.156.7.731. [DOI] [PubMed] [Google Scholar]
- 12.Beggs PW, Clark WJ, Williams SM, Coulter DM. A comparison of the use, effectiveness, and safety of bezafibrate, gemfibrozil, and simvastatin in normal clinical practice using the New Zealand Intensive Medicines Monitoring Programme (IMMP) Br J Clin Pharmacol. 1999;47:99–104. doi: 10.1046/j.1365-2125.1999.00846.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-term persistence in use of statin therapy in elderly patients. JAMA. 2002;288:455–461. doi: 10.1001/jama.288.4.455. [DOI] [PubMed] [Google Scholar]
- 14.Jackevicius CA, Mamdani M, Tu JV. Adherence with statin therapy in elderly patients with and without acute coronary syndrome. JAMA. 2002;288:462–467. doi: 10.1001/jama.288.4.462. [DOI] [PubMed] [Google Scholar]
- 15.Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerized data resource in the United Kingdom. Br Med J. 1991;302:766–768. doi: 10.1136/bmj.302.6779.766. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Jick H, Terris BZ, Derby LE, Jick SS. Further validation of information recorded on a general practitioner based computerized data resource in the United Kingdom. Pharmacoepidemiol Drug Saf. 1992;1:347–349. [Google Scholar]
- 17.Expert Panel on Detection, Evaluation, Treatment of High Blood Cholesterol in Adults. Second report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel II) Circulation. 1994;89:1333–1445. doi: 10.1161/01.cir.89.3.1333. [DOI] [PubMed] [Google Scholar]
- 18.Charlson ME, Pompei P, Ales KL, Mackenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 19.Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol. 1993;46:1075–1079. doi: 10.1016/0895-4356(93)90103-8. [DOI] [PubMed] [Google Scholar]
- 20.Hosmer DW, Lemeshow S. Applied Logistic Regression. 2. New York: John Wiley & Sons; 2000. [Google Scholar]
- 21.Safeer RS, Lacivita CL. Choosing drug therapy for patients with hyperlipidemia. Am Fam Physician. 2000;61:3371–3382. [PubMed] [Google Scholar]
- 22.Knopp RH. Drug treatment of lipid disorders. N Engl J Med. 1999;341:498–510. doi: 10.1056/NEJM199908123410707. [DOI] [PubMed] [Google Scholar]
- 23.Monane M, Bohn RL, Gurwitz JH, Glynn RJ, Levin R, Avorn J. The effects of initial drug choice and comorbidity on antihypertensive therapy compliance. Am J Hypertension. 1997;10:697–704. doi: 10.1016/s0895-7061(97)00056-3. [DOI] [PubMed] [Google Scholar]
- 24.Jones JK, Gorkin L, Lian JF, Staffa JA, Fletcher AP. Discontinuation of and changes in treatment after start of new courses of antihypertensive drugs: a study of a United Kingdom population. Br Med J. 1995;311:293–295. doi: 10.1136/bmj.311.7000.293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gillum RF, Neutra RR, Stason WB, Solomon HS. Determinants of dropout rate among hypertensive patients in an urban clinic. J Community Hlth. 1979;5:94–100. doi: 10.1007/BF01324011. [DOI] [PubMed] [Google Scholar]
- 26.Caldwell JR, Cobb SC, Dowling MD, Jongh DD. The dropout problem in antihypertensive treatment. J Chronic Dis. 1970;22:579–592. doi: 10.1016/0021-9681(70)90034-2. [DOI] [PubMed] [Google Scholar]
- 27.Janz NK, Becker MH. The health belief model: a decade later. Health Educ Q. 1984;11:1–47. doi: 10.1177/109019818401100101. [DOI] [PubMed] [Google Scholar]
- 28.Fairhurst K, Huby G. From trial data to practical knowledge: qualitative study of how general practitioners have assessed and used evidence about statin drugs in their management of hypercholesterolaemia. Br Med J. 1998;317:1130–1134. doi: 10.1136/bmj.317.7166.1130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.German PS. Compliance and chronic disease. Hypertension. 1988;11(Suppl II):II56–60. doi: 10.1161/01.hyp.11.3_pt_2.ii56. [DOI] [PubMed] [Google Scholar]
- 30.Redelmeier DA, Tan SH, Booth GL. The treatment of unrelated disorders in patients with chronic medical diseases. N Engl J Med. 1998;338:1516–1520. doi: 10.1056/NEJM199805213382106. [DOI] [PubMed] [Google Scholar]
- 31.Betteridge DJ, Dodson PM, Durrington PN, et al. Management of hyperlipidaemia: guidelines of the British Hyperlipidaemia Association. Postgrad Med J. 1993;69:359–369. doi: 10.1136/pgmj.69.811.359. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Wood D, Durrington P, McInnes G, Poulter N, Rees A, Wray R for the British Cardiac Society, British Hyperlipidaemia Association, British Hypertension Society and British Diabetic Association. Joint British recommendations on prevention of coronary heart disease in clinical practice. Heart. 1998;80:S1–S29. [PMC free article] [PubMed] [Google Scholar]
- 33.Majeed A, Moser K, Maxwell R. Age, sex and practice variations in the use of statins in general practice in England and Wales. J Public Health Med. 2000;22:275–279. doi: 10.1093/pubmed/22.3.275. [DOI] [PubMed] [Google Scholar]
- 34.Baxter C, Jones R, Corr L. Time trend analysis and variations in prescribing lipid lowering drugs in general practice. Br Med J. 1998;317:1134–1135. doi: 10.1136/bmj.317.7166.1134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Scandinavian Simvastatin Survival Study Group. Randomized trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S) Lancet. 1994;344:1383–1389. [PubMed] [Google Scholar]