Assessment of Patients’ and Physicians’ Compliance to an ACE Inhibitor Treatment Based on Urinary N-Acetyl Ser-Asp-Lys-Pro Determination in the Noninsulin-Dependent Diabetes, Hypertension, Microalbuminuria, Proteinuria, Cardiovascular Events, and Ramipril (DIABHYCAR) Study

The design of the study has been described elsewhere (1). DIABHYCAR was a 4-year, general practice–based, multicenter, randomized, double-blind, parallel-group trial comparing ramipril (1.25 mg o.d.) with placebo, in addition to usual antidiabetic and cardiovascular treatments, on the occurrence of cardiovascular and renal events in 4,912 patients with type 2 diabetes and persistent microalbuminuria or proteinuria.

Urine samples for the determination of AcSDKP and creatinine were obtained from all the 1,871 participants who completed follow-up and provided spot urine samples for UAE measurements. All AcSDKP determinations were performed blind to treatment using a commercially available enzyme immunoassay kit (AcSKDP EIA kit, SpiBio, CEA, Gif-sur-Yvette, France). The methods for UAE determination and the definition of significant albuminuria regression were as previously described (1).

Frequency histograms and nonparametric clustering methods (13) were used to define an optimum threshold for detectable ACE inhibition, using the log-transformed urinary AcSDKP-to-creatinine ratio. The algorithm used for the clustering method was based on spherical uniform kernel density estimations and was implemented in the MODECLUS procedure of SAS statistical software (14). The validity of the cutoff point of the log-transformed urinary AcSDKP-to-creatinine ratio was checked, using various clustering algorithms and other methods, such as receiver operating characteristic curves (not shown).

Depending on the distribution of the variables, unpaired t tests or nonparametric tests were used to compare the ramipril and placebo groups. Categorical variables were compared using the χ² test. Covariance analysis, with treatment group and the baseline value as covariates, was used to assess the effect of treatment on systolic blood pressure (SBP) and UAE at last evaluation. The probability of UAE regression in the two treatment groups was compared by means of a logistic regression model. Continuous data are expressed as means ± SD or medians (interquartile range). AcSDKP-to-creatinine ratios are expressed as medians with 95% CI. A P value of < 0.05 was considered significant. Statistical analyses were performed with SAS statistical software version 8.2 (SAS Institute, Cary, NC).

Nine hundred thirty-seven patients were randomly assigned to the ramipril group and 934 to the placebo group; patients were similar on entry into the study for all factors except prevalence of previous cardiovascular disease, which was higher in the placebo group than in the ramipril group (Table 1). These patients define the intention-to-treat population. At randomization, 59% of the patients in the ramipril group and 57% of the patients in the placebo group were taking 4–10 additional nonstudy drugs. In the intention-to-treat population, 281 patients of the ramipril group (30%) and 234 patients of the placebo group (25.1%, P = 0.017) prematurely discontinued their assigned double-blind treatment before the end of the trial because of either consent withdrawal or the occurrence of an adverse drug event, as reported in the case report form. In the intention-to-treat population, the prescription of a nonstudy open-label ACE inhibitor to 219 patients of the ramipril group (23.4%) and 212 patients of the placebo group (22.7%, P = 0.73) was reported in the case report form. Finally, 340 patients of the ramipril group (36.3%) and 313 patients of placebo group (33.5%, P = 0.21) had both conditions, i.e., premature discontinuation of their assigned double-blind treatment and open-label ACE inhibitor prescription.

The AcSDKP-to-creatinine ratio achieved in the ramipril group was six times higher than that in the placebo group (median 9.67 [95% CI 8.78–10.93] vs. 1.54 [1.47–1.59] nmol/mmol, respectively, P < 0.0001). Log-transformed AcSDKP-to-creatinine ratio displayed a bimodal distribution in both the ramipril and the placebo groups (Fig. 1). There was a very large overlap of values between the two groups: the range of AcSDKP-to-creatinine ratios was 0.42–131 nmol/mmol in the placebo group and 0.26–141 nmol/mmol in the ramipril group (Fig. 1).

The same analysis was performed after excluding patients known to have prematurely discontinued their allocated double-blind treatment and those known to have been treated with an open-label ACE inhibitor, as reported in the case report form (ramipril group, n = 340 and placebo group, n = 313). This defines the per-protocol population (n = 597 in the ramipril group and n = 621 in the placebo group). This per-protocol analysis increased the difference in urine AcSDKP-to-creatinine ratio between the ramipril (n = 597) and placebo (n = 621) groups (ramipril median 10.82 [95% CI 9.62–11.78] vs. placebo 1.45 [1.38–1.54] nmol/mmol, P < 0.0001), but there was still a large overlap of values between the two groups (Fig. 1).

Cluster analysis on log-transformed urinary AcSDKP-to-creatinine ratios confirmed the bimodal distribution of these ratios and delineated two separate clusters of patients in each group. We determined the threshold of detectable ACE inhibition by analyzing specifically the distribution of the AcSDKP-to-creatinine ratio in the cluster of patients treated with ramipril with high AcSDKP-to-creatinine ratios. Given the high sensitivity and specificity of urinary AcSDKP determination for assessing ACE inhibition (12), this cluster corresponds to patients who must have taken ramipril and/or a nonstudy ACE inhibitor. We selected a threshold of 4 nmol/mmol as the lower limit of the AcSDKP-to-creatinine ratio distribution in this cluster.

In intention-to-treat analysis, 331 of the 937 patients given ramipril (35.3%) had a ratio <4 nmol/mmol and 182 of the 934 (19.5%) patients given placebo had an AcSDKP-to-creatinine ratio ≥4 nmol/mmol. In the per-protocol analysis, 163 of the 597 patients in the ramipril group (27.3%) had a ratio <4 nmol/mmol and 60 of the 621 patients of the placebo group (9.7%) had a ratio ≥4 nmol/mmol. In the per-protocol population, we defined “truly adherent ramipril” patients (TA_R) as those with a ratio ≥4 nmol/mmol (n = 434) among the 597 patients of the ramipril group and “truly adherent placebo” patients (TA_P) as those with a ratio <4 nmol/mmol (n = 561) among the 621 patients of the placebo group (Fig. 1). The TA_R and TA_P groups had similar characteristics on inclusion (not shown).

In intention-to-treat analysis, the baseline-adjusted mean difference in final SBP between the ramipril and placebo group was −1.40 mmHg (95%CI −2.55 to −0.25; P = 0.017). In per-protocol analysis, this difference was −1.73 mmHg (−3.06 to −0.40; P = 0.011). The AcSDKP-guided analysis slightly amplified the baseline-adjusted mean difference in final SBP between the TA_R and TA_P groups, from −1.40 mmHg (−2.55 to −0.25) to −2.23 mmHg (−3.69 to −0.77, P < 0.0031).

In intention-to-treat analysis, the proportion of patients with albuminuria regression in the ramipril group (27.1%, 254 of 937 patients) did not differ significantly from that in the placebo group (23.2%, 217/934 patients). Changes in UAE were also similar in the two groups (ramipril group median +3% [interquartile range −59 to 161] vs. placebo group +4% [−57 to 188], NS). In the per-protocol analysis, changes in UAE were also similar (ramipril group +3% [−59 to 153] vs. placebo group +9% [−56 to 196], NS). In AcSDKP-guided analysis, 120 of 434 TA_R patients (27.7%) presented significant regression of albuminuria vs. 124 of 561 TA_P patients (22.1%, P = 0.044). UAE regressed in patients receiving 1.25 mg ramipril (−7% [−60 to 143]) whereas it increased in patients taking placebo (+9% [−55 to 188], P = 0.048).

We report for the first time the use of an endogenous biomarker of compliance, urinary AcSDKP determination (10,12) in a large-scale, double-blind, randomized, placebo-controlled clinical trial of an ACE inhibition in patients with type 2 diabetes and microalbuminuria/proteinuria. At the end of this trial, the median AcSDKP-to-creatinine ratio was six times higher in the ramipril group than in the placebo group, overall confirming the fact that 1.25 mg ramipril o.d. effectively inhibited ACE. Interestingly, the distribution of the AcSKDP-to-creatinine ratio was bimodal in both the placebo and ramipril groups. Based on this bimodal distribution, we identified a AcSKDP-to-creatinine ratio of 4 nmol/mmol as the threshold defining detectable ACE inhibition. This threshold was used to define compliance with active treatment by patients and compliance with the protocol by physicians. Both were less than expected from the information collected in the case report forms. Because the AcSDKP methodology was not available during the DIABHYCAR study, the consequences of noncompliance, such as that objectively assessed by spot urine AcSDKP measurements, can only be analyzed on blood pressure and UAE in patients who had a complete follow-up. Correcting for compliance in these patients minimally affected these surrogate outcome results.

In the ramipril group, 35.3% of the patients had urinary AcSKDP-to-creatinine ratios <4 nmol/mmol in intention-to-treat analysis, within the range of values indicating an absence of ACE inhibitor intake. In the per-protocol analysis, 27.3% of the patients had AcSKDP-to-creatinine ratios less than the 4 nmol/mmol threshold, indicating a lack of compliance with ramipril treatment. This proportion may have been overestimated if some spot urine samples were collected >24 h after the last administration of the drug. However, the urinary AcSDKP-to-creatinine ratio remained high for up to 4 days after the cessation of treatment with an ACE inhibitor, such as 50 mg captopril b.i.d (10), and delays in urine collection with respect to the last drug intake are therefore unlikely to overestimate the percentage of nonadherent patients. We can conclude with confidence that those patients with a low AcSKDP-to-creatinine ratio did not take their dose of ramipril for at least a few days before urine sampling. Conversely, the proportion of ramipril patients with AcSDKP-to-creatinine ratios ≥4 nmol/mmol may overestimate compliance with daily treatment intake due to “white coat” or partial compliance (15,16), as the AcSDKP-to-creatinine ratio is extremely sensitive to ACE inhibition, increasing within 0.5–1 h after a single oral dose of 50 mg captopril (10).

Conventional clinical assessment methods estimated compliance to be similar in the DIABHYCAR and MICROHOPE trials. The percentage of patients in the ramipril group still receiving the allocated treatment at the end of the study, as assessed clinically by pill counting, was 70% in DIABHYCAR vs. 65% in MICROHOPE. However, compliance was not measured objectively at the time of publication of these results. Biological monitoring with urinary AcSDKP in DIABHYCAR gives less reassuring results than clinical monitoring. In large scale morbidity/mortality trials, such as DIABHYCAR or MICROHOPE, there may be considerable selective or nonselective noncompliance with the prescribed treatment for at least two reasons. 1) Patients are already taking a large number of drugs (in DIABHYCAR, ∼60% of the patients took 4–10 other, nonstudy drugs). This factor, together with the number of daily intakes, is known to decrease compliance (8). 2) In such long-term randomized trials, patients may discard the experimental treatment (active or placebo) more readily than nonstudy drugs, as they are of unknown efficacy, as stated in informed consent forms (8). The ∼25% noncompliance observed when considering AcSKDP-to-creatinine ratio is consistent with the level of noncompliance observed in diabetic subjects treated with oral hypoglycemic agents (15–33%) in prospective studies using electronic compliance monitoring systems (3).

In the placebo group, 60 of the 621 placebo-treated patients (9.7%) had AcSDKP-to-creatinine ratios ≥4 nmol/mmol, in the absence of any ACE inhibitor prescription explicitly mentioned in their case report form. This surprising finding highlights the magnitude of under-reporting of critical drug treatment by patients (who may be treated by several different physicians) or by physicians involved in the trial at least at its end.

The percentage of patients in the placebo group taking nonstudy ACE inhibitors, as assessed on the basis of interviews, was 22.7% in DIABHYCAR and 15% in MICROHOPE (2). A “cross-contamination” of the placebo group with the active drug is frequently observed in clinical trials especially 1) in open-label trials (which was not the case for DIABHYCAR or MICROHOPE); 2) if a biological/clinical marker of efficacy, such as plasma lipid concentration, can be easily monitored, as observed in the FIELD (Fenofibrate Intervention and Event Lowering in Diabetes) (17) and ALLHAT-LLT (Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial) (18) trials; 3) if the clinical condition of the patient deteriorates; or 4) if the results of another clinical trial dealing with aspects of the disease under investigation are published before the end of the ongoing trial, with physicians anticipating the results of the new trial before publication (19). This may have been the case in the DIABHYCAR study, as the results of the MICROHOPE study were published on 22 January 2000 and DIABHYCAR ended on 21 March 2001 when urinary determinations were carried out.

Noncompliance with the active treatment and the use of nonstudy ACE inhibitor in the placebo group in the DIABHYCAR and MICROHOPE studies would have minimized the magnitude of the beneficial effect of ramipril. However, correcting for compliance in the DIABHYCAR study makes only a small difference in surrogate outcome results, i.e., decreases in UAE and SBP. This finding reinforces the main conclusion obtained from the comparison of the two studies results: a strong inhibition of the renin-angiotensin system with high doses of an ACE inhibitor is required to reduce both cardiovascular risk and UAE in patients with type 2 diabetes and micro/macroalbuminuria. In the MICROHOPE study, the prescription of high doses of the ACE inhibitor may have contributed per se to minimize the issue of noncompliance with the active treatment. High doses produce more sustained and prolonged ACE inhibition (20) but also attenuate the effects of an intentionally or unintentionally missed dose (21), thereby reducing the impact of partial or noncompliance.

There is no gold standard for measuring compliance with treatment. Direct determination of the plasma concentration of the drug is generally considered expensive, susceptible to white coat adherence bias, and complex (8). Urinary AcSDKP determination is inexpensive, based on a nonradioactive assay, and requires only a random spot urine sample at any moment of the day, with no special requirements for urine sampling. It can be performed by nonspecialist laboratories and does not require complex statistical models for its interpretation.

As shown in this large-scale ACE inhibition trial, the systematic use of this method could provide insight into patient noncompliance, which is underestimated by clinical methods, and compliance with the protocol, which is underestimated by the physicians’ recording of nonstudy drugs in the case report form. These results confirm that the effect of poor compliance with active treatment is probably underestimated in most large-scale randomized controlled trials (22,23).

In usual care, compliance with treatment may be even poorer than in clinical studies, reducing the benefits of treatment below expectations based on trial results. Because ACE inhibition is a cost-effective therapy in patients with diabetes (24) and poor compliance to treatment is independently associated with poor clinical outcome (23), we suggest that urinary AcSDKP determination should be more widely used to detect noncompliance in patients with diabetes displaying a less favorable than expected response to ACE inhibitor prescription.

The design, funding and interpretation of this adherence analysis were independent of the sponsor of the DIABHYCAR trial (Aventis, Paris). This study was supported by a grant from Association Naturalia et Biologia and Association Robert Debré, Paris, France.

We thank Christiane Dollin for her technical contribution.