Author + information
- Received March 2, 1999
- Revision received June 18, 1999
- Accepted August 23, 1999
- Published online December 1, 1999.
- Allan D Struthers, MD∗,*,
- Robert MacFadyen, MD∗,
- Callum Fraser, PhD†,
- Jess Robson∗,
- James J Morton, PhD§,
- Christophe Junot, BSc‡ and
- Eric Ezan, PhD‡
- ↵*Reprint requests and correspondence: Professor A. D. Struthers, Department of Clinical Pharmacology and Therapeutics, Ninewells Hospital, Dundee DD1 9SY, UK
This study was designed to compare different proposed methods of assessing adherence with angiotensin-converting enzyme (ACE) inhibitor (ACEI) therapy in chronic heart failure.
The use of ACEIs in chronic heart failure gives us a unique opportunity to assess a patient’s adherence by measuring whether the expected biochemical effect of an ACEI is present in the patient’s bloodstream. In fact, there are several different ways of assessing ACE in vivo: these are serum ACE activity itself, plasma N-acetyl-seryl-aspartyl-lysyl-proline (AcSDKP), urine AcSDKP, plasma angiotensin I (AI), plasma angiotensin II (AII), or the AII/AI ratio.
Patients with chronic heart failure (n = 39) were randomized to regimens of ACEI nonadherence for one week, ACEI adherence for one week or two versions of partial adherence for one week, after which the above six tests were performed.
All six tests significantly distinguished between full nonadherence for one week and full or partial adherence. Only plasma AcSDKP produced a significantly different result between partial adherence and either full adherence or full nonadherence for one week. In terms of their ability to distinguish full nonadherence from full adherence, plasma AcSDKP was 89% sensitive and 100% specific with an area under its ROC of 0.95. Corresponding figures for urine AcSDKP were 92%, 97% and 0.95 and for serum ACE they were 86%, 95% and 0.90.
All six tests distinguished full nonadherence from all other forms of adherence. The rank order of performance was plasma AcSDKP, urine AcSDKP, serum ACE, AII/AI ratio and plasma AII followed by plasma AI.
In chronic heart failure, the main therapeutic advance in the last 20 years has been the advent of angiotensin-converting enzyme (ACE) inhibitors (ACEIs) that improve mortality, morbidity and hospitalization rates (1). For an individual patient to gain maximum benefit from ACEI therapy, it is obviously important to adhere to this therapy. Yet most patients with chronic heart failure are elderly and receiving multiple drug therapy, which are both factors known to enhance nonadherence. Indeed, many studies do show that nonadherence with drug therapy is common in patients with chronic heart failure (2). In one study, elderly patients with chronic heart failure were estimated to take none of their therapy (digoxin) for an average of 111 d/yr (3). Importantly, in that study, there was a very large interindividual variability in adherence between different patients. The consequences of this nonadherence could be important since nonadherence is commonly present in hospitalized patients and improving adherence appears to reduce recurrent hospitalizations (2,4,5).
Identifying which patients are nonadherent with ACEI therapy may be possible by measuring in the patient’s bloodstream or urine whether the ACE enzyme is inhibited. There are, in fact, several ways to do this. First, one can measure serum ACE activity itself, which we have found to be a very sensitive indicator of whether an ACEI has been recently swallowed (6,7). Secondly, in 1997, Azizi et al. (8)suggested that the measurement of N-acetyl-seryl-aspartyl-lysl-proline (AcSDKP) is a better marker of adherence with ACE inhibition (8). This is an endogenous peptide that is metabolized by ACE, and ACE inhibition has been shown to induce AcSDKP in both plasma and urine (8). Thirdly, the ratio of angiotensin II (AII) to angiotensin I (AI) is probably the best index of whether ACE is inhibited in vivo, but the method to measure this is so cumbersome, that it is hard to imagine that this method could become a routine test to assess adherence with therapy unless it strongly outperformed the other methods (9). The main aim of our study was to perform a head-to-head comparison of these three biochemical techniques in their ability to detect nonadherence with ACEI therapy.
The problem of adherence is complex, however, because patients do not necessarily fall into the category of being either totally adherent or totally nonadherent. There are many different versions of partial compliance. The second aim of our study was therefore to compare the tests described above not only in the situation of full adherence and full nonadherence but also with various patterns of partial compliance.
Therefore we have not only compared three different ways of measuring adherence with therapy but have done so during various patterns of partial and full adherence.
Patient selection and protocol
Thirty-nine patients with chronic heart failure were recruited. The diagnosis of chronic heart failure was based on either echocardiographic or radionuclide evidence of LV systolic dysfunction (LVEF <45%). The study protocol was approved by our local research and ethics committee and written informed consent was obtained from all patients. All patients were receiving regular stable therapy with furosemide and lisinopril for at least two months before recruitment.
Each patient underwent the four seven-day treatment phases described below in a randomized fashion, interspersed with four weeks of their “normal” furosemide/lisinopril therapy. The four treatment phases were as follows: 1) full nonadherence: placebo therapy was substituted for their lisinopril for seven days (10); all other therapy was unchanged; 2) full adherence: lisinopril therapy (usual dose) was continued for seven days; all other therapy was unchanged; 3) clinic day adherence: placebo therapy was substituted for their lisinopril for six days followed by lisinopril (usual dose) on day 7 only; all other therapy was unchanged; and 4) half adherence: placebo therapy was substituted for their lisinopril on days 1, 3, 5 and 7; lisinopril therapy (usual dose) was taken on days 2, 4 and 6; and all other therapy was unchanged.
The patients attended on day 7 of each treatment period, at 3 to 6 h after their morning dose and having collected a 24-h urine sample. At each visit, they lay in supine position for 30 min, after which blood was taken for the measurement of serum ACE activity, plasma AcSDKP, plasma AI and plasma AII. The blood samples for AI and AII were collected into tubes containing 20 μmol/L of a renin inhibitor (H142) along with enalaprilat, neomycin, phenanthroline and EDTA. These substances, especially the renin inhibitor, were designed to prevent the in vitro generation of AII from the excess AI that was inevitably present due to the ACEI. All blood samples were collected on ice, centrifuged immediately at 4°C and the plasma or serum stored at −70°C prior to assaying them in batches.
Serum ACE activity was assayed by monitoring the change in absorbance at 340 nm of the hydrolysis of furylacrylolylphenylalanylglycylglycine (FAPGG) to FAP and GG (Sigma-Aldrich Chemical Company; Poole, Dorset GH12 4QH) on an analyzer (Roche MIRA Analyser; Roche Diagnostic Systems; Welwyn Garden City, Herts, AL7 3AY).
Plasma AI and AII were measured by Dr. J.J. Morton of the Department of Medicine and Therapeutics, University of Glasgow, using highly specific radioimmunoassays as described in more detail previously (11). Plasma and urine levels of AcSDKP were measured by a competitive enzyme immunoassay by Dr. Ezan in Paris using the technique described in more detail by those same workers (12,13).
Scattergrams were first constructed to visualize the data. For each biochemical test in isolation, two-way analysis of variance was then undertaken to see if the test was able to differentiate between the various patterns of adherence. Multiple comparisons were performed by Duncan’s test for homogenous subsets. Since all tests could readily distinguish between full nonadherence and all three other patterns of adherence, the main comparison between the tests was to see how each test performed in distinguishing full adherence from full nonadherence. To compare tests in this regard, we constructed ROC curves for each test and compared them. We also calculated the sensitivity, specificity, positive predictive accuracy and the negative predictive accuracy of each test as a way of comparing the various tests.
Our patients were typical patients with chronic heart failure (age 71 ± 6 years taking 75 ± 35 mg furosemide equivalents and 13 ± 6 mg lisinopril). There was no indication that the subsequent hormonal results varied according to the dose of furosemide or the dose of lisinopril.
Figure 1shows how each measurement performed on an individual basis. Analysis of variance was used to see how each test compared across a range of different forms of adherence. All tests were readily able to distinguish full adherence from full nonadherence. The two forms of partial adherence look indistinguishable from each other for all tests and they look similar to full adherence for all tests except plasma AcSDKP, in which partial adherence appears to produce values intermediate between full adherence and full nonadherence. The analysis of variance shown in Table 1confirms this impression, ie, for ACE, AII/AI ratio, AI and AII, full nonadherence was significantly different from full adherence, half adherence and from an ACEI on day 7 only, but there was no significant difference among the latter three. However, for plasma AcSDKP, there was not only a significant difference between full adherence and full nonadherence, but each of them was also significantly different from both forms of partial adherence. (Urine AcSDKP was not tested in its ability to detect partial adherence because the 24-h samples covered time periods when adherence was changing in the partial adherence situations. All patients did collect 24-h urine specimens on all treatment regimens so that they remained blind to which drug regimen they were receiving.)
To directly compare each test, we constructed ROC curves to compare the ability of all six tests to distinguish between full adherence and full nonadherence (Table 2). The area under the ROC was high for all tests and the only significant difference was between the very best (AcSDKP) and the very worst (plasma AI). Table 3shows the sensitivity and specificity of each test and again all tests performed well with a rank order of AcSDKP, (plasma and urine equal) ACE, AII/AI, AII and then AI. In absolute terms, both AcSDKP and ACE had sensitivities and specificities >85% and areas under ROC curves >0.9, which make both acceptable for clinical practice.
Our main conclusions are as follows. In general, all tests performed fairly well with a rank order of AcSDKP, ACE, AII/AI, AII and then AI. In particular, all tests above could distinguish full adherence from full nonadherence. Only AcSDKP could distinguish the two forms of partial adherence from both full adherence and full nonadherence.
The placebo tablets appeared identical to the lisinopril tablets, ie, this was a double dummy technique. From the ethical point of view, the replacement of lisinopril by placebo occurred only for a maximum of seven consecutive days and for a total of 17 days over 16 weeks. In the study of Pflugfelder et al. (10)in 1993, there was no clinical worsening of heart failure until treatment with an ACEI had been stopped for six weeks. We therefore believed that our patients were at very little (or no) risk when their lisinopril therapy was stopped for these brief periods.
It is widely recognized that nonadherence with therapy is an important practical problem (14). It is also widely appreciated that assessing adherence to therapy in the real world is notoriously difficult and that no method for assessing adherence will ever be perfect. Our proposal to take a blood sample from an unsuspecting patient has the major advantage that the patient is not alerted that his or her adherence is being monitored.
It is fortunate that AI, AII and the AII/AI ratio did not outperform AcSDKP or ACE, because the former are so technically difficult to measure that there are only three laboratories in Europe that can measure AII in the presence of an ACEI. As to AcSDKP and ACE, the former performs better because it can identify partial adherence. However, the performance of ACE is still fairly good and it does have the advantage that most routine hospital laboratories already measure ACE as a diagnostic test for sarcoidosis along with the fact that it is inexpensive and it can easily be automated on standard analyzers. The choice over whether to use AcSDKP or ACE in clinical practice will therefore depend on local circumstances.
Consequences of nonadherence
What are the consequences of nonadherence that might be preventable by targeting adherence-enhancing strategies? First, nonadherence probably leads to hospitalizations (2,4). Second, nonadherence leads to AII reactivation, which is known to produce a worse prognosis in terms of mortality and in terms of progression of LV dysfunction (15).
Our study applies only to lisinopril as the ACEI. It is uncertain how widely applicable these findings would be to other ACEIs. One particular ACEI in which serum ACE would not be a useful test is captopril because it uniquely has poor affinity for ACE in vitro that causes it to dissociate from ACE in a blood sample. This is why serum ACE is not able to be used to assess adherence when captopril is the prescribed ACEI. This specific problem with captopril does not appear to apply to the other measures of adherence (16). The other limitation in our study was that for ethical reasons, we switched patients to the placebo regimen for only seven days, and it is possible although unlikely that our biochemical measures had not fully returned to their baseline values. This seems unlikely to be a problem for serum ACE since serum ACE inhibition falls to only 40% at 72 h after lisinopril, which means that serum ACE inhibition is likely to be minimal by seven days (17). Furthermore, in our study, serum ACE, plasma AcSDKP and the AII/AI ratio were all in the expected population range after seven days of placebo therapy. A final limitation is that serum ACE was herein measured in samples that had been stored frozen, and freezing the samples is thought to cause a 20% increase in the absolute values for ACE. Therefore, the absolute cutoff value that we used in the present study for ACE may be a bit higher than would be the case if ACE was measured on fresh samples in routine practice. Our previous work with freshly assayed samples for ACE suggests that this is indeed the case (7).
Our study does not, in itself, answer all detailed questions on the practical use of these tests. One might think that the dose of the ACEI would be a factor in interpreting the assay results, but in 1998, van Veldhuisen et al. (18)found the same suppression of serum ACE with three different ACEI doses. Another factor that might be relevant to interpreting results is the patient’s ACE genotype but again at peak drug effect, serum ACE is only 3.6 U/L different between the II and DD genotype (19). Furthermore, in normal clinical practice, one does not know a patient’s genotype. The timing of the blood sample in relation to the timing of drug ingestion might also be important but with lisinopril, serum ACE suppression is 96% at 6 h and 80% at 24 h, which suggests that it might not matter that much when in the 24 h period the blood sample is taken since serum ACE suppression is fairly stable over this time period (18). Therefore, although these issues of detail ought to be formally addressed by future studies, it seems unlikely that dose, timing and genotype will have any major impact on the clinical utility of serum ACE as a measure of adherence, although as yet, little is known about these details and AcSDKP.
It is also worth emphasizing that the occasional high ACE activity values were not laboratory errors since quality control samples were run in all batches and long-term analytical variation was always <10%. Nevertheless, there were a few outlying points. These could be because some patients did not follow our instructions exactly, which again emphasizes the difficulties involved in doing research on adherence. However, any such problem would apply equally to all tests under comparison and we were anxious not to exclude any data. However, it is possible that clearly interpretable values will not be obtained in all patients in all circumstances. Nevertheless, the sensitivities and specifications of ACEI and AcSDKP are very encouraging.
Plasma and urine AcSDKP and serum ACE all perform well with a slight advantage for AcSDKP. Either of these techniques should now be assessed for its usefulness in targeting adherence-enhancing strategies toward nonadherent patients to see whether such targeted strategies can improve clinical events.
☆ This work was funded by the British Heart Foundation.
- angiotensin-converting enzyme
- angiotensin-converting enzyme inhibitor
- Angiotensin I
- Angiotensin II
- Received March 2, 1999.
- Revision received June 18, 1999.
- Accepted August 23, 1999.
- American College of Cardiology
- ↵Struthers AD, Anderson G, MacFadyen RJ, et al. Nonadherence with ACEI treatment is common in heart failure and can be detected by routine serum ACE Activity assays. Heart. In Press.
- Nussberger J,
- Brunner D,
- Keller I,
- Brunner H.R
- Pflugfelder P.W,
- Baird M.G,
- Tonken M.J,
- et al.
- Ezan E,
- Carde P,
- Le Kerneau J,
- et al.
- Corlett A.J
- Swedberg K,
- Eneroth P,
- Kjekshus J,
- et al.
- ↵Azizi M, Ezan E, Reny JL, et al. Renal and metabolic clearance of N-acetyl-seryl-aspartyl-lysyl-proline (AcSDKP) during ACE inhibition in humans. Hypertension. In Press.
- van Veldhuisen D.J,
- Genth-Zotz S,
- Braouwer J,
- et al.
- Davidson N.C,
- Coutie W.J,
- Webb D.J,
- Struthers A.D