This study examined factors that influenced the allocation of patients to different treatments for CAD. Three sets of analyses were conducted comparing actual decisions to undergo: PTCA rather than receive medication, CABG rather than PTCA, and CABG rather than medication.
Univariate analyses revealed a wide range of differentiating factors including, demographic variables, co-morbidities, family histories, symptom reports, functional limitations and angiogram indices (i.e. objective disease reports). While many of these factors are in the treatment guidelines (e.g. co-morbidities, angiogram indices); others are not (e.g. educational level, patient rated functioning or symptom reports), but appear to be, on empirical analysis, implicitly involved in the decision processes, particularly in analyses involving comparisons with PTCA. This indicates that clinicians are undoubtedly aware that multiple factors outside of those explicitly discussed in guidelines (which are usually technical and clinical in nature), should be addressed and guide decision making about treatment modality.
Gender has long been a controversial issue in CAD and much evidence has accumulated regarding poorer outcomes for women [16
]. Research investigating, gender differences in patients undergoing evaluation for CAD has found contradictory results but generally suggested that resource utilisation may be lower in women [17
]. In a large study, Miller et al. [20
] did find less coronary angiographies performed in women being assessed for CAD and lower referral for revascularisation in univariate analysis. Although, in this study, females were almost 70% less likely than males to have CABG when compared to medication, this effect was not robust enough to be retained in the multivariate analysis, suggesting that gender is not a primary concern during this decision process, between the two most disparate treatments. Further, in the other analyses there was no evidence to suggest a bias for males to receive particular interventions. However, these findings must be interpreted in light of the small number of female participants, which may limit the power to find gender effects.
Previous research has shown that mortality increases with age in both CABG and PTCA [21
]. In the present study age was not a significant factor in determining CABG from the other two treatments. The CABG results were unexpected but the increased risks may have been counter-balanced by judgements regarding graft patency and life expectancy. In contrast, an increase of ten years in age nearly doubled the likelihood of medication over PTCA in the multivariate analysis. While this may reflect a natural tendency to perform less invasive procedures in the elderly, it is also important that the PTCA group had significantly higher levels of education in comparison to both medication and CABG groups. One may speculate that the younger more educated individuals may be both more aware of and willing to try newer treatments and possibly more actively involved in their treatment decisions. It is however important that the PTCA group were also distinguished by their greater experience of CAD treatments and possible vicarious knowledge/experience from family members who had a history of IHD. PTCA patients were more likely to have received a previous intervention (mainly PTCA's) than both the medication and CABG patients in both the univariate and multivariate analyses. Given the relatively high rates of restenosis following PTCA, this finding is not unexpected. The previous experiences may have provided patients with clearer expectations upon which to base and participate in the decisions regarding their treatment.
Variables concerning patient history and co-morbidities appeared to appropriately reduce the likelihood of the more interventional treatments. Hypertension had a tendency to increase the odds of remaining on/starting medication. In the PTCA versus medication analysis, this effect was retained in the multivariate analysis. Hypercholesterolemia reduced the likelihood of CABG as opposed to medication, and the effects of a family history of neurological problems reduced the likelihood of CABG from PTCA. Less obviously a family history of IHD, reduced the odds of PTCA when compared to both medication and CABG. However, none of these findings were retained in the multivariate analyses.
The measures from the angiogram reports demonstrated a strikingly consistent effect in predicting the likelihood of CABG, as would be expected and is supportive of the guidelines. Within the univariate analysis, all five of the measures (levels of LAD, circumflex and right coronary artery disease, the number of disease arteries and the level of ventricular disease) increased the likelihood of CABG. The per unit increase in these measures led to a multiplication in odds of CABG ranging from 1.6 to 5.4 times in the medication versus CABG analyses, and between 1.5 and 2.5 times for the PTCA versus CABG analysis. For the most part, the odds ratios determining CABG from medication were larger than those determining CABG from PTCA, as would be expected. The other significant result from the univariate analyses of the angiogram data showed that increasing LAD artery disease predicts a greater likelihood of PTCA from medication; a finding consistent with treatment guidelines.
The multivariate analyses helped determine which of the angiogram measures was the most relevant to the treatment decision process. The pattern of results indicated that with increased LAD disease, medication was not considered a sufficient treatment; the likelihood of PTCA or CABG was increased. Moving from mild to moderate LAD disease increased the odds of PTCA from medication by ≈ 2.6 times, and the odds of CABG from medication by approximately twice this (5 times). Increasing LAD disease did not significantly alter the odds of CABG in comparison to PTCA.
Increasing levels of disease in the circumflex artery increased the likelihood that CABG would be the chosen option. A move from mild to moderate levels of circumflex artery disease more than doubled the odds of CABG from PTCA and almost trebled the odds of CABG as opposed to medication. Interestingly, the number of diseased arteries, which is often an indication for CABG rather than PTCA [6
], was not retained in the multivariate model. These results are reassuring in that they confirm the important role of the angiogram reports when deciding whether to proceed with interventional strategies, in particular CABG.
Patient reports of the frequency of their symptoms, and the manner in which their poor health is affecting their lives, are the customary way in which patients present the condition to their health care professionals. High levels of angina symptom reports and greater functional limitations increased the likelihood of receiving either of the two more invasive procedures. They did not, however, significantly affect the decision between PTCA and CABG. The symptom reports that appeared to drive the decisions were specific and non-cardiac symptoms did not affect the treatment decision. In the multivariate analysis, the frequency of angina was retained when comparing PTCA to medication, such that moving from occasional to frequent angina increased the odds of PTCA by ≈ 2.4 times. The finding that only one symptom measure entered each multivariate analysis is likely to be due to the degree of correlation between the various symptom report measures (range from 0.731 to 0.225 all p < 0.001).
Reports of restrictions on activities of daily living were also found to influence treatment decisions. Specifically, greater limitations in social behaviours and mobility increased the likelihood of receiving CABG, from both PTCA and medication. When entered into the multivariate analysis, the illnesses effect on mobility was the only measure of activities of daily living that was retained; increasing the odds of CABG from medication and PTCA by 80% and 60% respectively, for a unit increase in mobility range problems. This suggests that mobility restrictions are an important trigger to provoke more invasive treatments by cardiologists.
Overall each category of information was represented within the multivariate analyses, although all the categories were not represented in each of the analyses. This suggests that all types of information are being considered during the decision processes, although some may only be evident when examined within specific treatment comparisons. Two distinct patterns of outcomes can be discerned from the cumulative results of the three sets of analyses. First, the increased likelihood of CABG was primarily based on reports of greater objective disease (angiograms) and functional limitations. Second, the data suggests that decisions concerning PTCA can be interpreted to be driven by patient characteristics which may reflect the influence of patient preferences although these were not directly measured in this study. The findings reveal that treatment decisions are subject to a range of different influences and not simply guided by protocols.
It must be noted that treatment modalities/technologies and the guiding protocols are changing constantly. The management of coronary disease has changed over the period since the participants for this study were recruited (e.g. as many as three times more percutaneous coronary interventions are now done in the UK). However, although the manner in which factors are addressed in allocating patients to treatments modalities during the decision making process has altered in guidelines, the specific factors that are considered have not changed significantly in the recent updates [4
Potential limitations to this study include the relatively small sample of participants, which were predominately male and drawn from a limited geographical region, taken from a small pool of cardiologists; resulting in a highly selective patient population (with low incidences of some co-morbidities). Each of these factors may limit the generalisability of results (e.g. the small number of female participants may have precluded finding significant gender effects). A larger multi-centre study, with stratified sampling may help to resolve some of these issues, and it is acknowledged that both geographical and individual consultant factors may confound treatment decisions. For instance there may be a proclivity to perform interventional cardiology in certain regions and/or that individual cardiologists may be referred particular patients, and have preferences for particular treatments [23
]. More detailed analysis of such factors with a larger study sample may help to unravel how such factors confound treatment decision processes.
Additionally, a limitation of observational studies, such as the current one, is that they cannot exclude as explanations other unmeasured factors [11
]. It may be necessary to conduct further studies to see how recent changes in treatment and technology effects decision making. This study provides both a basis for this and evidence that treatment decisions are subject to a range of different influences including those in protocols.