Failure of Patients with Peripheral Arterial Disease to Accept the Recommended Treatment Results in Worse Outcomes Thomas E. Brothers, Charleston, South Carolina

Background: Strategies available to facilitate decision making for patients with peripheral arterial disease (PAD) include a Markov-based decision analysis (DA) model and the Lower Extremity Grading System (LEGS) score. Both have suggested inferior outcomes when the actual treatment received (ATX) differs from that predicted. This study focuses on patient outcomes when such discordance exists. Methods: All patients referred for symptomatic lower extremity PAD over a 3-year period were evaluated using the DA model and the LEGS score. Calculated quality of life (cQOL) values were assigned before treatment based on patient symptom, perfusion, and amputation status and at follow-up (range 1.000 [perfect health] to .000 [death]). The primary outcome of cQOL was compared according to whether the ATX matched that proposed by the surgeon or predicted by the DA model or LEGS score. Secondary outcomes for revascularized patients included major adverse limb event with perioperative death (MALE + POD) and amputation-free survival (AFS). Results: Among 375 procedures in 345 consecutive patients, the greatest improvement in cQOL at last follow-up (median 16 months) was observed with endovascular (0.23 ± 0.16, n ¼ 93) or open (0.21 ± 0.17, n ¼ 137) revascularization compared with primary amputation (0.10 ± 0.07, n ¼ 23) or medical therapy (0.04 ± 0.09, n ¼ 122). Multivariate regression showed discordance with the surgeon’s recommendation (P < 0.05) and/or the DA model (P < 0.05) to be independent predictors of improvement failure. ATX did not always agree with that proposed by the surgeon (89% agree, k ¼ 0.84), the DA model (68% agree, k ¼ 0.53), or the LEGS score (53% agree, k ¼ 0.32). Improvement in cQOL was greatest when ATX was concordant with treatment proposed by the surgeon (0.18 vs. 0.08, P < 0.01), the DA model (0.19 vs. 0.13, P < 0.01), or the LEGS score (0.23 vs. 0.10, P < 0.01). Patient refusal to follow the surgeon’s recommendations and continued smoking were associated with minimal improvement (cQOL ranges 0.05e0.07 and 0.00e0.02, respectively), while pursuing a less morbid procedure was associated with greater improvement (cQOL range 0.28e0.38). Among revascularized patients, MALE + POD was lower at 36 months after endovascular than open surgery (21% ± 5% vs. 36% ± 4%, P < 0.05), while AFS was not significantly different. Only discordance with the surgeon’s recommendation was an independent predictor of MALE + POD, possibly because of limitations in sample subset size. Conclusions: Mean cQOL improved most with direct revascularization, especially when the treatment received matched that predicted by the models or proposed by the surgeon. Type of treatment received was an independent predictor of agreement of treatment with recommendations. Patient refusal to follow the recommended treatment as well as the strategy not to revascularize claudicants who persist in smoking were associated with much less patient benefit from treatment.

Ralph H. Johnson Department of Veterans Affairs Medical Center, Medical University of South Carolina, Charleston, SC. Correspondence to: Thomas E. Brothers, MD, 25 Courtenay Drive, Charleston, SC, 29425-2950, Ralph H. Johnson Department of Veterans Affairs Medical Center, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC 29425-2950, USA; E-mail: [email protected]

244

Ann Vasc Surg 2015; 29: 244–259 http://dx.doi.org/10.1016/j.avsg.2014.08.006 Published by Elsevier Inc. Manuscript received: June 3, 2014; manuscript accepted: August 4, 2014; published online: October 8, 2014.

Vol. 29, No. 2, February 2015

Worse outcomes with failure to accept recommendations

INTRODUCTION

MATERIALS AND METHODS

The outcomes of treatment for peripheral arterial disease (PAD) vary greatly in terms of patient quality of life and are influenced by patient factors including comorbidities and preoperative functional status.1e3 However, decisions intended to maximum patient outcomes are complex and can be improved. At least 2 different models have been described to assist in decision making. First, a computerized Markov decision analysis (DA) model has been offered for patients with PAD, combining individualized probabilities for all potential outcomes for each treatment with patient-derived utility analysis to define expected quality of life.4,5 Using this model, it was observed that the treatment actually recommended by the surgeon to the patient frequently did not agree with the treatment mathematically predicted by the model to result in the best outcome.6 The DA model was subsequently expanded to include a broader spectrum of severity of PAD as well as adding the option of endovascular revascularization.7,8 An alternative route to decision making, the Lower Extremity Grading System (LEGS) score, derived by consensus from a panel of vascular specialists and validated on subsequent patients has also been reported to improve planning of therapy.9,10 The LEGS score appears to be less cumbersome than the Markov model, although it does not address primary medical therapy alone as an option. Interestingly, it was reported for both the DA model and the LEGS score that patients experienced inferior outcomes when the treatment that they actually received did not match that predicted as optimal.4,11 Satisfactory explanation for these observations has yet to be elucidated, and while it does not necessarily follow that universal concordance of treatment with these models would have led to better patient results, the potential for such tools to improve patient outcome remains intriguing. The objective of this study is to compare decisions regarding the therapy received by patients with PAD with the therapy proposed by the treating vascular surgeon, the DA model, and LEGS score to determine why such decisions might be discordant and how such discordance affects patient outcomes. The hypotheses to be tested were that decisions regarding the best therapy for PAD often differ among the patient, the treating vascular surgeon, and existing decision-making models and that patient outcome is affected by discordance with these decisions.

Research Design

245

The protocol for this retrospective caseecontrol study including waiver of consent was approved by the Institutional Review Board of the Medical University of South Carolina and the Research and Development Committee of the Ralph H. Johnson Department of Veterans Affairs Medical Center in Charleston. The primary outcome of interest was calculated quality of life (cQOL). The secondary outcomes for patients revascularized with endovascular or open surgery included major adverse limb event with perioperative death (MALE + POD) and amputation-free survival (AFS). As MALE is presumed to indicate major undesirable and even catastrophic outcomes from intervention, POD was included in this measure. Patients were compared according to whether their actual treatment received (ATX) was concordant with the therapy initially proposed by the surgeon (SURG) or that predicted as optimal by the DA model or LEGS score. For each group, controls were defined as those with concordant treatments, while cases were defined as those with discordant treatments. Data Collection All patients referred to one vascular surgeon over a 3-year period for management of symptomatic lower extremity PAD including both tertiary university and Veterans Administration practices were eligible for study inclusion and included patients who were managed with primary amputation or medical therapy alone. Patients without at least one follow-up encounter more than 30 days after the initial intervention (or encounter if medical therapy only) in which the results of therapy could not be ascertained were excluded from analysis. Exceptions were patients who died or had a MALE within 30 days of intervention. The medical records of study patients were abstracted for pertinent clinical information including location and severity of symptoms as well as anatomic and demographic information. This information was used to predict the optimum type of therapy using 2 published clinical vascular surgical decision-making tools, the DA model, and the LEGS score.4,9 In the current retrospective study, all patient treatment decisions were made before and independently of either model. The treatment recommended to the patient as well as that actually received was recorded, as were results including relief of ischemic symptoms,

246 Brothers

subsequent reintervention, above ankle amputations, graft or arterial patency, and death. Assessment was performed at the time of initial referral and each subsequent follow-up encounter. Patients were not stratified according to the degree or anatomic level of ischemia because of the limited sample size available.

DA Model The Markov-based, individualized vascular surgery DA model has previously been described.8 Briefly, the probability that any one of the 4 different treatments (primary amputation, medical therapy only, open surgery, or endovascular intervention) would result in a specific outcome is estimated by an experienced surgeon based on clinical experience and published literature. The presence of renal failure and heel ulcer further downgrade the expectation of limb salvage in the DA model, while the prediction of ability to ambulate with patent revascularization indirectly account for the presence of diabetes and extent of gangrene. Utility scores were assigned for each of these outcomes. Derivation of these utility values using a visual analog scale in patients with symptomatic PAD has previously been described and is provided in Appendix 1.4 The probabilities of outcome from an intervention as well as the utility values for these outcomes were then entered into the Markov analysis, which was performed using TreeAge Pro 2011 software (TreeAge Software, Inc, Williamstown, MA). Probabilities for average annual patient mortality were derived from actuarial tables supplied by the Social Security Administration (http://www.ssa.gov/ OACT/STATS/table4c6.html). Thirty-day perioperative mortality for open surgery or primary amputation was obtained from the Veterans Affairs Surgical Quality Improvement Program mortality calculator for 2013 (https://vaww.nso1.med.va.gov/vasqip/). Mortality for endovascular therapy was assumed to be zero. In this model, utility values were discounted at an annual rate of 5%, with negative utilities of 0.005 or 0.001 assigned for the performance of each additional open operation or endovascular intervention, respectively. The model was rolled back on a monthly cycle and terminated when 99.9% of the model was absorbed within the transitional state for death or at a maximum of 600 cycles (corresponding to 50 years after entry) in the unlikely event that the first condition was not met. Sensitivity analysis was not performed for each patient for alternative outcome probability estimates, as commitment to a single prediction estimate at

Annals of Vascular Surgery

the time and point of care was considered necessary for this exercise. LEGS Score The LEGS score was derived as described by Taylor et al.9 by grading each involved limb within the clinical categories of ‘‘Arteriographic Findings,’’ ‘‘Presentation,’’ ‘‘Functional Status,’’ ‘‘Comorbidities,’’ and ‘‘Technical Factors.’’ Points assigned to each category were combined to obtain the LEGS score. Point totals of 0e9 indicated preference for open surgical revascularization, 10e19 for endovascular intervention, and >20 for primary amputation. However, for the purposes of this study, all patients were assessed with the LEGS score, including highrisk operative candidates and claudicants who continued to use tobacco. cQOL Scores For this retrospective study based on chart documentation, prospective assessment of patient quality of life with existing instruments was not feasible. Instead, cQOL scores were assigned according to patient symptoms, limb perfusion status, limb salvage, type of intervention, wound healing, and ambulatory ability as well as the requirement for subsequent intervention and were considered to represent a generic measure of quality of life. These cQOL scores were based on prior published utility analysis of patient preferences for outcome, with anchors of 1.000 (perfect health), 0.732 (claudication), 0.456 (critical limb ischemia), and 0.000 (death) using visual analog scale methodology4 (Table I). Scores were assigned preoperatively and at each follow-up encounter according to symptoms, physical findings, and treatment results. cQOL values for each patient were assumed to be static over time unless there was a change in patient ischemic symptoms, ambulatory ability, or reintervention status, at which time the values were assumed to change as determined from patient history. ATX was recorded as the initial treatment provided. Patients who died during the course of follow-up were assigned a monthly score of 0.000 for each subsequent month after death until the date of study terminus (October 31, 2013). This allowed the outcome to be counted in the initial months after study entry in calculation of the average cQOL, but also acknowledged limitations in the long-term benefit of improvement in cQOL for the entire group. The reason for this was 2fold: first, when considering how aggressively to treat a patient for PAD, it is important to consider that patient’s expected survival. Intervention for

Vol. 29, No. 2, February 2015

Worse outcomes with failure to accept recommendations

247

Table I. Values assigned as cQOL according to prior treatment and anatomic and functional result Value

Treatment

Description of outcome

0.876 0.864 0.852 0.840 0.816 0.756 0.744 0.732 0.708 0.696 0.660 0.636 0.624 0.612 0.600 0.576 0.552 0.528 0.516 0.468 0.456 0.444 0.396 0.372 0.360 0.312 0.000

Medical Endovascular Endovascular Endovascular Bypass Bypass Bypass Medical Endovascular Medical Bypass Amputation Endovascular Amputation Endovascular Bypass Bypass Amputation Endovascular Bypass Medical Endovascular Bypass Amputation Endovascular Bypass Any

Improved, no rest pain, no additional procedures Patent, healed, no additional procedures Patent, healed, additional procedures, never thrombosed Patent, healed, additional procedures after thrombosis Patent, healed, no additional procedures Patent, healed, additional procedures, never thrombosed Patent, healed, additional procedures after thrombosis Stable claudication, no additional procedures Occluded, claudication only, no amputation Worsened claudication or additional procedures Occluded, no rest pain, no additional procedures Healed amputation, ambulatory, no additional procedures Healed amputation, ambulatory, no additional procedures Healed amputation, ambulatory, additional procedures Healed amputation, ambulatory, additional procedures Healed amputation, ambulatory, no additional procedures Healed amputation, ambulatory, additional procedures Healed amputation, not ambulatory Healed amputation, not ambulatory Healed amputation, not ambulatory Rest pain/gangrene, no amputation Rest pain/gangrene, no amputation Rest pain, gangrene, no amputation Not healed and/or rest pain, not ambulatory Not healed and/or rest pain, not ambulatory Not healed and/or rest pain, not ambulatory Death

claudication for a patient with a severely shortened lifespan because of comorbidities such as metastatic disease would infrequently be appropriate. Second, if a patient suffered early mortality but was ambulatory with a patent revascularization, it would still seem inappropriate to award this result as an improved outcome, when in fact the patient died.

outcomes of MALE and AFS were analyzed with KaplaneMeier life-table analysis using the logrank test. Agreement beyond chance alone between the treatment recommended and that actually received was assessed using the Kappa statistic.

RESULTS Statistical Analysis Chi-square test or Fisher’s exact test (as appropriate) was used to compare demographic, risk factor, worst ischemic symptom, level of intervention, and hybrid intervention data as well as reasons for discordance between treatment groups. Analysis of variance was used to compare cQOL scores. For this univariate analysis, the primary outcome (cQOL score) was reported according to patient status at last follow-up. Multivariate logistic regression was employed to identify independent predictors of concordance with therapy or patient outcomes using SPSS 12.0.0 for Windows software (SPSS, Inc, Chicago, IL). In addition, using data from all visits, patient status was also integrated to generate mean cQOL and mean change in cQOL from baseline. Secondary

Three hundred fifty-nine consecutive patients were eligible for participation, among whom no follow-up data were available in 14 patients, all of whom had been advised to have medical therapy alone. This left 345 patients with 375 initial treatments for study enrollment and evaluation, with a median 16 months (range 1e38 months) of follow-up. Medical therapy alone was provided in 122 (32%) patients, open revascularization in 137 (37%), endovascular therapy in 93 (25%), and primary amputation in 23 (6%). Patient demographic information, risk factors, and presenting ischemic symptoms according to initial treatment information are listed in Table II. Hybrid procedures were common; 22% of open surgery also involved endovascular therapy, while 32% of endovascular

248 Brothers

Annals of Vascular Surgery

Table II. Demographics, risk factors, and presenting ischemic symptoms according to initial treatment Actual treatment received

Age (IQR) Follow-up months (IQR) Male Black Current smoking Type I diabetes Type II diabetes Hypertension Hyperlipidemia Cardiac disease Renal insufficiency COPD Claudication Rest pain Ischemic ulcer Gangrene Elective Suprainguinal Infrainguinal Hybrid procedure

Primary amputation (n ¼ 23)

Medical only (n ¼ 122)

Open surgery (n ¼ 137)

Endovascular (n ¼ 93)

72 (59e76) 23 (10e35)

65 (59e71) 21 (16e31)

64 (59e70) 28 (17e35)

65 (61e70) 25 (14e34)

83% 78% 17% 61% 13% 78% 17% 48% 26% 9% 0%a 9%a 13% 83% 48% d d d

85% 36% 40% 15% 18% 74% 64% 48% 15% 17% 77%a 10%a 11% 9% 97% d d d

81% 39% 23% 23% 12% 79% 38% 35% 8% 10% 32% 41% 16% 12% 72% 21% 79% 25%

75% 44% 11% 31% 15% 78% 45% 38% 18% 3% 56%b 22%b 12% 14% 78% 45% 62% 30%

P value

0.267 0.034 0.326 0.002

Failure of patients with peripheral arterial disease to accept the recommended treatment results in worse outcomes.

Strategies available to facilitate decision making for patients with peripheral arterial disease (PAD) include a Markov-based decision analysis (DA) m...
843KB Sizes 0 Downloads 6 Views