- Email: [email protected]
Predictors of compliance with surveillance after endovascular aneurysm repair and comparative survival outcomes
From the Midwestern Vascular Surgical Society
Predictors of compliance with surveillance after endovascular aneurysm repair and comparative survival outcomes Chris Y. Wu, BS, Huiting Chen, MD, Katherine A. Gallagher, MD, Jonathan L. Eliason, MD, John E. Rectenwald, MD, and Dawn M. Coleman, MD, Ann Arbor, Mich Objective: Although imaging surveillance is mandatory for all patients after endovascular aneurysm repair (EVAR), many patients are not compliant with follow-up. We sought to determine predictors of compliance with EVAR surveillance and to examine how compliance with current surveillance protocols correlates with survival. Methods: We analyzed 188 patients who underwent EVAR at our institution for infrarenal abdominal aortic aneurysms (AAAs) between 2001 and 2011. The primary end point was compliance with post-EVAR surveillance recommendations. Univariate analysis included patient demographics and socioeconomic information, AAA characteristics, EVAR hospital course variables, late complications and secondary interventions, length of follow-up, smoking status, family history of AAA, driving distances, primary care providers, and medical comorbidities. Mortality was determined by the Social Security Death Index. Multinomial logistic regressions were fit to identify independent predictors of compliance. Survival plots were generated with the Kaplan-Meier method and compared with the log-rank test. Univariate and multivariate Cox regression analysis was used to determine effect of compliance on survival after adjusting for confounders. Results: Of 188 patients, 89 (47.3%) were 100% compliant with follow-up visits and imaging, 21 (11.1%) were moderately compliant by missing appointments, and 78 (41.4%) were lost to follow-up completely. Overall median age was 74 years, and 81.9% of patients were male. Late complications occurred in 77 patients (40.9%), secondary interventions were performed in 32 patients (17%), and 5-year mortality was 21.2%. Mean follow-up interval was >40 months for 100% compliant and moderately compliant patients and <20 months for those lost to follow-up (P < .0001). In adjusted analysis, late complications (odds ratio [OR], 2.71; 95% confidence interval [CI], 1.32-5.55; P [ .007), absence of social work consultation (OR, 2.43; 95% CI, 1.12-5.27; P [ .024), and family history of AAA (OR, 2.67; 95% CI, 1.06-6.75; P [ .037) were associated with 100% compliance, whereas shorter driving distances (P [ .051) and shorter hospital stay (P [ .056) approached significance. Transient ischemic attack or stroke (OR, 3.59; 95% CI, 1.18-10.91; P [ .024) was the only variable independently associated with moderate compliance. Compared with patients lost to follow-up, 100% compliant patients had worse survival (log-rank test, P [ .033), whereas moderately compliant patients’ survival was not significantly different (log-rank test, P [ .149). In adjusted Cox regression analysis, 100% compliant patients had decreased survival duration (rate ratio, 2.67; 95% CI, 1.18-6.06; P [ .018) compared with those lost to follow-up. Conclusions: Follow-up surveillance is incomplete for more than half of patients who undergo EVAR at our institution, and patient compliance can be predicted by covariates mentioned before. Compliance with current surveillance regimens does not confer a survival benefit. Further research individualizing surveillance protocols based on risk level of late complications and noncompliance and prospective studies examining resulting survival benefits of compliance are warranted. (J Vasc Surg 2015;-:1-9.)
Endovascular aneurysm repair (EVAR) has progressively replaced open aneurysm repair for the treatment of infrarenal abdominal aortic aneurysms (AAAs) since its introduction two decades ago because of reduced operative morbidity, shorter intensive care unit (ICU) and hospital From the Department of Surgery, Section of Vascular Surgery, University of Michigan. Author conflict of interest: none. Presented as a poster at the Thirty-eighth Annual Meeting of the Midwestern Vascular Surgical Society, Coralville, Iowa, September 4-6, 2014. Reprint requests: Dawn M. Coleman, MD, University of Michigan Cardiovascular Center, 1500 E Medical Center Dr, Ann Arbor, MI 48109-5867 (e-mail: [email protected]). The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest. 0741-5214 Copyright Ó 2015 by the Society for Vascular Surgery. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jvs.2015.02.023
length of stay, and lower perioperative mortality rates.1-4 Despite its increasing popularity, EVAR is associated with a high rate of secondary interventions, which is necessary in up to 8.7% of patients at 12 months with a cumulative incidence of 14% at 4 years.5 Endoleak is the most frequent complication after EVAR, seen in up to 25% of patients during follow-up.1 Late rupture is an additional risk with an incidence of 1.5%.6 As such, the Society for Vascular Surgery practice guidelines recommend imaging surveillance after EVAR to detect these rupture-predisposing late complications.1 Research regarding the optimal modality and schedule of surveillance is ongoing, although computed tomography angiography (CTA) is typically used, which subjects the patient to chronic exposure to radiation and iodinated contrast agents. Compliance with post-EVAR imaging surveillance remains suboptimal. Compliance rates range in the literature, and risk factors for such remain unclear. Jones et al7 reported that nearly one third of patients had incomplete 1
2 Wu et al
JOURNAL OF VASCULAR SURGERY --- 2015
Fig 1. Flow diagram showing selection of patients. EVAR, Endovascular aneurysm repair; PCP, primary care provider.
follow-up and that these patients had shorter follow-up intervals (34.7 vs 18.8 months; P < .001) and an increase in major adverse events (6.1% vs 0.5% in compliant patients), concluding that patients deemed at high risk for noncompliance be advised against EVAR. Leurs et al8 identified smoking, hyperlipidemia, unfitness for open surgery or anesthesia, and presence of late complication as significantly associated with compliance patterns; in addition, after 84 months of follow-up, compliant patients had higher mortality rates, leading them to conclude that patients with more medical comorbidities received more intensive follow-up. Kret et al,9 during a median follow-up period of 28 6 10.5 months, found that only 44% of patients were compliant with follow-up, and in this cohort the presence of late complications was the only significant association. These authors also found that compliance was associated with decreased overall survival, a result further confirmed by Waduud et al.9,10 Building on these referenced studies, we attempted to identify clinically useful predictors of post-EVAR surveillance compliance as well as to comprehensively examine the effect of compliance on patient survival. METHODS Study cohort. The University of Michigan Institutional Review Board approved this study (No. 00076472), and data were obtained retrospectively through review of electronic medical records. Patient consent was not required by our institution’s Institutional Review Board, and all patient data were stored without identifiers. All consecutive patients who underwent EVAR between 2001 and 2011 at the University of Michigan performed by the Section of Vascular Surgery were included from a prospectively maintained database. Fig 1 illustrates exclusion criteria, which included patients who died within 1 year of endograft placement, those who had EVAR performed for non-AAAs, those who were Veterans Affairs patients referred to the University of
Michigan for EVAR because of case complexity or equipment malfunction, those who were dialysis dependent or transplant patients (who have care coordinators that may confound compliance rates), and those for whom local or primary care follow-up was arranged (whereas these patients may have obtained some form of surveillance imaging with their primary care providers [PCPs], there is no evidence that the images were reviewed by a vascular surgeon or that the PCPs were aware of surveillance guidelines). Outcomes and covariates. Institutional practice pattern secures 1-month follow-up after EVAR before hospital discharge. Surveillance imaging is performed at this visit, usually in the form of CTA. Although it is subject to individual surgeon variability, surveillance imaging is then performed at 12 months after EVAR and then yearly thereafter, with more frequent imaging performed for complicated cases or those with identified endoleaks. Other forms of surveillance, such as ultrasound or radiography, were also acceptable. When patients miss surveillance appointments, clinic nurses contact them by mail and by phone multiple times. In addition, there would be intermittent correspondence with patients’ PCPs, depending on the attending physician. Compliance with vascular surgeons’ recommendations was the main outcome of interest. All but one of our vascular surgeons followed the described surveillance protocol. The one surgeon who deviated would, in certain cases, obtain imaging surveillance between longer intervals. Patients were considered 100% compliant with follow-up surveillance if they followed all of their vascular surgeon’s recommendations regarding surveillance imaging. Patients were considered moderately compliant if they missed appointments or surveillance imaging (either one appointment or multiple ones) but continued to follow up thereafter. Patients were considered lost to follow-up if they had no further vascular surgical follow-up, surveillance imaging, or further documentation beyond their last
JOURNAL OF VASCULAR SURGERY Volume -, Number -
missed appointment (despite the recommendation of their vascular surgeon to continue surveillance). Patients who missed clinic visits but still obtained surveillance imaging and patients who obtained imaging remotely were considered compliant if their vascular surgeon documented review of the imaging. Covariates extracted from the medical record included: patient age at the time of operation; gender; preoperative AAA characteristics, such as diameter, ruptured, inflammatory, or dissecting; length of hospital and ICU stay; presence of endoleak on completion angiography; American Society of Anesthesiologists (ASA) physical status score; presence of social work consultation, which was used as a surrogate for an unstable socioeconomic environment; family history of AAA; smoking history; home zip code; having a PCP; ethnicity; and insurance status. Driving distance from the patient’s home to the University of Michigan was calculated using open-access online software. Patient mortality and, if applicable, date of death was determined using open-access online Social Security Death Index databases. Medical comorbidities were extracted from the electronic record. Follow-up intervals were calculated as months from time of operation to last recorded vascular surgery clinic visit or surveillance imaging. Late complications were defined as endograft-related complications that occurred after initial hospital discharge, including endoleak, endograft or limb occlusion or stenosis, endograft kink, migration, late rupture, and endograft infection. Secondary interventions, defined as any open or endovascular procedures performed to fix graft-related complications, were recorded. Statistical analysis. All statistical analyses were performed with commercially available software: IBM SPSS Smartreader version 20 (IBM Corporation, Armonk, NY). The primary outcome of analysis was compliance with the vascular surgeon’s recommendations regarding EVAR surveillance as previously described. Univariate analyses were used to determine which covariates were associated with compliance. Continuous variables were assessed by the independent t-test or analysis of variance as appropriate, and categorical data were assessed by the c2 test or Fisher exact test as appropriate. We then fit separate multinomial logistic regression models for the 100% compliant and moderately compliant groups using the patients lost to follow-up as the reference group. In the initial multinomial regression models, we included all covariates with P value < .2 from the univariate analysis to adjust for potential confounders. By backward elimination, we excluded covariates not significantly associated with the outcomes of interest (defined as P value > .2) from subsequent iterations of the model. Final results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Adjusted covariates with P values < .05 were considered significant. The Kaplan-Meier technique and the log-rank test were then used to examine patient survival in each compliance group. The Cox proportional hazards model was used to identify univariate and multivariate predictors of survival.
Wu et al 3
Unadjusted and adjusted rate ratios (RRs), 95% CIs, and P values were calculated. All covariates associated with compliance identified from the multinomial analysis as well as covariates known to affect survival, such as baseline medical comorbidities, various patients’ demographics, and case complexity, were included in the Cox model. RESULTS Cohort. In total, 188 patients met selection criteria for inclusion. There was a male predominance (n ¼ 154; 81.9%), with a median age of 74 years (interquartile range [IQR], 67-79 years) and mostly white ethnicity (n ¼ 182; 96.8%). Patient demographics and risk factors are described in Table I. Eighty-nine patients (47.3%) were 100% compliant with post-EVAR surveillance, 21 patients (11.1%) were moderately compliant, and 78 patients (41.4%) were lost to follow-up. One hundred eighty patients (95.7%) carried some form of insurance (private, public, or both), and 182 patients (96.8%) had a PCP. Patients traveled a median distance of 43 miles (IQR, 27-76 miles) for care. Late complications were detected in 77 patients (40.9%; Table II), and secondary interventions were performed in 32 patients (17%). Overall median follow-up length was 25 months (IQR, 9-45 months). Of 188 patients, there were 40 deaths within 5 years after surgery (21.2% 5-mortality rate). None of the fully compliant patients were enrolled in EVAR clinical trials. Four patients in the moderately compliant or lost to follow-up cohort were initially in endograft device clinical trials. Univariate analysis. Table I reveals the univariate analysis results comparing covariates of interest between the 100% compliant, moderately compliant, and lost to follow-up patients. Variables significantly associated with compliance include length of hospital stay (P ¼ .04), average length of follow-up (P < .0001), active smoking (P ¼ .045), family history of AAA (P ¼ .034), and TIA or stroke (P ¼ .007). Social work consultation (P ¼ .077), ASA physical status (P ¼ .832), late complications (P ¼ .101), secondary interventions (P ¼ .234), medical comorbidities except for TIA and stroke, uninsured status (P ¼ .848), PCP (P ¼ .309), and driving distance (P ¼ .078) were not significantly associated with compliance outcomes. Multinomial analysis. Table III reveals variables associated with 100% compliance using the patients lost to follow-up as a reference group. All covariates from univariate analysis (Table I) with P values < .2 were included in the original multinomial regression model. Inflammatory AAA, ICU stay, active smoking, hyperlipidemia, and transient ischemic attack (TIA) or stroke were excluded from subsequent iterations because of nonsignificance. Variables achieving significant correlation with 100% compliance include late complications (P ¼ .007), absence of social work consultation (P ¼ .024), and family history of AAA (P ¼ .037). Variables approaching significance include shorter driving distance (P ¼ .051) and shorter length of hospital stay (P ¼ .056). Table IV reveals variables correlating with moderate compliance using the patients lost to follow-up as a
JOURNAL OF VASCULAR SURGERY --- 2015
4 Wu et al
Table I. Univariate analysis for 188 patients who underwent endovascular aneurysm repair (EVAR) between 2001 and 2011 Variable Age, years Gender Male Female ASA physical status, grade AAA diameter, cm Ruptured AAA Dissecting AAA Inflammatory AAA Endoleak on completion angiography Length of hospital stay, days ICU stay, days Absence of social work consultation Average length of follow-up, months Late complications Secondary intervention Active smoking Family history of AAA Hypertension Hyperlipidemia Coronary artery disease Congestive heart disease COPD Diabetes mellitus Chronic kidney disease Peripheral vascular disease TIA or stroke Uninsured PCP Minority race Driving distance, miles
100% Compliant (n ¼ 89)
Moderately compliant (n ¼ 21)
Lost to follow-up (n ¼ 78)
73.69 6 1.74
72.19 6 3.65
73.05 6 1.88
73 (82) 16 (18) 3.23 6 0.11 5.81 6 0.24 6 (6.7) 2 (2.2) 3 (3.4) 24 (27) 2.55 6 0.55 6 (6.7) 68 (76.4) 40.86 6 5.73 43 (48.3) 19 (21.3) 19 (21.3) 23 (25.8) 75 (84.3) 65 (73) 43 (48.3) 12 (13.5) 26 (29.2) 13 (14.8) 13 (14.6) 16 (18) 15 (16.9) 3 (3.4) 88 (98.9) 2 (2.2) 52.79 6 9.20
17 (81) 4 (19) 3.28 6 0.23 5.73 6 0.46 1 (4.8) 0 (0) 2 (9.5) 4 (19) 3.52 6 0.97 4 (19) 15 (71.4) 42.08 6 7.34 9 (42.9) 4 (19) 2 (9.5) 2 (9.5) 16 (76.2) 17 (81) 11 (52.4) 3 (14.3) 7 (35) 5 (23.8) 6 (28.6) 4 (19) 10 (47.6) 1 (4.8) 20 (95.2) 1 (4.8) 80.27 6 37.61
64 (82.1) 14 (17.9) 3.20 6 0.11 5.97 6 0.29 5 (6.4) 0 (0) 1 (1.3) 23 (29.5) 5.41 6 2.36 12 (15.4) 47 (60.3) 18.59 6 5.20 25 (32.1) 9 (11.5) 26 (33.3) 9 (11.7) 66 (84.6) 47 (60.3) 43 (55.1) 11 (14.1) 25 (32.1) 13 (16.7) 18 (23.1) 14 (18.2) 15 (19.2) 4 (5.1) 74 (94.9) 3 (3.8) 95.74 6 39.58
P value .734 .993 .832 .606 .946 .325 .161a .633 .040a .12a .077a <.0001 .101a .234 .045a .034a .63 .09a .678 .991 .851 .606 .216 .993 .007a .848 .309 .766 .078a
AAA, Abdominal aortic aneurysm; ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; PCP, primary care provider; TIA, transient ischemic attack. Continuous variables are displayed as mean 6 95% confidence interval and categorical variables as number (%). a P value < .2 will be included in multinomial analysis (excluding mean follow-up intervals).
Table II. Incidence of each late complication occurring in 77 patients
Table III. Multinomial logistic regression analysis results showing independent predictors of 100% compliance with recommended surveillance protocols
No. (%) Type I endoleak Type II endoleak Type III endoleak Type IV endoleak Type V endoleak Graft limb occlusion or stenosis Graft kink Graft migration Late rupture Graft infection
12 47 3 1 4 11 2 2 2 1
(6.3) (25) (1.6) (0.5) (2.1) (5.9) (1.1) (1.1) (1.1) (0.5)
reference group. Late complications, length of hospital stay, ICU stay, absence of social work consultation, family history of AAA, and driving distance were excluded after the first iteration of the multinomial regression analysis because of nonsignificance. The only significant association was TIA or stroke (P ¼ .024). Insignificant variables
Variable
OR
95% CI
Late complications 2.71 1.322-5.559 Absence of social work consultation 2.43 1.127-5.270 Family history of AAA 2.67 1.062-6.753 Driving distance, miles 0.99 0.986-1 Length of hospital stay 0.9 0.814-1.003
P value .007a .024a .037a .051 .056
AAA, Abdominal aortic aneurysm; CI, confidence interval; OR, odds ratio. a Significant.
included hyperlipidemia (P ¼ .074), inflammatory AAA (P ¼ .077), and active smoking (P ¼ .110). Survival analysis. Table V reveals number of mortality events in each compliance group and number of patients censored. The 100% compliant patients experienced 29 events (deaths) during the study period, moderately compliant patients experienced 6 events, and lost to followup patients experienced 20 events.
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Wu et al 5
Table IV. Multinomial logistic regression analysis results showing independent predictors of moderate compliance Variable TIA or stroke Hyperlipidemia Inflammatory AAA Active smoking
OR
95% CI
P value
3.59 3.44 13.3 0.27
1.185-10.917 0.888-13.375 0.758-235.1 0.057-1.34
.024a .074 .077 .110
AAA, Abdominal aortic aneurysm; CI, confidence interval; OR, odds ratio; TIA, transient ischemic attack. a Significant.
Table V. Kaplan-Meier survival analysis showing number of patients and events (deaths) in each group Censored Groups 100% compliant Moderately compliant Lost to follow-up
Patients
Events
No.
%
89 21 78
29 6 20
60 15 58
67.4 71.4 74.4
Figs 2 and 3 demonstrate survival plots comparing 100% compliant and moderately compliant groups with the lost to follow-up group. The 100% compliant patients had significantly worse survival compared with the lost to follow-up patients (P ¼ .033) as determined by the log-rank test, whereas no survival differences were found between moderately compliant patients and lost to follow-up patients (P ¼ .149). Table VI reveals univariate and multivariate predictors of shorter survival duration. In the univariate analysis, 100% compliance (RR, 1.9; 95% CI, 1.05-3.43; P value, .034), ASA physical status (RR, 1.82; 95% CI, 1.09-3.03; P value, .02), and age (RR, 1.07; 95% CI, 1.03-1.1; P value, .0001) were significantly associated with shorter survival duration. In adjusted analysis, 100% compliance (RR, 2.67; 95% CI, 1.18-6.06; P value, .018), ASA physical status (RR, 3.2; 95% CI, 1.57-6.53; P value, .001), age (RR, 1.11; 95% CI, 1.05-1.17; P value, <.0001), and social work consultation (RR, 3.24; 95% CI, 1.44-7.27; P value, .004) were independently associated with shorter survival duration. DISCUSSION The objective of this study was to identify clinically useful predictors of post-EVAR surveillance compliance as well as to comprehensively examine how compliance correlates with survival. We found that 47.3% of patients were 100% compliant, 11.1% of patients were moderately compliant, and 41.4% of patients were lost to follow-up. This is concordant with prior studies, which identified that only 35% to 67.2% of patients were deemed to have complete follow-up after EVAR.7-10 In addition, we found that late complications, stable socioeconomic situations, family history of AAA, and TIA or stroke were independently associated with either full or moderate compliance.
Finally, in our survival analysis, we found that 100% compliant patients had decreased survival duration, a result that was consistent after adjusting for confounders in a multivariate Cox regression analysis. Jones et al published on 302 patients and found that mean follow-up length was increased in the compliant group; however, 5-year intervention rates, 5-year survival, and incidence of known endoleak were all found to be similar between groups. Major events that required urgent surgery were increased in the noncompliant group (6.1% vs 0.5%; P ¼ .006). They also found that follow-up duration was better in patients enrolled in clinical trials, which they attributed to the persistence of research nurses. They concluded that patients at risk for compliance should be advised against EVAR. Our results affirm several of these conclusions as we also found that follow-up duration was better in patients who had follow-up (>40 months mean follow-up interval for 100% compliant and moderately compliant patients vs <20 months mean follow-up interval for lost to follow-up patients; P < .0001). We were unable to assess cause of mortality in the lost to follow-up group. Two studies have sought to determine predictors of compliance with post-EVAR surveillance. In their study of 4433 patients enrolled in the European Collaborating Group on Stent-Graft Techniques for Abdominal Aortic Aneurysm Repair (EUROSTAR) who underwent EVAR between 1996 and 2004, Leurs et al8 identified certain comorbidities (active smoking, hyperlipidemia, and unfitness for open surgery or general anesthesia) as predictive of compliance. In agreement with our study, complications during follow-up were significantly more common in the compliant group. Cumulative survival rates were found to be worse in the compliant group (71% vs 74%; P < .0001). These results were confirmed in 2013 by Kret et al,9 who reported that during a median follow-up of 28 6 10.5 months, only late complications were associated with better compliance. Medical comorbidities and socioeconomic factors including distance from the patient’s home zip code to the medical center’s zip code were not significantly associated with compliance. They also found that noncompliant patients had improved survival outcomes (P < .001). These counterintuitive results positively associating compliance with increased mortality were confirmed by Waduud et al,10 who found that patients with incomplete follow-up had improved survival duration compared with the complete follow-up group (P < .001) in Cox regression analysis; however, their multivariate analysis adjusted only for age, gender, location, AAA diameter, and year of EVAR, leaving out variables known to affect mortality, such as medical comorbidities and case complexity. Similar to these studies examining post-EVAR surveillance compliance, we found that late complications (OR, 2.71; P ¼ .007) were independently predictive of 100% compliance. In addition, we found that a stable socioeconomic situation (using absence of social work consultation as a surrogate; OR, 2.43; P ¼ .024) and family history of AAA (OR, 2.67; P ¼ .037) were independently predictive
6 Wu et al
JOURNAL OF VASCULAR SURGERY --- 2015
Fig 2. Kaplan-Meier survival curve representing 100% compliant patients with patients completely lost to follow-up as a reference. The 100% compliant patients had worse survival compared with the lost to follow-up group (P ¼ .033). The number at risk is shown below the plot; the asterisk denotes time points at which standard error exceeds 10% for each curve.
of being compliant with surveillance. Shorter driving distances (P ¼ .051) and shorter EVAR hospital stay (P ¼ .056) approached significant association with compliance. In addition, in our analysis of predictors of moderate compliance, presence of TIA or stroke in the patient’s medical history was independently predictive (OR, 3.59; P ¼ .024). In addition, our survival analysis as well as both the univariate and multivariate Cox regression analyses showed that 100% compliant patients had an increased risk of mortality compared with the lost to follow-up patients. Partly owing to few events, we found no survival difference between moderately compliant and lost to follow-up patients. Including our institutional results, all three studies examining predictors of post-EVAR surveillance have found that late complications are associated with compliance. Several explanations can be proposed for this finding. First, as raised by the previous authors, it may be that greater efforts are made to ensure follow-up for patients who develop late complications.8,9 Second, the knowledge that they have aneurysm-related complications may extend to the patient’s ownership of a more conscientious effort to follow medical (and surveillance) recommendations. Third, selection bias may be a confounding factor. Specifically, the longer patients are followed up with surveillance imaging, the more opportunities there are for complications to occur and to be detected. If all the lost to follow-up patients are subjected to CTA, some of them are likely to have asymptomatic late complications that were previously undetected. It would make sense, then, that secondary interventions were not
found to be predictive, given that most interventions are triggered by symptoms and that only 1.4% to 9% of secondary interventions are indicated by asymptomatic complications detected on imaging.11-15 In addition, on the basis of how we defined “compliant” and “moderately compliant,” patients who initially miss surveillance imaging but return with a late complication and then obtain surveillance imaging afterward would have been categorized as moderately compliant. This may contribute to the lost to follow-up group’s having fewer late complications. In contrast to the previous studies, we did identify several additional independent predictors of compliance. The fact that social work documentation was significantly more common in lost to follow-up patients may speak to an unstable home environment with poorer financial resources that leads to poorer medical compliance, although it does not appear that social work services helped ensure compliance for these patients. In addition, it is likely that patients with family members who have or have had AAAs are more knowledgeable about the illness and its consequences and thus are more likely to obtain followup. Shorter driving distances approached significance (P ¼ .051) and may also have been predictive of compliance with a larger cohort of patients. Finally, shorter hospital stay after EVAR also approached significant association with compliance (P ¼ .056). Because this is adjusted for medical comorbidities, social work issues, and ICU stay, perhaps longer hospital stays negatively influenced patients’ perception of the institution and this contributed to their being medically noncompliant.
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Wu et al 7
Fig 3. Kaplan-Meier survival curve representing moderately compliant patients with patients completely lost to followup as a reference. No differences in survival were seen (P ¼ .149). The number at risk is shown below the plot; the asterisk denotes time points at which standard error exceeds 10% for each curve.
In agreement with previous studies, we also found worse survival for compliant patients compared with the lost to follow-up group.8-10 However, unlike previous studies, we performed a multivariate Cox regression analysis to adjust for covariates known to affect survival, such as medical comorbidities, late complications and secondary interventions, and case complexity. The finding that compliance is independently associated with increased mortality is difficult to explain. Leurs et al8 proposed that medically complex patients were more likely to be compliant, and as a result, these patients hindered the compliant groups’ survival, whereas Kret et al9 raised the idea that compliance was an indicator of late complications. If these previous authors’ hypotheses were correct, however, the negative effect of compliance on survival should have been mitigated in multivariate analyses after adjusting for these potential confounders. The fact that ASA physical status, age, and poor socioeconomic status are associated with shorter survival duration validates our model. We conclude that either compliance with the current recommended surveillance regimen has a negative impact on survival duration or confounders are poorly adjusted for or not adjusted for. For instance, medical comorbidities in our Cox model were taken at baseline, whereas compliance was defined during surveillance follow-up; as a result, newly diagnosed or worsening medical comorbidities that occurred during follow-up were not accounted for and may have affected patients’ behavior and their survival. We also extracted medical comorbidities from the electronic medical record, which is relatively unstandardized
and not nuanced with regard to severity. Finally, perhaps there are discrepancies or gaps in the Social Security Death Index records (as one of the referenced studies also used it to assess survival).9 Unavoidable sequelae of EVAR, such as late complications and secondary interventions, were found to have no impact on survival in our multivariate Cox analysis. The fact that compliance can be predicted but that improved survival of compliant patients cannot be demonstrated warrants further study. This, combined with the fact that secondary interventions are typically triggered by symptoms rather than by asymptomatic imaging findings, suggests that perhaps surveillance protocols need to be revised.11-15 One example proposed by Tse et al15 is to risk stratify patients at multiple time points: immediately after the procedure on the basis of aneurysm size and unfavorable anatomy (ie, neck length <10 mm and angulation >60 degrees), before discharge by CTA findings that confer higher risk for endograft-related complications, and after 1 year. Patients deemed to be at low risk undergo a less burdensome first postoperative year surveillance protocol with subsequent annual abdominal ultrasound and radiographic examination, whereas those deemed at high risk receive annual CTA. Those patients at risk for noncompliance, who experience fewer late complications and are healthier, would in all likelihood be deemed low risk and offered the least burdensome surveillance regimen. This could potentially increase compliance rates without compromising surveillance quality. After surveillance protocols are individualized, compliance and its
JOURNAL OF VASCULAR SURGERY --- 2015
8 Wu et al
Table VI. Univariate and multivariate predictors of shorter survival duration Univariate analysis Variable Compliance 100% compliant Moderately compliant Lost to follow-up ASA physical status, per grade Age, per year Gender Male Female AAA diameter, per cm Ruptured AAA Dissecting AAA Inflammatory AAA Endoleak on completion angiogram Length of hospital stay, per day ICU stay Social work consultation Late complications Secondary interventions Active smoking Family history of AAA Hypertension Hyperlipidemia Coronary artery disease Congestive heart failure COPD Diabetes mellitus Peripheral vascular disease TIA or stroke Driving distance, per mile
Multivariate analysis
RR
95% CI
P value
RR
95% CI
P value
1.9 1.76 1 1.82 1.07
1.05-3.43 0.69-4.49 Reference 1.09-3.03 1.03-1.10
.034a .232 Reference .02a .0001a
2.67 1.24 1 3.2 1.11
1.18-6.06 0.34-4.47 Reference 1.57-6.53 1.05-1.17
.018a .741 Reference .001a <.0001a
1 1.6 0.95 0.89 0.8 0.99 1.51 0.98 0.8 0.99 1 1.74 1.007 1 1.37 1.004 0.94 1.16 0.78 1.02 1.2 1.001 1
Reference 0.87-2.96 0.74-1.22 0.27-2.90 0.10-5.95 0.98-1.01 0.77-2.93 0.94-1.03 0.37-1.72 0.99-1.003 0.99-1.006 0.78-3.89 0.99-1.01 0.99-1.007 0.68-2.74 0.99-1.01 0.55-1.61 0.54-2.46 0.44-1.38 0.47-2.18 0.58-2.48 0.99-1.008 0.998-1.001
Reference .13 .709 .857 .829 .796 .224 .563 .583 .386 .931 .17 .101 .997 .366 .14 .834 .697 .395 .954 .61 .661 .565
1 1.18 0.87 1.32 0.82 6.38 0.52 0.93 1.78 3.24 1.01 0.41 0.56 1.14 0.45 1.07 1.29 0.68 1.7 1.42 0.88 0.5 1.002
Reference 0.51-2.71 0.63-1.2 0.29-6.02 0.08-8.3 0.94-43.2 0.22-1.19 0.85-1.02 0.62-5.10 1.44-7.27 0.46-2.20 0.13-1.25 0.19-1.64 0.44-2.92 0.019-1.11 0.51-2.23 0.65-2.54 0.26-1.74 0.80-3.61 0.51-3.93 0.33-2.35 0.23-1.08 0.99-1.004
Reference .694 .426 .711 .873 .058 .124 .157 .277 .004a .976 .119 .292 .785 .085 .846 .458 .425 .165 .497 .809 .080 .281
AAA, Abdominal aortic aneurysm; ASA, American Society of Anesthesiologists; CI, confidence interval; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; RR, rate ratio; TIA, transient ischemic attack. In the multivariate Cox regression analysis, 100% compliance remains predictive of decreased survival duration after adjusting for potential confounders. a Significant.
effect on survival outcomes can then be re-examined, preferably prospectively in randomized controlled trials to avoid the limitations of retrospective studies examining survival outlined in the previous paragraph. Those patients at risk for noncompliance can then receive further resources to ensure compliance. Our study has further limitations beyond what was mentioned. Our study is retrospective and data collection was unmasked, making our study vulnerable to bias. Also, with almost all patients having PCPs, some may be obtaining follow-up with their PCPs without documentation in the medical record, in which case these patients would have been erroneously counted as lost to followup instead of being excluded. Our patient cohort was also relatively small, leading to possible type II statistical errors. In addition, there was some surgeon-specific variability in surveillance recommendations, and therefore small differences in quality of the surveillance regimen to which the patient was assigned may confound the survival analysis. Finally, we did not contact our patients or the families of patients to determine reasons for noncompliance or causes of mortality.
CONCLUSIONS More than half of our patients were noncompliant with recommended post-EVAR surveillance protocols, although this did not result in decreased survival. Late complications, absence of social work documentation, family history of AAA, certain comorbidities, and, potentially, shorter driving distances were associated with obtaining some form of follow-up after EVAR. In agreement with prior studies, compliance with surveillance was associated with decreased survival duration. Further studies are needed to determine the best surveillance protocol and modality for patients based on their risk level of aneurysm-related late complications and to determine optimal methods of ensuring compliance in those at risk for noncompliance with these protocols. Prospective studies are warranted to examine survival outcomes of compliance with these surveillance protocols. AUTHOR CONTRIBUTIONS Conception and design: JR, DC Analysis and interpretation: CW, JR, DC
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Data collection: CW, DC Writing the article: CW Critical revision of the article: CW, HC, KG, JE, JR, DC Final approval of the article: CW, HC, KG, JE, JR, DC Statistical analysis: CW Obtained funding: Not applicable Overall responsibility: DC REFERENCES 1. Chaikof EL, Brewster DC, Dalman RL, Makaroun MS, Illig KA, Sicard GA, et al. The care of patients with an abdominal aortic aneurysm: the Society for Vascular Surgery practice guidelines. J Vasc Surg 2009;50:S2-49. 2. Volodos NL, Karpovich IP, Troyan VI, Kalashnikova YuV, Shekhanin VE, Ternyuk NE, et al. Clinical experience of the use of selffixing synthetic prostheses for remote endoprosthetics of the thoracic and the abdominal aorta and iliac arteries through the femoral artery and as intraoperative endoprosthesis for aorta reconstruction. Vasa Suppl 1991;33:93-5. 3. Parodi JC, Palmaz JC, Barone HD. Transfemoral intraluminal graft implantation for abdominal aortic aneurysms. Ann Vasc Surg 1991;5: 491-9. 4. Chaikof EL, Blankensteijn JD, Harris PL, White GH, Zarins CK, Bernhard VM, et al. Reporting standards for endovascular aortic aneurysm repair. J Vasc Surg 2002;35:1048-60. 5. Hobo R, Buth J. Secondary interventions following endovascular abdominal aortic aneurysm repair using current endografts. J Vasc Surg 2006;43:896-902. 6. Mehta M, Paty PS, Roddy SP, Taggert JB, Sternbach Y, Kreienberg PB, et al. Treatment options for delayed AAA rupture following endovascular repair. J Vasc Surg 2011;53:14-20.
Wu et al 9
7. Jones WB, Taylor SM, Kalbaugh CA, Joels CS, Blackhurst DW, Langan EM 3rd, et al. Lost to follow-up: a potential under-appreciated limitation of endovascular aneurysm repair. J Vasc Surg 2007;46: 434-40. 8. Leurs LJ, Laheij RJ, Buth J. What determines and are the consequences of surveillance intensity after endovascular abdominal aortic aneurysm repair? Ann Vasc Surg 2005;19:868-75. 9. Kret MR, Azarbal AF, Mitchell EL, Liem TK, Landry GJ, Moneta GL. Compliance with long-term surveillance recommendations following endovascular repair or type B dissection. J Vasc Surg 2013;58:25-31. 10. Waduud MA, Choong WL, Ritchie M, Williams C, Yadavali R, Lim S, et al. Endovascular aneurysm repair: is imaging surveillance robust, and does it influence long-term mortality? Cardiovasc Intervent Radiol 2015;38:33-9. 11. Nordon IM, Karthikesalingam A, Hinchliffe RJ, Holt PJ, Loftus IM, Thompson MM. Secondary interventions following endovascular aneurysm repair (EVAR) and the enduring value of graft surveillance. Eur J Vasc Endovasc Surg 2010;39:547-54. 12. Black SA, Carrell TW, Bell RE, Waltham M, Reidy J, Taylor PR. Longterm surveillance with computed tomography after endovascular aneurysm repair may not be justified. Br J Surg 2009;96:1280-3. 13. Dias NV, Riva L, Ivancev K, Resch T, Sonesson B, Malina M. Is there a benefit of frequent CT follow-up after EVAR? Eur J Vasc Endovasc Surg 2009;37:425-30. 14. Karthikesalingam A, Holt PJ, Hinchliffe RJ, Nordon IM, Loftus IM, Thompson MM. Risk of reintervention after endovascular aortic aneurysm repair. Br J Surg 2010;97:657-63. 15. Tse DM, Tapping CR, Patel R, Morgan R, Bratby MJ, Anthony S, et al. Surveillance after endovascular abdominal aortic aneurysm repair. Cardiovasc Intervent Radiol 2014;37:875-88.
Submitted Nov 12, 2014; accepted Feb 9, 2015.
Predictors of compliance with surveillance after endovascular aneurysm repair and comparative survival outcomes Chris Y. Wu, BS, Huiting Chen, MD, Katherine A. Gallagher, MD, Jonathan L. Eliason, MD, John E. Rectenwald, MD, and Dawn M. Coleman, MD, Ann Arbor, Mich Objective: Although imaging surveillance is mandatory for all patients after endovascular aneurysm repair (EVAR), many patients are not compliant with follow-up. We sought to determine predictors of compliance with EVAR surveillance and to examine how compliance with current surveillance protocols correlates with survival. Methods: We analyzed 188 patients who underwent EVAR at our institution for infrarenal abdominal aortic aneurysms (AAAs) between 2001 and 2011. The primary end point was compliance with post-EVAR surveillance recommendations. Univariate analysis included patient demographics and socioeconomic information, AAA characteristics, EVAR hospital course variables, late complications and secondary interventions, length of follow-up, smoking status, family history of AAA, driving distances, primary care providers, and medical comorbidities. Mortality was determined by the Social Security Death Index. Multinomial logistic regressions were fit to identify independent predictors of compliance. Survival plots were generated with the Kaplan-Meier method and compared with the log-rank test. Univariate and multivariate Cox regression analysis was used to determine effect of compliance on survival after adjusting for confounders. Results: Of 188 patients, 89 (47.3%) were 100% compliant with follow-up visits and imaging, 21 (11.1%) were moderately compliant by missing appointments, and 78 (41.4%) were lost to follow-up completely. Overall median age was 74 years, and 81.9% of patients were male. Late complications occurred in 77 patients (40.9%), secondary interventions were performed in 32 patients (17%), and 5-year mortality was 21.2%. Mean follow-up interval was >40 months for 100% compliant and moderately compliant patients and <20 months for those lost to follow-up (P < .0001). In adjusted analysis, late complications (odds ratio [OR], 2.71; 95% confidence interval [CI], 1.32-5.55; P [ .007), absence of social work consultation (OR, 2.43; 95% CI, 1.12-5.27; P [ .024), and family history of AAA (OR, 2.67; 95% CI, 1.06-6.75; P [ .037) were associated with 100% compliance, whereas shorter driving distances (P [ .051) and shorter hospital stay (P [ .056) approached significance. Transient ischemic attack or stroke (OR, 3.59; 95% CI, 1.18-10.91; P [ .024) was the only variable independently associated with moderate compliance. Compared with patients lost to follow-up, 100% compliant patients had worse survival (log-rank test, P [ .033), whereas moderately compliant patients’ survival was not significantly different (log-rank test, P [ .149). In adjusted Cox regression analysis, 100% compliant patients had decreased survival duration (rate ratio, 2.67; 95% CI, 1.18-6.06; P [ .018) compared with those lost to follow-up. Conclusions: Follow-up surveillance is incomplete for more than half of patients who undergo EVAR at our institution, and patient compliance can be predicted by covariates mentioned before. Compliance with current surveillance regimens does not confer a survival benefit. Further research individualizing surveillance protocols based on risk level of late complications and noncompliance and prospective studies examining resulting survival benefits of compliance are warranted. (J Vasc Surg 2015;-:1-9.)
Endovascular aneurysm repair (EVAR) has progressively replaced open aneurysm repair for the treatment of infrarenal abdominal aortic aneurysms (AAAs) since its introduction two decades ago because of reduced operative morbidity, shorter intensive care unit (ICU) and hospital From the Department of Surgery, Section of Vascular Surgery, University of Michigan. Author conflict of interest: none. Presented as a poster at the Thirty-eighth Annual Meeting of the Midwestern Vascular Surgical Society, Coralville, Iowa, September 4-6, 2014. Reprint requests: Dawn M. Coleman, MD, University of Michigan Cardiovascular Center, 1500 E Medical Center Dr, Ann Arbor, MI 48109-5867 (e-mail: [email protected]). The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest. 0741-5214 Copyright Ó 2015 by the Society for Vascular Surgery. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jvs.2015.02.023
length of stay, and lower perioperative mortality rates.1-4 Despite its increasing popularity, EVAR is associated with a high rate of secondary interventions, which is necessary in up to 8.7% of patients at 12 months with a cumulative incidence of 14% at 4 years.5 Endoleak is the most frequent complication after EVAR, seen in up to 25% of patients during follow-up.1 Late rupture is an additional risk with an incidence of 1.5%.6 As such, the Society for Vascular Surgery practice guidelines recommend imaging surveillance after EVAR to detect these rupture-predisposing late complications.1 Research regarding the optimal modality and schedule of surveillance is ongoing, although computed tomography angiography (CTA) is typically used, which subjects the patient to chronic exposure to radiation and iodinated contrast agents. Compliance with post-EVAR imaging surveillance remains suboptimal. Compliance rates range in the literature, and risk factors for such remain unclear. Jones et al7 reported that nearly one third of patients had incomplete 1
2 Wu et al
JOURNAL OF VASCULAR SURGERY --- 2015
Fig 1. Flow diagram showing selection of patients. EVAR, Endovascular aneurysm repair; PCP, primary care provider.
follow-up and that these patients had shorter follow-up intervals (34.7 vs 18.8 months; P < .001) and an increase in major adverse events (6.1% vs 0.5% in compliant patients), concluding that patients deemed at high risk for noncompliance be advised against EVAR. Leurs et al8 identified smoking, hyperlipidemia, unfitness for open surgery or anesthesia, and presence of late complication as significantly associated with compliance patterns; in addition, after 84 months of follow-up, compliant patients had higher mortality rates, leading them to conclude that patients with more medical comorbidities received more intensive follow-up. Kret et al,9 during a median follow-up period of 28 6 10.5 months, found that only 44% of patients were compliant with follow-up, and in this cohort the presence of late complications was the only significant association. These authors also found that compliance was associated with decreased overall survival, a result further confirmed by Waduud et al.9,10 Building on these referenced studies, we attempted to identify clinically useful predictors of post-EVAR surveillance compliance as well as to comprehensively examine the effect of compliance on patient survival. METHODS Study cohort. The University of Michigan Institutional Review Board approved this study (No. 00076472), and data were obtained retrospectively through review of electronic medical records. Patient consent was not required by our institution’s Institutional Review Board, and all patient data were stored without identifiers. All consecutive patients who underwent EVAR between 2001 and 2011 at the University of Michigan performed by the Section of Vascular Surgery were included from a prospectively maintained database. Fig 1 illustrates exclusion criteria, which included patients who died within 1 year of endograft placement, those who had EVAR performed for non-AAAs, those who were Veterans Affairs patients referred to the University of
Michigan for EVAR because of case complexity or equipment malfunction, those who were dialysis dependent or transplant patients (who have care coordinators that may confound compliance rates), and those for whom local or primary care follow-up was arranged (whereas these patients may have obtained some form of surveillance imaging with their primary care providers [PCPs], there is no evidence that the images were reviewed by a vascular surgeon or that the PCPs were aware of surveillance guidelines). Outcomes and covariates. Institutional practice pattern secures 1-month follow-up after EVAR before hospital discharge. Surveillance imaging is performed at this visit, usually in the form of CTA. Although it is subject to individual surgeon variability, surveillance imaging is then performed at 12 months after EVAR and then yearly thereafter, with more frequent imaging performed for complicated cases or those with identified endoleaks. Other forms of surveillance, such as ultrasound or radiography, were also acceptable. When patients miss surveillance appointments, clinic nurses contact them by mail and by phone multiple times. In addition, there would be intermittent correspondence with patients’ PCPs, depending on the attending physician. Compliance with vascular surgeons’ recommendations was the main outcome of interest. All but one of our vascular surgeons followed the described surveillance protocol. The one surgeon who deviated would, in certain cases, obtain imaging surveillance between longer intervals. Patients were considered 100% compliant with follow-up surveillance if they followed all of their vascular surgeon’s recommendations regarding surveillance imaging. Patients were considered moderately compliant if they missed appointments or surveillance imaging (either one appointment or multiple ones) but continued to follow up thereafter. Patients were considered lost to follow-up if they had no further vascular surgical follow-up, surveillance imaging, or further documentation beyond their last
JOURNAL OF VASCULAR SURGERY Volume -, Number -
missed appointment (despite the recommendation of their vascular surgeon to continue surveillance). Patients who missed clinic visits but still obtained surveillance imaging and patients who obtained imaging remotely were considered compliant if their vascular surgeon documented review of the imaging. Covariates extracted from the medical record included: patient age at the time of operation; gender; preoperative AAA characteristics, such as diameter, ruptured, inflammatory, or dissecting; length of hospital and ICU stay; presence of endoleak on completion angiography; American Society of Anesthesiologists (ASA) physical status score; presence of social work consultation, which was used as a surrogate for an unstable socioeconomic environment; family history of AAA; smoking history; home zip code; having a PCP; ethnicity; and insurance status. Driving distance from the patient’s home to the University of Michigan was calculated using open-access online software. Patient mortality and, if applicable, date of death was determined using open-access online Social Security Death Index databases. Medical comorbidities were extracted from the electronic record. Follow-up intervals were calculated as months from time of operation to last recorded vascular surgery clinic visit or surveillance imaging. Late complications were defined as endograft-related complications that occurred after initial hospital discharge, including endoleak, endograft or limb occlusion or stenosis, endograft kink, migration, late rupture, and endograft infection. Secondary interventions, defined as any open or endovascular procedures performed to fix graft-related complications, were recorded. Statistical analysis. All statistical analyses were performed with commercially available software: IBM SPSS Smartreader version 20 (IBM Corporation, Armonk, NY). The primary outcome of analysis was compliance with the vascular surgeon’s recommendations regarding EVAR surveillance as previously described. Univariate analyses were used to determine which covariates were associated with compliance. Continuous variables were assessed by the independent t-test or analysis of variance as appropriate, and categorical data were assessed by the c2 test or Fisher exact test as appropriate. We then fit separate multinomial logistic regression models for the 100% compliant and moderately compliant groups using the patients lost to follow-up as the reference group. In the initial multinomial regression models, we included all covariates with P value < .2 from the univariate analysis to adjust for potential confounders. By backward elimination, we excluded covariates not significantly associated with the outcomes of interest (defined as P value > .2) from subsequent iterations of the model. Final results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Adjusted covariates with P values < .05 were considered significant. The Kaplan-Meier technique and the log-rank test were then used to examine patient survival in each compliance group. The Cox proportional hazards model was used to identify univariate and multivariate predictors of survival.
Wu et al 3
Unadjusted and adjusted rate ratios (RRs), 95% CIs, and P values were calculated. All covariates associated with compliance identified from the multinomial analysis as well as covariates known to affect survival, such as baseline medical comorbidities, various patients’ demographics, and case complexity, were included in the Cox model. RESULTS Cohort. In total, 188 patients met selection criteria for inclusion. There was a male predominance (n ¼ 154; 81.9%), with a median age of 74 years (interquartile range [IQR], 67-79 years) and mostly white ethnicity (n ¼ 182; 96.8%). Patient demographics and risk factors are described in Table I. Eighty-nine patients (47.3%) were 100% compliant with post-EVAR surveillance, 21 patients (11.1%) were moderately compliant, and 78 patients (41.4%) were lost to follow-up. One hundred eighty patients (95.7%) carried some form of insurance (private, public, or both), and 182 patients (96.8%) had a PCP. Patients traveled a median distance of 43 miles (IQR, 27-76 miles) for care. Late complications were detected in 77 patients (40.9%; Table II), and secondary interventions were performed in 32 patients (17%). Overall median follow-up length was 25 months (IQR, 9-45 months). Of 188 patients, there were 40 deaths within 5 years after surgery (21.2% 5-mortality rate). None of the fully compliant patients were enrolled in EVAR clinical trials. Four patients in the moderately compliant or lost to follow-up cohort were initially in endograft device clinical trials. Univariate analysis. Table I reveals the univariate analysis results comparing covariates of interest between the 100% compliant, moderately compliant, and lost to follow-up patients. Variables significantly associated with compliance include length of hospital stay (P ¼ .04), average length of follow-up (P < .0001), active smoking (P ¼ .045), family history of AAA (P ¼ .034), and TIA or stroke (P ¼ .007). Social work consultation (P ¼ .077), ASA physical status (P ¼ .832), late complications (P ¼ .101), secondary interventions (P ¼ .234), medical comorbidities except for TIA and stroke, uninsured status (P ¼ .848), PCP (P ¼ .309), and driving distance (P ¼ .078) were not significantly associated with compliance outcomes. Multinomial analysis. Table III reveals variables associated with 100% compliance using the patients lost to follow-up as a reference group. All covariates from univariate analysis (Table I) with P values < .2 were included in the original multinomial regression model. Inflammatory AAA, ICU stay, active smoking, hyperlipidemia, and transient ischemic attack (TIA) or stroke were excluded from subsequent iterations because of nonsignificance. Variables achieving significant correlation with 100% compliance include late complications (P ¼ .007), absence of social work consultation (P ¼ .024), and family history of AAA (P ¼ .037). Variables approaching significance include shorter driving distance (P ¼ .051) and shorter length of hospital stay (P ¼ .056). Table IV reveals variables correlating with moderate compliance using the patients lost to follow-up as a
JOURNAL OF VASCULAR SURGERY --- 2015
4 Wu et al
Table I. Univariate analysis for 188 patients who underwent endovascular aneurysm repair (EVAR) between 2001 and 2011 Variable Age, years Gender Male Female ASA physical status, grade AAA diameter, cm Ruptured AAA Dissecting AAA Inflammatory AAA Endoleak on completion angiography Length of hospital stay, days ICU stay, days Absence of social work consultation Average length of follow-up, months Late complications Secondary intervention Active smoking Family history of AAA Hypertension Hyperlipidemia Coronary artery disease Congestive heart disease COPD Diabetes mellitus Chronic kidney disease Peripheral vascular disease TIA or stroke Uninsured PCP Minority race Driving distance, miles
100% Compliant (n ¼ 89)
Moderately compliant (n ¼ 21)
Lost to follow-up (n ¼ 78)
73.69 6 1.74
72.19 6 3.65
73.05 6 1.88
73 (82) 16 (18) 3.23 6 0.11 5.81 6 0.24 6 (6.7) 2 (2.2) 3 (3.4) 24 (27) 2.55 6 0.55 6 (6.7) 68 (76.4) 40.86 6 5.73 43 (48.3) 19 (21.3) 19 (21.3) 23 (25.8) 75 (84.3) 65 (73) 43 (48.3) 12 (13.5) 26 (29.2) 13 (14.8) 13 (14.6) 16 (18) 15 (16.9) 3 (3.4) 88 (98.9) 2 (2.2) 52.79 6 9.20
17 (81) 4 (19) 3.28 6 0.23 5.73 6 0.46 1 (4.8) 0 (0) 2 (9.5) 4 (19) 3.52 6 0.97 4 (19) 15 (71.4) 42.08 6 7.34 9 (42.9) 4 (19) 2 (9.5) 2 (9.5) 16 (76.2) 17 (81) 11 (52.4) 3 (14.3) 7 (35) 5 (23.8) 6 (28.6) 4 (19) 10 (47.6) 1 (4.8) 20 (95.2) 1 (4.8) 80.27 6 37.61
64 (82.1) 14 (17.9) 3.20 6 0.11 5.97 6 0.29 5 (6.4) 0 (0) 1 (1.3) 23 (29.5) 5.41 6 2.36 12 (15.4) 47 (60.3) 18.59 6 5.20 25 (32.1) 9 (11.5) 26 (33.3) 9 (11.7) 66 (84.6) 47 (60.3) 43 (55.1) 11 (14.1) 25 (32.1) 13 (16.7) 18 (23.1) 14 (18.2) 15 (19.2) 4 (5.1) 74 (94.9) 3 (3.8) 95.74 6 39.58
P value .734 .993 .832 .606 .946 .325 .161a .633 .040a .12a .077a <.0001 .101a .234 .045a .034a .63 .09a .678 .991 .851 .606 .216 .993 .007a .848 .309 .766 .078a
AAA, Abdominal aortic aneurysm; ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; PCP, primary care provider; TIA, transient ischemic attack. Continuous variables are displayed as mean 6 95% confidence interval and categorical variables as number (%). a P value < .2 will be included in multinomial analysis (excluding mean follow-up intervals).
Table II. Incidence of each late complication occurring in 77 patients
Table III. Multinomial logistic regression analysis results showing independent predictors of 100% compliance with recommended surveillance protocols
No. (%) Type I endoleak Type II endoleak Type III endoleak Type IV endoleak Type V endoleak Graft limb occlusion or stenosis Graft kink Graft migration Late rupture Graft infection
12 47 3 1 4 11 2 2 2 1
(6.3) (25) (1.6) (0.5) (2.1) (5.9) (1.1) (1.1) (1.1) (0.5)
reference group. Late complications, length of hospital stay, ICU stay, absence of social work consultation, family history of AAA, and driving distance were excluded after the first iteration of the multinomial regression analysis because of nonsignificance. The only significant association was TIA or stroke (P ¼ .024). Insignificant variables
Variable
OR
95% CI
Late complications 2.71 1.322-5.559 Absence of social work consultation 2.43 1.127-5.270 Family history of AAA 2.67 1.062-6.753 Driving distance, miles 0.99 0.986-1 Length of hospital stay 0.9 0.814-1.003
P value .007a .024a .037a .051 .056
AAA, Abdominal aortic aneurysm; CI, confidence interval; OR, odds ratio. a Significant.
included hyperlipidemia (P ¼ .074), inflammatory AAA (P ¼ .077), and active smoking (P ¼ .110). Survival analysis. Table V reveals number of mortality events in each compliance group and number of patients censored. The 100% compliant patients experienced 29 events (deaths) during the study period, moderately compliant patients experienced 6 events, and lost to followup patients experienced 20 events.
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Wu et al 5
Table IV. Multinomial logistic regression analysis results showing independent predictors of moderate compliance Variable TIA or stroke Hyperlipidemia Inflammatory AAA Active smoking
OR
95% CI
P value
3.59 3.44 13.3 0.27
1.185-10.917 0.888-13.375 0.758-235.1 0.057-1.34
.024a .074 .077 .110
AAA, Abdominal aortic aneurysm; CI, confidence interval; OR, odds ratio; TIA, transient ischemic attack. a Significant.
Table V. Kaplan-Meier survival analysis showing number of patients and events (deaths) in each group Censored Groups 100% compliant Moderately compliant Lost to follow-up
Patients
Events
No.
%
89 21 78
29 6 20
60 15 58
67.4 71.4 74.4
Figs 2 and 3 demonstrate survival plots comparing 100% compliant and moderately compliant groups with the lost to follow-up group. The 100% compliant patients had significantly worse survival compared with the lost to follow-up patients (P ¼ .033) as determined by the log-rank test, whereas no survival differences were found between moderately compliant patients and lost to follow-up patients (P ¼ .149). Table VI reveals univariate and multivariate predictors of shorter survival duration. In the univariate analysis, 100% compliance (RR, 1.9; 95% CI, 1.05-3.43; P value, .034), ASA physical status (RR, 1.82; 95% CI, 1.09-3.03; P value, .02), and age (RR, 1.07; 95% CI, 1.03-1.1; P value, .0001) were significantly associated with shorter survival duration. In adjusted analysis, 100% compliance (RR, 2.67; 95% CI, 1.18-6.06; P value, .018), ASA physical status (RR, 3.2; 95% CI, 1.57-6.53; P value, .001), age (RR, 1.11; 95% CI, 1.05-1.17; P value, <.0001), and social work consultation (RR, 3.24; 95% CI, 1.44-7.27; P value, .004) were independently associated with shorter survival duration. DISCUSSION The objective of this study was to identify clinically useful predictors of post-EVAR surveillance compliance as well as to comprehensively examine how compliance correlates with survival. We found that 47.3% of patients were 100% compliant, 11.1% of patients were moderately compliant, and 41.4% of patients were lost to follow-up. This is concordant with prior studies, which identified that only 35% to 67.2% of patients were deemed to have complete follow-up after EVAR.7-10 In addition, we found that late complications, stable socioeconomic situations, family history of AAA, and TIA or stroke were independently associated with either full or moderate compliance.
Finally, in our survival analysis, we found that 100% compliant patients had decreased survival duration, a result that was consistent after adjusting for confounders in a multivariate Cox regression analysis. Jones et al published on 302 patients and found that mean follow-up length was increased in the compliant group; however, 5-year intervention rates, 5-year survival, and incidence of known endoleak were all found to be similar between groups. Major events that required urgent surgery were increased in the noncompliant group (6.1% vs 0.5%; P ¼ .006). They also found that follow-up duration was better in patients enrolled in clinical trials, which they attributed to the persistence of research nurses. They concluded that patients at risk for compliance should be advised against EVAR. Our results affirm several of these conclusions as we also found that follow-up duration was better in patients who had follow-up (>40 months mean follow-up interval for 100% compliant and moderately compliant patients vs <20 months mean follow-up interval for lost to follow-up patients; P < .0001). We were unable to assess cause of mortality in the lost to follow-up group. Two studies have sought to determine predictors of compliance with post-EVAR surveillance. In their study of 4433 patients enrolled in the European Collaborating Group on Stent-Graft Techniques for Abdominal Aortic Aneurysm Repair (EUROSTAR) who underwent EVAR between 1996 and 2004, Leurs et al8 identified certain comorbidities (active smoking, hyperlipidemia, and unfitness for open surgery or general anesthesia) as predictive of compliance. In agreement with our study, complications during follow-up were significantly more common in the compliant group. Cumulative survival rates were found to be worse in the compliant group (71% vs 74%; P < .0001). These results were confirmed in 2013 by Kret et al,9 who reported that during a median follow-up of 28 6 10.5 months, only late complications were associated with better compliance. Medical comorbidities and socioeconomic factors including distance from the patient’s home zip code to the medical center’s zip code were not significantly associated with compliance. They also found that noncompliant patients had improved survival outcomes (P < .001). These counterintuitive results positively associating compliance with increased mortality were confirmed by Waduud et al,10 who found that patients with incomplete follow-up had improved survival duration compared with the complete follow-up group (P < .001) in Cox regression analysis; however, their multivariate analysis adjusted only for age, gender, location, AAA diameter, and year of EVAR, leaving out variables known to affect mortality, such as medical comorbidities and case complexity. Similar to these studies examining post-EVAR surveillance compliance, we found that late complications (OR, 2.71; P ¼ .007) were independently predictive of 100% compliance. In addition, we found that a stable socioeconomic situation (using absence of social work consultation as a surrogate; OR, 2.43; P ¼ .024) and family history of AAA (OR, 2.67; P ¼ .037) were independently predictive
6 Wu et al
JOURNAL OF VASCULAR SURGERY --- 2015
Fig 2. Kaplan-Meier survival curve representing 100% compliant patients with patients completely lost to follow-up as a reference. The 100% compliant patients had worse survival compared with the lost to follow-up group (P ¼ .033). The number at risk is shown below the plot; the asterisk denotes time points at which standard error exceeds 10% for each curve.
of being compliant with surveillance. Shorter driving distances (P ¼ .051) and shorter EVAR hospital stay (P ¼ .056) approached significant association with compliance. In addition, in our analysis of predictors of moderate compliance, presence of TIA or stroke in the patient’s medical history was independently predictive (OR, 3.59; P ¼ .024). In addition, our survival analysis as well as both the univariate and multivariate Cox regression analyses showed that 100% compliant patients had an increased risk of mortality compared with the lost to follow-up patients. Partly owing to few events, we found no survival difference between moderately compliant and lost to follow-up patients. Including our institutional results, all three studies examining predictors of post-EVAR surveillance have found that late complications are associated with compliance. Several explanations can be proposed for this finding. First, as raised by the previous authors, it may be that greater efforts are made to ensure follow-up for patients who develop late complications.8,9 Second, the knowledge that they have aneurysm-related complications may extend to the patient’s ownership of a more conscientious effort to follow medical (and surveillance) recommendations. Third, selection bias may be a confounding factor. Specifically, the longer patients are followed up with surveillance imaging, the more opportunities there are for complications to occur and to be detected. If all the lost to follow-up patients are subjected to CTA, some of them are likely to have asymptomatic late complications that were previously undetected. It would make sense, then, that secondary interventions were not
found to be predictive, given that most interventions are triggered by symptoms and that only 1.4% to 9% of secondary interventions are indicated by asymptomatic complications detected on imaging.11-15 In addition, on the basis of how we defined “compliant” and “moderately compliant,” patients who initially miss surveillance imaging but return with a late complication and then obtain surveillance imaging afterward would have been categorized as moderately compliant. This may contribute to the lost to follow-up group’s having fewer late complications. In contrast to the previous studies, we did identify several additional independent predictors of compliance. The fact that social work documentation was significantly more common in lost to follow-up patients may speak to an unstable home environment with poorer financial resources that leads to poorer medical compliance, although it does not appear that social work services helped ensure compliance for these patients. In addition, it is likely that patients with family members who have or have had AAAs are more knowledgeable about the illness and its consequences and thus are more likely to obtain followup. Shorter driving distances approached significance (P ¼ .051) and may also have been predictive of compliance with a larger cohort of patients. Finally, shorter hospital stay after EVAR also approached significant association with compliance (P ¼ .056). Because this is adjusted for medical comorbidities, social work issues, and ICU stay, perhaps longer hospital stays negatively influenced patients’ perception of the institution and this contributed to their being medically noncompliant.
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Wu et al 7
Fig 3. Kaplan-Meier survival curve representing moderately compliant patients with patients completely lost to followup as a reference. No differences in survival were seen (P ¼ .149). The number at risk is shown below the plot; the asterisk denotes time points at which standard error exceeds 10% for each curve.
In agreement with previous studies, we also found worse survival for compliant patients compared with the lost to follow-up group.8-10 However, unlike previous studies, we performed a multivariate Cox regression analysis to adjust for covariates known to affect survival, such as medical comorbidities, late complications and secondary interventions, and case complexity. The finding that compliance is independently associated with increased mortality is difficult to explain. Leurs et al8 proposed that medically complex patients were more likely to be compliant, and as a result, these patients hindered the compliant groups’ survival, whereas Kret et al9 raised the idea that compliance was an indicator of late complications. If these previous authors’ hypotheses were correct, however, the negative effect of compliance on survival should have been mitigated in multivariate analyses after adjusting for these potential confounders. The fact that ASA physical status, age, and poor socioeconomic status are associated with shorter survival duration validates our model. We conclude that either compliance with the current recommended surveillance regimen has a negative impact on survival duration or confounders are poorly adjusted for or not adjusted for. For instance, medical comorbidities in our Cox model were taken at baseline, whereas compliance was defined during surveillance follow-up; as a result, newly diagnosed or worsening medical comorbidities that occurred during follow-up were not accounted for and may have affected patients’ behavior and their survival. We also extracted medical comorbidities from the electronic medical record, which is relatively unstandardized
and not nuanced with regard to severity. Finally, perhaps there are discrepancies or gaps in the Social Security Death Index records (as one of the referenced studies also used it to assess survival).9 Unavoidable sequelae of EVAR, such as late complications and secondary interventions, were found to have no impact on survival in our multivariate Cox analysis. The fact that compliance can be predicted but that improved survival of compliant patients cannot be demonstrated warrants further study. This, combined with the fact that secondary interventions are typically triggered by symptoms rather than by asymptomatic imaging findings, suggests that perhaps surveillance protocols need to be revised.11-15 One example proposed by Tse et al15 is to risk stratify patients at multiple time points: immediately after the procedure on the basis of aneurysm size and unfavorable anatomy (ie, neck length <10 mm and angulation >60 degrees), before discharge by CTA findings that confer higher risk for endograft-related complications, and after 1 year. Patients deemed to be at low risk undergo a less burdensome first postoperative year surveillance protocol with subsequent annual abdominal ultrasound and radiographic examination, whereas those deemed at high risk receive annual CTA. Those patients at risk for noncompliance, who experience fewer late complications and are healthier, would in all likelihood be deemed low risk and offered the least burdensome surveillance regimen. This could potentially increase compliance rates without compromising surveillance quality. After surveillance protocols are individualized, compliance and its
JOURNAL OF VASCULAR SURGERY --- 2015
8 Wu et al
Table VI. Univariate and multivariate predictors of shorter survival duration Univariate analysis Variable Compliance 100% compliant Moderately compliant Lost to follow-up ASA physical status, per grade Age, per year Gender Male Female AAA diameter, per cm Ruptured AAA Dissecting AAA Inflammatory AAA Endoleak on completion angiogram Length of hospital stay, per day ICU stay Social work consultation Late complications Secondary interventions Active smoking Family history of AAA Hypertension Hyperlipidemia Coronary artery disease Congestive heart failure COPD Diabetes mellitus Peripheral vascular disease TIA or stroke Driving distance, per mile
Multivariate analysis
RR
95% CI
P value
RR
95% CI
P value
1.9 1.76 1 1.82 1.07
1.05-3.43 0.69-4.49 Reference 1.09-3.03 1.03-1.10
.034a .232 Reference .02a .0001a
2.67 1.24 1 3.2 1.11
1.18-6.06 0.34-4.47 Reference 1.57-6.53 1.05-1.17
.018a .741 Reference .001a <.0001a
1 1.6 0.95 0.89 0.8 0.99 1.51 0.98 0.8 0.99 1 1.74 1.007 1 1.37 1.004 0.94 1.16 0.78 1.02 1.2 1.001 1
Reference 0.87-2.96 0.74-1.22 0.27-2.90 0.10-5.95 0.98-1.01 0.77-2.93 0.94-1.03 0.37-1.72 0.99-1.003 0.99-1.006 0.78-3.89 0.99-1.01 0.99-1.007 0.68-2.74 0.99-1.01 0.55-1.61 0.54-2.46 0.44-1.38 0.47-2.18 0.58-2.48 0.99-1.008 0.998-1.001
Reference .13 .709 .857 .829 .796 .224 .563 .583 .386 .931 .17 .101 .997 .366 .14 .834 .697 .395 .954 .61 .661 .565
1 1.18 0.87 1.32 0.82 6.38 0.52 0.93 1.78 3.24 1.01 0.41 0.56 1.14 0.45 1.07 1.29 0.68 1.7 1.42 0.88 0.5 1.002
Reference 0.51-2.71 0.63-1.2 0.29-6.02 0.08-8.3 0.94-43.2 0.22-1.19 0.85-1.02 0.62-5.10 1.44-7.27 0.46-2.20 0.13-1.25 0.19-1.64 0.44-2.92 0.019-1.11 0.51-2.23 0.65-2.54 0.26-1.74 0.80-3.61 0.51-3.93 0.33-2.35 0.23-1.08 0.99-1.004
Reference .694 .426 .711 .873 .058 .124 .157 .277 .004a .976 .119 .292 .785 .085 .846 .458 .425 .165 .497 .809 .080 .281
AAA, Abdominal aortic aneurysm; ASA, American Society of Anesthesiologists; CI, confidence interval; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; RR, rate ratio; TIA, transient ischemic attack. In the multivariate Cox regression analysis, 100% compliance remains predictive of decreased survival duration after adjusting for potential confounders. a Significant.
effect on survival outcomes can then be re-examined, preferably prospectively in randomized controlled trials to avoid the limitations of retrospective studies examining survival outlined in the previous paragraph. Those patients at risk for noncompliance can then receive further resources to ensure compliance. Our study has further limitations beyond what was mentioned. Our study is retrospective and data collection was unmasked, making our study vulnerable to bias. Also, with almost all patients having PCPs, some may be obtaining follow-up with their PCPs without documentation in the medical record, in which case these patients would have been erroneously counted as lost to followup instead of being excluded. Our patient cohort was also relatively small, leading to possible type II statistical errors. In addition, there was some surgeon-specific variability in surveillance recommendations, and therefore small differences in quality of the surveillance regimen to which the patient was assigned may confound the survival analysis. Finally, we did not contact our patients or the families of patients to determine reasons for noncompliance or causes of mortality.
CONCLUSIONS More than half of our patients were noncompliant with recommended post-EVAR surveillance protocols, although this did not result in decreased survival. Late complications, absence of social work documentation, family history of AAA, certain comorbidities, and, potentially, shorter driving distances were associated with obtaining some form of follow-up after EVAR. In agreement with prior studies, compliance with surveillance was associated with decreased survival duration. Further studies are needed to determine the best surveillance protocol and modality for patients based on their risk level of aneurysm-related late complications and to determine optimal methods of ensuring compliance in those at risk for noncompliance with these protocols. Prospective studies are warranted to examine survival outcomes of compliance with these surveillance protocols. AUTHOR CONTRIBUTIONS Conception and design: JR, DC Analysis and interpretation: CW, JR, DC
JOURNAL OF VASCULAR SURGERY Volume -, Number -
Data collection: CW, DC Writing the article: CW Critical revision of the article: CW, HC, KG, JE, JR, DC Final approval of the article: CW, HC, KG, JE, JR, DC Statistical analysis: CW Obtained funding: Not applicable Overall responsibility: DC REFERENCES 1. Chaikof EL, Brewster DC, Dalman RL, Makaroun MS, Illig KA, Sicard GA, et al. The care of patients with an abdominal aortic aneurysm: the Society for Vascular Surgery practice guidelines. J Vasc Surg 2009;50:S2-49. 2. Volodos NL, Karpovich IP, Troyan VI, Kalashnikova YuV, Shekhanin VE, Ternyuk NE, et al. Clinical experience of the use of selffixing synthetic prostheses for remote endoprosthetics of the thoracic and the abdominal aorta and iliac arteries through the femoral artery and as intraoperative endoprosthesis for aorta reconstruction. Vasa Suppl 1991;33:93-5. 3. Parodi JC, Palmaz JC, Barone HD. Transfemoral intraluminal graft implantation for abdominal aortic aneurysms. Ann Vasc Surg 1991;5: 491-9. 4. Chaikof EL, Blankensteijn JD, Harris PL, White GH, Zarins CK, Bernhard VM, et al. Reporting standards for endovascular aortic aneurysm repair. J Vasc Surg 2002;35:1048-60. 5. Hobo R, Buth J. Secondary interventions following endovascular abdominal aortic aneurysm repair using current endografts. J Vasc Surg 2006;43:896-902. 6. Mehta M, Paty PS, Roddy SP, Taggert JB, Sternbach Y, Kreienberg PB, et al. Treatment options for delayed AAA rupture following endovascular repair. J Vasc Surg 2011;53:14-20.
Wu et al 9
7. Jones WB, Taylor SM, Kalbaugh CA, Joels CS, Blackhurst DW, Langan EM 3rd, et al. Lost to follow-up: a potential under-appreciated limitation of endovascular aneurysm repair. J Vasc Surg 2007;46: 434-40. 8. Leurs LJ, Laheij RJ, Buth J. What determines and are the consequences of surveillance intensity after endovascular abdominal aortic aneurysm repair? Ann Vasc Surg 2005;19:868-75. 9. Kret MR, Azarbal AF, Mitchell EL, Liem TK, Landry GJ, Moneta GL. Compliance with long-term surveillance recommendations following endovascular repair or type B dissection. J Vasc Surg 2013;58:25-31. 10. Waduud MA, Choong WL, Ritchie M, Williams C, Yadavali R, Lim S, et al. Endovascular aneurysm repair: is imaging surveillance robust, and does it influence long-term mortality? Cardiovasc Intervent Radiol 2015;38:33-9. 11. Nordon IM, Karthikesalingam A, Hinchliffe RJ, Holt PJ, Loftus IM, Thompson MM. Secondary interventions following endovascular aneurysm repair (EVAR) and the enduring value of graft surveillance. Eur J Vasc Endovasc Surg 2010;39:547-54. 12. Black SA, Carrell TW, Bell RE, Waltham M, Reidy J, Taylor PR. Longterm surveillance with computed tomography after endovascular aneurysm repair may not be justified. Br J Surg 2009;96:1280-3. 13. Dias NV, Riva L, Ivancev K, Resch T, Sonesson B, Malina M. Is there a benefit of frequent CT follow-up after EVAR? Eur J Vasc Endovasc Surg 2009;37:425-30. 14. Karthikesalingam A, Holt PJ, Hinchliffe RJ, Nordon IM, Loftus IM, Thompson MM. Risk of reintervention after endovascular aortic aneurysm repair. Br J Surg 2010;97:657-63. 15. Tse DM, Tapping CR, Patel R, Morgan R, Bratby MJ, Anthony S, et al. Surveillance after endovascular abdominal aortic aneurysm repair. Cardiovasc Intervent Radiol 2014;37:875-88.
Submitted Nov 12, 2014; accepted Feb 9, 2015.