Increasing clinicians’ suspicion of ATTR amyloidosis using a retrospective algorithm

Abstract

Background

This study aimed to increase the index of suspicion for transthyretin amyloidosis (ATTR) among cardiologists leading to increased screening for amyloidosis.

Methods

A retrospective algorithm was created to identify patients at risk for ATTR. A list of these patients and instructions on how to order amyloidosis testing were given to cardiologists, who then determined if further evaluation was warranted. The ordering trends of Technetium 99 m-Pyrophosphate (PYP) scans and the number of ordering physicians before and after this intervention were recorded across the entire practice.

Results

The algorithm identified 349 potential high-risk patients of which only 23 eventually had PYP scans performed resulting in 2 equivocal and 1 positive results. Across the practice, over the 28 months before initiating this protocol, PYP scans were ordered for 22 patients of which 6 were equivocal or positive. Over the 23-month course of this project, 142 PYP scans were ordered of which 18 were equivocal or positive. The number of ordering providers increased from 7 prior to the protocol’s implementation to 22 by the end of this project within 23 months. On change point analysis, PYP scan ordering increased after protocol initiation (regression coefficient 1.27 vs. 6.31, p < 0.001), as well as equivocal or positive PYP results (regression coefficient 0.38 vs. 0.52, p < 0.01).

Conclusion

The results of this study suggest that using this algorithm, despite it not being independently predictive of ATTR, did result in our clinicians having a lower threshold for testing for ATTR. More clinicians ordered appropriate testing, and more positive tests were obtained.

Background

Amyloidosis is a disease characterized by extracellular tissue deposition of insoluble fibrils composed of abnormally folded proteins. Two types of amyloidosis are responsible for 95% of cardiac amyloidosis cases, immunoglobulin light chain amyloidosis (AL) and transthyretin amyloidosis (ATTR). ATTR has two main types: hereditary transthyretin amyloidosis (hATTR) and wild-type transthyretin amyloidosis (wtATTR). The Val122Ile is the most common hATTR variant in the US and this mutation is present in approximately 4% of African Americans. This may even be an underestimation due to healthcare disparities and underdiagnosis of cardiac amyloidosis (CA), as these patients are underrepresented in the Southern U.S. despite having larger proportions of self-identified African Americans [1,2,3]. The exact penetrance is unknown, though some data evaluations estimate 44% of Val122Ile carriers > 50 years old have heart failure (HF) or cardiomyopathy, and the incidence increases to 70% and 100% for 70- and 80-year-olds, respectively, though there may be competing causes of heart failure (such as wtATTR) in some of these instances [4].

ATTR has historically been considered a rare disease, though it is underdiagnosed due to its vague, multisystem involvement and common HF symptoms that lead clinicians to a nonspecific diagnosis of HFpEF. The phenotype of hATTR varies depending on the specific mutation, but extracardiac manifestations may include autonomic and peripheral neuropathy, carpal tunnel syndrome (CTS), spinal stenosis, and more. Transthyretin is a tetrameric protein made by the liver, choroid plexus, and retinal epithelium. It is involved in transporting thyroid hormones and retinol. In ATTR amyloidosis, this tetramer becomes destabilized, loses its quaternary structure, and can circulate and deposit systemically, including within the interstitium of the heart, leading to disease. The diffuse deposition of amyloid fibrils in the cardiac interstitial space leads to cardiomyocyte necrosis and interstitial fibrosis, subsequently, there is biventricular wall thickening and stiffening resulting in restrictive cardiomyopathy with diastolic dysfunction. It’s estimated 10% of African Americans > 60 years old with HF are Val122Ile positive [4].

To make an accurate and timely diagnosis, clinicians must have a high index of suspicion to execute a rapid, targeted diagnostic evaluation which would allow for sooner initiation of treatment that can improve prognosis, all-cause mortality, and decrease hospitalization rates. Previous studies using electronic medical record (EMR)-based data extraction to improve screening have shown promise to offer effective, low-cost screening in high-risk patients, limiting low-value testing and unnecessary testing [5, 6]. Our hybrid academic/private cardiology practice consists of 29 cardiologists in the Greater Memphis, TN area serving roughly 60,000 outpatients per year of which 31,500 were African American. Our study utilized electronic medical records to identify examples of patients at risk for amyloidosis and notify their respective providers who would then determine further need for testing. The primary outcome was the number of patients diagnosed with ATTR amyloidosis. Secondary outcomes observed include the following: 

• Number of patients receiving amyloidosis workup before and after being identified by the algorithm and the provider being informed.

• Total number of PYP orders and number of ordering physicians before and after this intervention.

Methods

Study design

After obtaining IRB approval from the University of Tennessee Health Science Center (UTHSC) Institutional Review Board. Our intervention’s targets were the individual providers, and the IRB deemed the protocol a retrospective quality improvement study not requiring patient consent. The study was conducted in accordance with the Declaration of Helsinki. We retrospectively identified patients at risk for amyloidosis via chart review of echocardiographic data and ICD-10 codes from the Methodist-Le Bonheur Healthcare system consisting of 6 hospitals located in Tennessee and Mississippi, including a secondary and tertiary care academic hospital. The total clinical volume over the preceding 2 years prior to the study consisted of over 120,000 outpatients, of which over 63,000 were African American.

First, patients from the previous three-year period were identified from the electronic echocardiography interface system (Change Healthcare Cardiology version 14.2 P7; Change Healthcare Canada) meeting the following criteria: LV posterior wall (LVPW) and interventricular septal thickness (IVS) of ≥ 1.4 cm, age > 65 years old, and left ventricular ejection fraction (LVEF) ≥ 50%. These patients were then matched to our outpatient clinic electronic medical record system (Touchworks/Allscripts version 19.4; Veridigm Corporation) if they had a clinic note within the past 3 years with an ICD-10 code for either nonspecific or diastolic heart failure. Patients with known amyloidosis were excluded. Patients that did not have a clinic note and echocardiogram in our system over the corresponding timeframe were excluded. This process was repeated every 5 months to identify new patients meeting the screening criteria. Only outpatients actively followed in our cardiology clinics were included. A list of patients fulfilling the criteria was emailed to the outpatient provider after each data pull. Included with the list was a standardized explanation of our approach and concern for ATTR amyloidosis, and a brief journal article outlining a simplified approach to the diagnosis of ATTR amyloidosis, and specific instructions on how to order a Technetium 99 m-Pyrophosphate (PYP) imaging and appropriate serum/urine testing in our EMR system for the necessary step of ruling out AL amyloidosis. Instructions provided included information regarding subspecialist follow-up and referral for treatment for any positive results; however, the ultimate decision to screen or refer was left to the discretion of the individual providers.

Data collected included cardiac PYP scan testing frequency and results for the entire inpatient and outpatient practice starting 28 months before the protocol through 23 months after its initiation. Patients found to have AL amyloidosis through this protocol were excluded from analysis (n = 2). PYP scan results were individually adjudicated by a level 3 nuclear radiologist.

Statistical analysis

The data from multiple sites was reviewed and quality checked for preprocessing in terms of coding categorical variables, converting continuous variables into categorical variables. The data from multiple spreadsheets were organized, appended, and merged as needed to analyze the trends in PYP scans and positive rates. Baseline univariate analysis was carried out using Chi-Square for categorical variables and from ANOVA for continues variables.

A change point analysis was performed to identify a unique time point in time where there was a significant change in the total number of ordered PYP scans and positive results. Linear regression fit using data from December 18 to March 15th, 2021, as well as after March 15th, 2021, was performed. A t-test comparing the two slopes before and after the change point was performed.

Results

A total of 349 patients who were deemed high-risk for ATTR amyloidosis were identified and this information was presented to the patient’s outpatient cardiologist. Of these, 23 eventually underwent amyloidosis testing of which there were 2 equivocal and 1 positive PYP scans.

In the 28 months prior to the start of the protocol, there were a total of 22 PYP orders across our healthcare system corresponding to an average of 0.8 PYP orders per month (Fig. 1). Of the 22 orders, 16 were negative (72.7%), 3 tests were equivocal (13.6%), and 3 tests were positive (13.6% each). Over the next 23 months after implementing the new protocol, there were a total of 142 PYP orders out of the 349 patients identified as high-risk: corresponding to an average of 5.7 PYP (120 more total orders corresponding to a 713% increase per month) orders per month (Table 1). There was a total of 124 negative tests (87.3%), 5 tests were equivocal (3.5%), and 13 tests were positive (9.2%), for a total of 18 equivocal or positive tests (12.7%). There were no statistically significant characteristics found between patients who tested negative vs. equivocal or positive (Table 2).

Fig. 1

PYP Orders Over time. PYP orders (blue line) and ordering providers (red line) over time before and after protocol initiation (grey dashed line). Both the total number of PYP orders and number of ordering providers substantially increased following the protocol initiation date

Table 1 PYP orders and results summary

Table 2 Lab values comparison based on PYP results

Change point analysis of the total PYP orders showed a significant difference in the number of PYP orders on March 15th, 2021 (Fig. 2). Similar change point analysis on the total number of equivocal or positive PYP results detected the date of September 15th, 2021, as the change point where equivocal or positive result patterns were significantly changing (Fig. 3). The linear regression fit using data from December 18 to September 15th, 2021, resulted in slope with 95% confidence interval of (which would be the regression coefficient) 0.38 (0.34–0.42). On the other hand, using data after September 15th, 2021, a slope of 0.52 (0.44–0.62). A t-test comparing two slopes, before and after change point, resulted in statistically significant difference (p < 0.01) confirming that the amount monthly increase in the amount equivocal or positive PYP are significantly higher after September 15th, 2021. This ultimately showed that while there was an increase in the orders starting on March 15th, 2021, and implementation of the protocol several weeks later on April 4th, 2021; however, it took approximately 5 months until results were seen.

Fig. 2

PYP Change Point Analysis. Change point analysis of PYP orders over time. The cumulative PYP orders (blue line) significantly increased after March 15th, 2021 (green line). Regression coefficient (orange line) before change point was 1.27 (95% CI 1.25–1.40) vs. after 6.31 (95% CI 6.11–6.52) (p < 0.001). Knowing that the new protocol was initiated on April 4th, 2021, the results suggest that the new protocol resulted in increased PYP orders

Fig. 3

Change Point Timing. A change point analysis of equivocal or positive PYP scans (blue line) over time. The change point (green line) occurred September 15th, 2021 (green line), roughly 5 months after the protocol was initiated. Regression coefficients (orange line) before change point was 0.38 (0.34–0.42) and following was 0.52 (0.44–0.62) (p < 0.01)

Discussion

Our study used a simple algorithm to identify high-risk patients for cardiac amyloidosis and relied on providers to decide whether further testing was warranted. Following this protocol’s initiation, both testing and newly diagnosed ATTR amyloidosis cases increased dramatically. It is important to note that it was not the purpose of this study to produce a successful TTR-detection algorithm. The purpose of this study was to increase our clinicians’ suspicions for testing for amyloidosis, regardless of whether or not we produced a successful algorithm. Perhaps the most intriguing finding is that providers did not test many individuals identified in the high-risk algorithm, yet their testing frequency still increased significantly in their general practice. There are several reasons for this. First, the protocol to identify high-risk patients was perhaps too rudimentary to be very helpful, particularly in a population with a high prevalence of confounding conditions. Our protocol did not exclude patients with ventricular hypertrophy from other causes, such as hypertension, hypertrophic cardiomyopathy, or end-stage renal disease. Second, many patients, especially on initial data capture, may not have been actively followed by our clinic due to the up to 3 years lag in time from the last clinic contact to the point of identification. Several patients in this cohort may have died, changed providers to a different practice, or may not have been interested in further testing. Thirdly, some of the providers may have ignored our notification for further testing due to persistent alarm fatigue which is becoming increasingly prevalent in electronic medical records. Nevertheless, amyloidosis is currently a niche area of cardiology and is still unfamiliar to many cardiology providers who may focus primarily on coronary disease or arrhythmias.

A previous iteration of this study attempted to identify patients at high-risk for amyloidosis by including additional ICD-10 codes for comorbidities seen in amyloidosis (i.e., carpal tunnel syndrome, spinal stenosis, joint replacement, glaucoma, orthostasis). However, it was thought to be too insensitive as these comorbidities were not routinely collected in our cardiology clinic. As more complex screening tools develop, it would be useful to inquire about these comorbidities in cardiology clinics with subsequent incorporation of the data into the algorithm. One readily available feature that may hold significant predictive value is body mass index, as obesity is a relatively rare finding in ATTR-CM (transthyretin amyloid cardiomyopathy), while it is common in other causes of heart failure with preserved ejection fraction [7]. Standardizing and incorporating advanced echocardiographic techniques, like myocardial strain imaging in patients with significant hypertrophy, might improve the specificity of the algorithm but may be limited by additional cost, technical skills, and time constraints in clinical practice [8].

Our study showed that periodically informing providers of their high-risk patients for cardiac amyloidosis resulted in providers screening their new patients for amyloidosis, as well as a marked increase in the overall rate of screening in the selected patient groups. The primary outcome assessed was the total number of patients diagnosed with ATTR-CM, which increased after the initiation of the protocol. However, the equivocal or positive rate decreased after protocol initiation. This drop-off in specificity may be partially explained by providers being unfamiliar with screening for ATTR amyloidosis and, therefore, not having yet developed a good clinical pre-test probability for this disease. Indeed, a number of new PYP scans after protocol initiation were ordered on patients that were < 65 years old. Additionally, it could be attributed to screening a group with a lower prevalence of ATTR-CM. The equivocal or positive rate may have been higher if we combined the initial inclusion criteria with some of the “red flag” signs/symptoms proposed by experts in amyloidosis [9]. Unfortunately, our study did not identify any statistically significant characteristics with discriminatory power to predict if a patient would have a positive PYP scan. This may be due to a small sample size or a true lack of statistical significance. Additionally, it is possible that the echocardiographic criteria were insensitive in a population with high rates of comorbid and confounding conditions. Our study demonstrates that pointing out high-risk patients for ATTR-CM in a clinicians own practice does result in more testing, as demonstrated by the increase in PYP scans obtained and number of patients diagnosed with ATTR-CM. Future studies should include a larger sample size and a refined screening protocol that identifies a population with a higher prevalence of ATTR-CM, and should factor in provider education.

The delayed diagnosis of ATTR-CM impacts a patient’s quality of life and contributes to financial burdens and worse prognosis. Misdiagnosis delays the time to initiation of appropriate therapy and increases the risk of inappropriate treatment. Typical heart failure therapies, including beta-blockers, angiotensin-converting enzyme inhibitors, or angiotensin II receptor antagonists, are often ineffective in this patient population and may even lead to clinical worsening [9]. Disease-modifying therapies acting on various steps in the amyloid production process exist for ATTR-CM. TTR stabilizers, such as Tafadimis and Diflunisal, bind to TTR tetramers, preventing dissociation into monomers and decreasing amyloid formation. TTR silences, such as Inotersen and Patisiran, suppress hepatic TTR mRNA, reducing TTR amyloid production. Early and correct diagnosis of ATTR-CM allows for targeted, appropriate treatment [10].

Strengths and limitations

A limitation of our study is our EMR system which collected limited non-cardiac disease data and thus hindered our ability to produce an effective algorithm, though this did not stop us from producing an effective measure for lowering our clinicans’ diagnostic thresholds. Second, there are many disease phenocopies to amyloidosis, and thus, the characteristics used to define high-risk patients are likely not specific enough to TTR amyloidosis to produce an effective algorithm.

This study was designed to test the hypothesis that identifying and then alerting physicians of patients at high risk of ATTR-CM would impact the general ordering practices of physicians for testing for ATTR-CM. However, increased screening does not necessarily guarantee improved outcomes. The ability to improve individual patient outcomes will depend on several factors, most notably the stage at diagnosis, as early identification and treatment offers the best hope for meaningful clinical outcomes [11]. Clinicians should focus on early detection of the disease in order to prolong the lives of their patients. Unfortunately, many of the clinical diagnostic clues used to guide physicians in recognizing this devastating disease are not seen until the disease has already reached advanced stage. Future iterations of the current screening model, consisting of more complex variable analysis, coupled with machine learning and artificial intelligence may potentially produce a better algorithm.

One of the strengths of the trial design was reduction of bias. Previous studies have found that even the language used in the echo report significantly influences screening rates for cardiac amyloidosis. Simply mentioning ‘amyloid’ in the TTE report, as opposed to less specific terms, increased the likelihood of evaluation for amyloidosis by nearly 90% . Our study bypassed this barrier by directly informing the providers of the possibility of amyloid and took it one step further by providing guidance on next steps to guide their workup, although the ultimate decision to screen the patients was left to the discretion of the individual provider. Multiple studies have noted racial disparities in screening and outcomes in black patients, who tend to present with more advanced disease including more severe heart failure due in part to them being screened half as frequently as their white counterparts despite similar echocardiograms [12, 13]. On analysis, our providers independently chose a substantial black majority of patients for testing. On the background of having a 50% overall black patient population, perhaps this trend in testing marks getting one small step closer towards bridging the racial disparity gap.

Conclusion

In summary, our EMR-based algorithm was successful in increasing clinicians’ suspicion for ATTR amyloidosis, despite not being useful as an independent algorithm, which resulted in significantly more providers ordering testing for ATTR amyloidosis and more positive tests.