What I Learned about Anti-Black Racism from the Digitalis Data Set
In 2020, during the first year of my Master’s program, in a course on basic statistical methods, I needed to produce a final project that utilized a variety of statistical techniques to analyze an existing data set. As an RN with experience working with heart failure patients, I thought the digitalis data set would be an interesting choice, and that I had an above-average chance of being able to use my domain knowledge to craft, refine, and interpret the results of a useful model applied to this data. What I didn’t know when I started the project, was that I would come away from it with data-driven evidence of systemic racism in our health care system, and its ongoing impact on Black lives.
Of course, like many people with an interest in the history of the American health care system, including its shortcomings, I was aware of Linda Villarosa’s important articles in the New York Times on both the crisis of maternal mortality among Black women, and ongoing beliefs in false racial differences that continue among health care providers — beliefs that, for no good reason, fuel ongoing disparities in care. I believed these reports, and already strove to speak up for my patients wherever I could make a difference in their care, but I never saw data speak directly to me on this subject until the day I chose to compare rates of hospitalization for worsening heart failure in Black and Hispanic people to those of white people enrolled in this study.
The Digitalis Investigation Group (DIG) study was a prospective, randomized clinical trial involving 7,788 patients with congestive heart failure who were randomized to treatment with digoxin or a placebo and then followed for an average of 37 months. Data collection began in June of 1990 and continued until June of 1998, and patients were enrolled between February of 1991 and September of 1993. In the initial study, interactions between age and clinical outcomes were studied, including mortality, all-cause hospitalization, hospitalization for any cardiovascular cause, hospitalization for worsening heart failure, and hospitalization for digoxin toxicity, among others. Since the study reported its primary outcomes in the New England Journal of Medicine in February of 1997, the study has been cited more than 1000 times and 20 additional distinct studies have been published analyzing various aspects of the data collected.
One of the things that makes the Digitalis Investigative Group data somewhat unique is the lengths the researchers went to in attempting to recruit a sample of patients from a variety of hospitals that would approximately proportionally represent the demographic composition of the US population with congestive heart failure. (This should be a typical research practice, but we all know it generally is not.) Due to the involvement of the VA hospital system and the large number of minorities who get their health care from the VA, the study was hoping to recruit 15–20% non-white patients. In the end, about 12% of the patients enrolled were African American or Hispanic. All the participants were required to have an EF (ejection fraction) < 45% and to be in a normal sinus rhythm at the beginning of the trial. All patients enrolled were randomized equally to receive either digoxin or placebo.
As I compared the outcomes for hospitalization for worsening heart failure among the demographic groups present in the data set, I found a statistically significant difference between these outcomes for white people vs. People of Color. The People of Color, most of them self-identifying as Black or African-American, were significantly more likely to be hospitalized for a worsening in their condition. I decided to look into this and try to see what the data, and the literature, had to say about why this might be the case. What I found convinced me once again that medicine and research still have a long way to go before they can claim fair treatment of and best practices with respect to Patients of Color.
In spite of the statistically significant differences in outcomes for non-white vs. white heart failure patients in the study, the patients had many similarities. The mean EF for the sample was 28%, the min was 3% and max 45% (per the study recruitment criteria). Mean cardiothoracic ratio (represented in this study by a variable called “CHESTX”) — often a measure of the amount of thickening of the left ventricle, usually due to pumping against vascular resistance, or heart enlargement in heart failure patients — was .52 for the sample. Mean heart rate, systolic and diastolic pressure were normally distributed, with mean values 79, and 126/75 respectively. Mean diagnostic duration was 30 months, and unlike most other clinical heart failure trials reviewed for this paper, mean New York Heart Association functional class was II.
Other notable facts about the data set include the nearly equal distribution of diabetes (28% overall) between white and non-white patients in the sample, as well as normal and nearly equal BMI distribution (70% of the male sample had BMI 18.5–29.9, placing them in the normal to overweight categories). Regarding prevalent conditions, 45% of the sample were identified as hypertensive, 27% had angina, and 65% had had a prior MI. While 50% were provided with digoxin (were in the treatment group), 77% were on diuretics, 94% were on ACE inhibitors and 43% were on nitrates. Only 7.3% were being treated with thiazide diuretics, 2% were on hydralazine and 1% took another vasodilator. So what were the possible reasons for the worse outcomes in non-white patients?
The person who deserves credit for putting me on the right track to sleuthing out the revealing facts within the data set is Dr. Chinyere Obimba, MD MPH of the University of Washington Medical School. In February of 2020, Dr. Obimba gave a Grand Rounds presentation that was preserved on YouTube, where she went over in detail the disparities between the health of white Americans and Americans of Color, particularly Black Americans and discussed in detail disparities in treatment that have long been justified by erroneous beliefs in treatment response or physiological differences between white and non-white patients. After watching this video a few times, I was able to conduct a revealing literature review.
Sure enough, one of the most notable variations in the data was the significantly higher proportion of white patients with heart failure of ischemic cause (78%), while only 42% of the non-white patients had heart failure of ischemic cause, but 27% had heart failure of hypertensive cause. This means that the vast majority of the white patients had heart failure because either a clot or plaque-blockage in a vessel in the heart led to a heart attack, causing some tissue to die and the heart to no longer pump as well. While 42% of the non-white patients also had this cause, 27% (almost a third) had a very different cause — high blood pressure in their systemic circulation produced a situation where their hearts got ‘worn out’ pumping against high resistance. Imagine you had a hose that was full of water at high pressure, and you were trying to push more water into that hose. You’d have to push really hard in order to force more liquid in and that would make you tired fast. This is what happens to the heart in hypertensive heart failure, the action of pushing blood into an already high-pressure system enlarges and fatigues the heart. This gap in etiology between white and non-white patients in the study is concerning because there were almost no differences in treatment strategy among all patients — patients were getting mostly similar treatments regardless of the primary cause of their heart failure.
In 1986, Cohn et al.’s V-HeFT trial found a reduction in mortality and hospitalization among heart failure patients who were treated with hydralazine and isosorbide dinitrate (also called ‘I-H therapy’). In spite of these findings coming out 7 years before the DIG trial’s enrollment period, study physicians were prescribing I-H therapy or thiazides to treat only very few of the patients enrolled in this study, even when hypertension was the etiology for their heart failure — a decision that would clearly have a greater impact on patients of color.
My literature review of the state and history of cardiovascular medicine ended up confirming both Dr. Obimba’s presentation and Linda Villarosa’s findings — a lot of what has been previously published about physiological differences between treatment responses in self-identified racial groups have only anecdotal basis in fact, yet people are still — to this day — treating patients and speaking about differences in medication efficacy or physiologic response to treatment as if they were based on something more.
The most recent findings made possible by a greater understanding of genetic variation among perceived ethnic and racial groups, casts a lot of doubt on whether previous race-based guidelines should routinely be used in choosing treatment regimens for African American patients. In 2008, in a comprehensive literature review on ethnic differences in cardiovascular drug response, Pharmacist Julie Johnson astutely pointed out that the perceived increased efficacy of the nitrate/hydralazine combination in African American patients described by A-HeFT may be because “[B]lacks are significantly more likely to have hypertensive heart failure, whereas the underlying cause is more likely to be ischemic heart disease in whites. Thus, it is possible that [B]lacks and whites with hypertensive heart failure would respond equally well to I-H and that ethnicity per se is not the source of response differences”.
In 2010, Flack et al. again cast doubt on the idea of Black patients being different from whites, citing the large overlap in response to hypertensives between African American and white patients identified by Chrysant in 2003. Additionally, Seghal (2004) cites wide variability in response to hypertensive medications within ethnic and racial groups, a finding that appears to be greatly reinforced by the research conducted by Wilson et al. (2001) and published in the journal Nature Genetics, revealing that perceived racial or ethnic differences could not reliably predict drug metabolism variations, and that these variations were only reliably detected by AI-determined genetic sub clustering.
As can easily be seen in the above table, 21% of the Afro-Caribbean sample and 62% of the Ethiopian sample were in the same genetic subcluster as 96% of Norwegians, 90% of Armenians and 96% of Ashkenazi Jews. Additionally, every subcluster had some (even if very few) members from all the sample populations studied, making it highly improbable that a clinician could determine an individual’s subcluster based on external physical characteristics of race or ethnicity alone.
So, while prior findings on racial differences in efficacy and risk-profile between drugs such as thiazides and lisinopril persist into modern studies and will likely hold true for some Black patients, it is becoming increasingly clear that clinicians will need to do more than follow a treatment guideline when deciding how best to assist their African American and Hispanic patients in attaining blood pressure control or preventing worsening heart failure.
According to the latest information from the American Heart Association, heart failure currently impacts approximately 6 million Americans. The Department of Health and Human Services’ Office of Minority Health cites CDC’s Summary Health Statistics released in 2021 showing that in 2018, Black Americans are still 30 percent more likely to die from heart disease than non-Hispanic whites and less likely to be properly diagnosed (and therefore, properly treated) for the disease. What this means is that the racism I discovered within the digitalis data set lives on, twenty years after that study’s last day of activity, and many of the diagnosis and care issues I found direct evidence of when analyzing this data are still with us to this very day.
It is natural to ask — when will our health care system finally change in response to the truths evidence has revealed time and time again? I believe it is not enough for us to study the problem, or even admit that it exists. We must take the lessons of Wilson et al (2001)’s Nature Genetics article and of Dr. Obimba’s presentation to heart, and advocate tirelessly as nurses, providers, patients and family members for true equality in both care-tailoring and aggressive treatment for heart failure patients. This means we provide care and treatment specific to each cause of heart failure in every patient, and that we look out especially for Black heart failure patients who have been undertreated and overlooked for far too long.
