November 27, 2012
Table of Contents
Richard D. Moore et al. Clinical Infectious Diseases 2012;55(9):1242-51. Read the abstract.
Disparities in health outcomes are pervasive in the U.S., with some in our big pot melting better than others. Racial and ethnic minorities, and those living close to or in poverty, typically get burned. When it comes to HIV care, there have been data pointing to significant disparities, including racial differences, in the rates at which antiretroviral therapy (ART) is prescribed, viral suppression is achieved and mortality is averted.
In an important and provocative paper from the Johns Hopkins University School of Medicine, differences in HIV outcomes, including receipt of combination ART, HIV RNA levels, CD4+ cell counts, opportunistic infections and mortality, were examined and stratified by HIV risk group, sex and race. At their Baltimore, Md.-based clinic, 6,366 patients were followed from 1995 to 2010; about two thirds were men, three quarters were African American, 45% were injection drug users (IDU), 30% were men who have sex with men (MSM) and 60% had either Medicaid or private insurance.
Over the course of the period studied, there was an increasing trend for receipt of ART across all demographic and behavioral risk groups, with uptake occurring earlier and more briskly among MSM, all men and white patients. However, by 2010, 87% of the clinic cohort was prescribed ART, and there was no significant difference (after multivariable adjustment) seen in receipt of ART among the groups.
Looking at HIV RNA levels, by 2010 the median viral load was less than 200 copies/mL for all demographic and risk groups. In multivariate analysis, the log10 HIV RNA level was slightly but significantly higher -- by 0.28 log in the IDU group compared to the MSM group in 2010 (P < .001), although there was no significant difference between the MSM and heterosexual groups. The log10 HIV RNA level was also 0.24 log higher in blacks compared to whites in 2010 (P < .001). There was no difference by sex.
Similarly, median CD4+ cell counts were high, at almost 500 cells/mm3 by 2010. Slightly lower counts were seen among IDU compared to MSM, as well as among men compared to women. Opportunistic infections became virtually extinct by 1998 and there were no differences between groups seen in more recent years.
Lastly, there was no significant difference in mortality among the groups. Life expectancy computed for the entire sample found that, for a 28-year-old HIV-infected patient in their care in 2009, remaining life expectancy was calculated to be 45.4 years (95% confidence interval, 39.6-51.3 years). That is, a 28-year-old patient -- regardless of race, gender or risk category -- could be expected to live to 73 years of age.
Other studies have detected differences that were not seen in the analyses of this cohort; however, the Hopkins study includes more current data. There was a striking "catch-up" effect among minorities, women and IDU over time: By 2010, the benefits of ART and advances in care appeared to be universal regardless of race, gender and risk category.
These results fly in the face of overall health disparity trends in the U.S., wherein African Americans and those with lower socioeconomic status have higher rates of morbidity and mortality. Although HIV disproportionately affects those same populations that are at risk for worse outcomes and could be expected to compound these disparities, this study finds the opposite.
So, why do populations vulnerable to disparities in health care outcomes fare better when it comes to HIV care? Two words: big government. While that term may often find itself slung like an insult in a political scrap, in the case of HIV, it is public funding through the Ryan White Care Act that provides the primary care, supportive care and medication assistance that are critical to achieving the very outcomes studied by the Hopkins investigators.
HIV is a complex and expensive disease to manage, yet the results coming from Baltimore are mirrored in clinics across the nation precisely because we have a national safety net that sets up those with HIV for success rather than failure.
While the investigators stress their finding of little to no difference in HIV outcomes across populations, what they also discovered is how different HIV is from other serious conditions in achieving this degree of parity. This lesson should be Exhibit A when the Ryan White Care Act comes up for reauthorization, and as we look to ways in which the Affordable Care Act will be implemented.
Ank E. Nijhawan et al. J Acquir Immune Defic Syndr 2012;61:349-358. Read the abstract.
It is sad that the house staff at my hospital -- those who admit patients with HIV infection -- see a side of the disease that is completely warped compared to the reality of widespread suppression of HIV RNA levels, boosted CD4+ cell counts and jolly, round faces that characterize my clinic population. While these interns and residents have heard of the fruits of advances in HIV care, they might as well be in a third-world nation as they take histories from patients gasping with Pneumocystis pneumonia or battling an aggressive lymphoma.
Few medical residents spend time in our outpatient clinics, and thus they experience HIV through the lens of "difficult cases" -- the non-adherent, the out-of-care and the very unlucky. It is no wonder that few young physicians are gunning to become HIV docs.
Further disheartening to these clinicians are the "frequent flyers": patients who are admitted, treated and released, but "bounce back" soon after discharge with the same -- or a new -- set of problems and issues.
To get a better sense for which patients with HIV infection tend to return after a hospital stay, a team from the Dallas Parkland Hospital system examined hospital records of HIV-infected inpatients from 2005 to 2008 in an attempt to develop a predictive model of 30-day risk of readmission or death.
Looking at 1,500 cases (three quarters African American, 16% Latino, 60% male, less than half with outpatient medical insurance), one quarter were readmitted. We should let that settle in: One in four patients with HIV in this medical system were readmitted within 30 days of discharge. Dallas, we have a problem.
Delving into the details, a range of clinical factors were associated with risk of readmission, including AIDS-defining illness, CD4+ cell count less than 92 cells/mm3 within 12 weeks of the first admission, creatinine greater than 1.77 mg/dL, HCO3 less than 19 mmol/L, ALT or AST greater than 35 U/L, HCT less than 28.4 or greater than 48.8%, anion gap greater than 12 mmol/L, and an ALC less than 0.34x109/L (probably reflecting low CD4+ cell count).
In addition, and tellingly, patients were more likely to be readmitted within 30 days of release if they were homeless, lived more than a dozen or so miles from the hospital, were insured by Medicaid (i.e., were disadvantaged or sick enough to qualify) or had higher rates of utilization of inpatient and emergency department services.
Three percent (74 patients) of those examined were known to have died within 30 days of hospital discharge. Factors associated with mortality were basically the same laboratory abnormalities I listed above, including immunosuppression.
This study is interesting in that the elements included in its models for readmission are readily available to clinicians. Recognizing these red flags for readmission during hospitalization should prompt discharge planning that addresses this risk. Careful coordination of post-release services -- including immediate, community-based medical care appointments and check-ins by care coordinators -- could go a long way toward reducing bounce backs. Of course, not releasing patients until they are really ready to be discharged, while difficult, will also help.
Gilles Wandeler et al. Clinical Infectious Diseases 2012;55(10):1408-16. Read the abstract.
You do not have to be that old to recall when hepatitis C (HCV) was known as "non-A non-B hepatitis" and was considered an emergent stalker of blood banks, tissue banks and dialysis machines. Even after the wastepaper-basket moniker was traded for the next available letter in the alphabet, we continued to consider this virus as either a nosocomial infection or one that was spread outside of clinical settings by needle sharing and tattooing.
More recently, we have come to understand that HCV can be transmitted sexually -- particularly by MSM. Investigators from the Swiss HIV Cohort Study examined data collected from 1998 to 2011 to gauge HCV incidence among HIV-infected MSM, heterosexuals and IDU, excluding IDU in the other categories. Since 1998, HCV antibody testing has been conducted every other year during cohort study visits on everyone not already known to be HCV infected.
At first test, 3% of the 4,629 MSM, 11% of the 4,530 heterosexuals and a whopping 92% of the 2,678 IDU screened positive for HCV infection. A total of 6,534 patients, of whom 3,333 were MSM, 123 were IDU and 3,078 were heterosexuals, were included in the HCV incidence analyses.
Over a total follow-up period of 23,707 patient-years (py) for the MSM group, 733 py for the IDU group and 20,752 py for the heterosexual group, 101 (3.0%), 41 (33.3%) and 25 (0.8%) patients, respectively, experienced an HCV seroconversion during follow-up. Incidence rates increased among MSM by 18-fold (!), while they fell for IDU from 1998 to 2011; there were very few cases among heterosexuals overall.
Predictors of HCV infection among the MSM included past syphilis and hepatitis B (HBV) infection, as well as inconsistent condom use.
These results punctuate a few important points:
Jennifer H. Han et al. AIDS 2012; 26:2087-2095. Read the abstract.
There are mixed data regarding the presence of a heightened risk of diabetes mellitus for persons living with HIV. As with other metabolic conditions, there can be considerable confounding factors that make associations difficult to establish. Certainly, HIV does not protect against the development of diabetes in those destined to this fate.
In a clever paper from the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS) cohort, a study of HIV-infected patients in clinical care nationwide, responses to type-2 diabetes therapies were assessed in 286 HIV-infected patients and 858 age- and sex-matched controls without HIV infection. The HIV-infected cases were those in care at any of the 8 CNICS sites, while the controls were enrolled in a large cohort study out of Philadelphia. For both cases and controls, ICD-9 codes and prescriptions of diabetes-specific medications were used to establish prevalent disease, and the initial date of first-ever diabetes therapy was ascertained.
The main outcome was change in HbA1c over the first year of therapy. The patients with HIV infection had lower mean baseline HbA1c values: 7.82% (standard deviation [SD] 2.3) versus 8.62% (SD 2.4), respectively; P < .001. The most common diabetes treatment was metformin, followed by sulfonylureas. HIV-infected patients achieved significantly smaller decreases in HbA1c compared to patients without HIV infection, with an absolute mean difference in HbA1c reduction of -0.17% during the first year of treatment (95% CI, -0.28 to -0.06; P < .003), after adjustment for baseline HbA1c. However, in adjusted subgroup analyses, the difference in the change in HbA1c with therapy between HIV-infected and HIV-uninfected patients was seen in those on a protease inhibitor, but was not seen in those on regimens excluding this class of antiretrovirals.
Interestingly, despite the observed smaller decrease in HbA1c for the HIV-infected patients, they were more likely to achieve a value that is below the American Diabetes Association goal of 7% for this parameter. This finding is partly explained by the fact that the HIV-infected folks started with a lower HbA1c. In addition, others have described an HbA1c-glycemia disconnect in HIV, wherein the HbA1c underestimates the level of hyperglycemia. A paper by Peter S. Kim et al in Diabetes Care associated elevated mean corpuscular volume and abacavir (Ziagen) use with this disconnect.
Overall, these data show that there was a small but significant difference in response rates, and that protease inhibitor therapy was implicated as a culprit. It is unclear whether these drugs are having a direct effect on glycemic control or whether those taking protease inhibitors were different in some other ways (e.g., more advanced HIV disease, longer duration of treatment, higher number of risk factors for diabetes).
The results also suggest that treatment of diabetes and other cardiovascular disease risks may need to be more aggressive among those living with HIV.
Mathieu Bastard et al. PLoS ONE; 7(11): e49091. Read the full text.
Predicting adherence to medications is a fool's game. Who knows how many patients I have seen -- smiley people with winning personalities, some of whom even send me holiday cards -- who I would wager are great at taking their meds, yet who secretly end the month with more than a few extra tablets in their bottles. Most of these folks are lucky in that their viral loads are as suppressed as their mischief, but their non-adherence can catch up with them.
The most common method to assess adherence is to ask people about missed doses, which is helpful only when the beans are spilled; otherwise, such a survey method is useless (see the next summary for more on this). There are some nifty ways to gauge adherence, including pill counts, pharmacy refills and checking drug levels in blood or even hair. However, these are not perfect, nor are they always practical and available.
Investigators from Médecins Sans Frontières (MSF) looked at something much more simple: whether patients had delays in making their clinic appointments. They calculated a metric of the number of appointments attended, with lateness (in days) divided by the number of months between ART initiation and date of virological testing, and then multiplied the result by 100 (simple!). Applying this metric, they categorized 3,580 adults and 253 children receiving care at MSF clinics in Africa and Asia as "good," "moderate" or "poor" adherers to clinic appointments.
In their analysis, 58.0% of patients showed good adherence, 35.6% showed moderate and 6.4% showed poor. Patients in the poor and moderate categories were two to three times more likely to experience virological failure and HIV drug resistance compared to those within the good category -- even after adjusting for initial age and CD4+ cell count, previous ART experience, type of regimen and tuberculosis diagnosis at start of therapy. Results were similar in children. Interestingly, the delays in making it to an appointment at an MSF clinic were not huge -- they were generally on the order of a few days.
It can be debated whether these results, obtained from facilities in some of the poorest places on Earth, can be translated to our more cushy world of flat screen-equipped waiting rooms and automated appointment reminders here in the U.S. However, I suspect the same concepts apply -- and, although the forces that make getting to a doctor's appointment a challenge may be different in Burkina Faso than in Buffalo, N.Y., they nonetheless impact medication adherence regardless of geography.
It is the patient who misses visits or shows up late for his/her clinic appointments that we should watch carefully. These data and our own experiences tell us they may be at greatest risk of falling off the ART wagon.
Betsy J Feldman et al. AIDS Behav 2012: epublished ahead of print. Read the abstract.
Looking at your patient's appointment tardiness may be useful, but is likely to be specific and not sensitive to suboptimal medication adherence. The same can be said for self-reported adherence. Although asking folks how well they are taking their meds is standard operating procedure in clinical research, so is fibbing; for most clinicians, their patients' responses are considered about as reliable as those dealing with number of beers drunk and when crack was last smoked (i.e., not).
What if there was a question that patients did respond to and, like Wonder Woman's lasso, revealed all?
A group of collaborators from the CNICS cohort don't have anything nearly so foolproof, but they do suggest that one single question was as good as, or even better than, longer batteries of adherence questions.
They looked at data collected during self-reported assessments completed during routine clinic visits by those on ART who attended clinics in Birmingham, Ala., or Seattle, Wash. These assessments are done using a touch-screen tablet and include surveys of medication adherence, substance/alcohol use and depression. The medication adherence questions include an item that, in at least one study comparing it to the gold standard (electronic caps that record the opening of a medication bottle), was found to produce less overestimation of adherence.
What is this question, you may ask? The way the authors refrain from explicitly reporting it in this paper, you would think it was David Petraeus' personal email account password. Fortunately, I found the actual wording in the paper describing the electronic cap study, and the amazingly revealing question is fiendishly simple: "Rate your ability to take all your medications as prescribed" over the past month (in one of six categories: very poor, poor, fair, good, very good and excellent). That's it.
In the CNICS study, this one question was compared to the other adherence survey items, as well as viral load results, in 2,399 patients. Most patients rated their adherence to be pretty decent, with very few rating their adherence as "poor" or "very poor." Among those reporting poor or very poor adherence, 51% had a detectable viral load; by comparison, a detectable viral load was found among 42% of those reporting fair adherence, 23% of those reporting good adherence, 13% of those reporting very good adherence and 9% of those reporting excellent adherence.
The single-question responses correlated well with other survey items and, in terms of predicting viral load suppression, the single question was just as good as using a visual analogue scale to indicate adherence -- but it was quicker to complete, clocking in at a mean of just 13.5 seconds.
Who knew that asking a direct question could lead to more honest answers? Of course, almost 1 in 10 of those rating their adherence over the past month as "excellent" had a detectable viral load, raising suspicious eyebrows. But the study suggests that we can do away with longer and more cumbersome surveys when this one question is used. At 13.5 seconds, it may be worth a try.
Copyright © 2012 Remedy Health Media, LLC. All rights reserved.