Jeffrey Kohlwes

Test characteristics of α-fetoprotein for detecting hepatocellular carcinoma in patients with hepatitis C: A systematic review and critical analysis

People with hepatitis C virus (HCV) have a 2% annual risk and a 7% to 14% five-year risk for hepatocellular carcinoma (1-3), a tumor with an estimated median survival duration of 4.3 to 20 months after diagnosis (4-7). Some studies suggest a possible survival advantage when small tumors are detected (8, 9), but no randomized, controlled trials of screening for hepatocellular carcinoma in patients with HCV have been conducted. Although the National Cancer Institute currently recommends against screening for hepatocellular carcinoma (10), many physicians currently screen high-risk populations with various strategies, including serum -fetoprotein (AFP), ultrasonography, and computed tomography (11). The use of AFP, a tumor marker variably secreted by hepatocellular carcinomas, to detect these tumors has been widely debated (12-14). Many conclude that AFP is not a useful diagnostic test (12, 15), but AFP continues to be commonly used (11). To determine a summary estimate of the test characteristics of AFP for detecting hepatocellular carcinoma in patients with HCV, we conducted a systematic review. Methods Study Search Protocol We performed a MEDLINE search from 1966 through December 2002 for English- and nonEnglish-language articles using the following search terms: hepatitis C, hepatocellular carcinoma, screening, diagnosis, -fetoprotein, sensitivity, and specificity. Bibliographies of all reviewed articles were searched to identify additional relevant titles. Titles that mentioned hepatocellular carcinoma or HCV and screening were identified for abstract review. Abstracts that described the use of AFP as a diagnostic or screening test for hepatocellular carcinoma were marked for full article review. Inclusion and Exclusion Criteria Study designs accepted for analysis included randomized, controlled trials, cohort studies, or casecontrol studies that used AFP to detect hepatocellular carcinoma in HCV-infected patients with or without cirrhosis. We required that the authors report sensitivity and specificity for the use of serum AFP (or data sufficient to calculate these test characteristics) and that they identify some gold standard for diagnosis. Computed tomography, magnetic resonance imaging, histopathology, and disease-free time greater than 2 years were considered adequate gold standards. Ultrasonography was not considered an adequate gold standard because its sensitivity for hepatocellular carcinoma is controversial (12, 14, 16, 17). Studies were excluded from analysis if the cause of viral hepatitis was unclear, if at least 50% of the study patients did not have HCV, and if the same data were presented in a separate article by the same investigators. Data abstracted were study design, cause of hepatitis, whether the AFP test was used for diagnosis or screening, type of gold standards used, percentage of the study sample with cirrhosis, and reported sensitivity and specificity of AFP for detecting hepatocellular carcinoma. Analysis To grade the quality of evidence for use of serum AFP as a screening test for hepatocellular carcinoma in patients with HCV, we independently determined study design, whether application of the gold standard for each study was blinded to AFP result, whether the patient selection was independent, the type of gold standard implemented, and presence of partial verification bias. Disparity in grade was resolved by discussion and consensus among all three authors. Results A total of 1239 titles were identified, 55 relevant abstracts were reviewed, and 18 articles were identified as potentially relevant. Five studies met all inclusion criteria and were included in the analysis (15, 18-21). Of the 18 potentially relevant articles, 5 were excluded because they were uncontrolled case series (22-26), 6 were excluded because most study patients did not have HCV (27-32), 1 was excluded because it did not identify the cause of hepatitis in all study patients (33), and 6 were excluded because they did not provide both sensitivity and specificity data for AFP in their study sample (8, 34-38). Characteristics of the 5 studies meeting all inclusion criteria and no exclusion criteria are shown in Table 1 (15, 18-21). Table 1. Characteristics of Included Studies Two studies were prospective cohort studies (15, 18) and 3 were casecontrol studies (19-21). One study (15) universally applied an acceptable gold standard test to both case-patients and controls, and each study used a different gold standard. Table 2 presents abstracted sensitivity and specificity data and abstracted or calculated positive and negative likelihood ratios for the diagnosis of hepatocellular carcinoma at an AFP cutoff value of 20 g/L. This cutoff value was chosen because each included article provided data for a cutoff value of 20 g/L, and an AFP level of 20 g/L is considered a level that prompts further testing (19). Other cutoff values were not reported in every article. Exclusive data for patients with HCV were available for all studies but one (18); for the latter study, only combined data for patients with HCV and hepatitis B virus were available. Table 2. Abstracted Test Characteristics of -Fetoprotein Levels Higher than 20 g/L for Detecting Hepatocellular Carcinoma Sensitivity of AFP levels higher than 20 g/L ranged from 41% to 65%, while specificity ranged from 80% to 94%. Positive likelihood ratios for AFP levels higher than 20 g/L ranged from 3.1 to 6.8 and negative likelihood ratios ranged from 0.4 to 0.6. Table 3 shows the sensitivity and specificity data for an AFP cutoff value higher than 200 g/L, a value that is frequently reported to be specific for the diagnosis of hepatocellular carcinoma (19, 21). Four of the 5 studies reported sensitivity and specificity for this cutoff value. The range of specificities was very high at this cutoff value (99% to 100%), but the sensitivity was very low (20% to 45%). Table 3. Abstracted Test Characteristics for -Fetoprotein Levels Higher than 200 g/L for Detecting Hepatocellular Carcinoma Discussion Our systematic review of the literature shows that the quality of evidence describing the characteristics of AFP as a diagnostic test for hepatocellular carcinoma in patients with HCV is limited. Three of the reviewed studies were casecontrol studies (19-21), which potentially overestimate the sensitivity and specificity of the test in question (39, 40). In contrast, cohort studies are less susceptible to bias because they are more likely to include patients with a varying spectrum of disease, particularly those patients who present more subtly, and therefore more closely reflect the manner in which a test will be implemented in clinical practice (39). Two studies (15, 20) may have partial verification bias, which occurs when the result of the test being evaluated (in this case, AFP or ultrasonography) influences the decision to administer the gold standard test (39, 40). This may falsely elevate sensitivity and specificity (39). Four of five studies (18-21) applied a gold standard of uncertain validity to both case-patients and controls, resulting in a possible underestimation of disease prevalence and an unknown ultimate effect on sensitivity and specificity. Blinding was not reported in four studies (15, 19-21) and may have affected interpretation of gold standard test results. Without systematic blinding, investigators may be more vigilant in applying gold standards to those patients with positive test results and thereby falsely elevate specificity (39, 40). Finally, four studies (15, 18, 20, 21) included patients with and without cirrhosis. Patients with cirrhosis have a higher risk for cancer (41) but commonly have elevated levels of AFP thought to be unrelated to hepatocellular carcinoma (42), leading to an unknown effect on sensitivity and specificity. Notably, one excluded study reported sensitivity of 80% and specificity of 95% for AFP levels higher than 10 g/L applied to a subgroup of patients with histologically severe liver injury (35). Given the significant concerns about the validity of the data generated by the studies reviewed, we could not calculate conclusive summary estimates of the sensitivity and specificity of AFP as a diagnostic test for hepatocellular carcinoma. The biases previously mentioned that affect the reported sensitivities and specificities tend to overestimate the utility of AFP as a diagnostic test, but to guide current interpretation of AFP in practice, we can consider the use of this test if these best-case estimates are true. The most common use of AFP is to screen for hepatocellular carcinoma in asymptomatic patients with HCV. In this scenario, reported prevalence data indicate a pretest probability of hepatocellular carcinoma in patients with HCV is 5% to 12% (41, 43). Using a prevalence of 5% with the range of positive likelihood ratios for an AFP level higher than 20 g/L (3.16.8), results in a post-test probability of 14% to 25%, while an AFP level lower than 20 g/L results in a post-test probability of 2% to 3%. Although a post-test probability of 25% would prompt further work-up with imaging, a post-test probability of 2% is unlikely to be reassuring enough to preclude the use of other screening strategies, including ultrasonography or computed tomography. The other common use of AFP involves the evaluation of patients presenting with one or more high-risk features, including a hepatic nodule (found incidentally or with a screening test) or decompensated liver failure. Data for AFP at higher cutoff values, such as an AFP level higher than 200 g/L (Table 3), suggest that AFP, although not sensitive, can be highly specific for hepatocellular carcinoma. A low AFP level (<200 g/L) would not be informative enough to stop further search for hepatocellular carcinoma, but an AFP level higher than 200 g/L would strongly suggest that cancer is present, allowing for earlier counseling of a patient. In addition to the quality of articles review

The prognosis and treatment of acquired hemophilia: a systematic review and meta-analysis

The inhibition of factor VIII by autoantibody development, or acquired hemophilia, occurs in approximately one person per million each year and can cause life-threatening bleeding. Due to the disease rarity, there are no randomized studies addressing prognostic features and treatment. The goal of this study is to identify prognostic indictors in acquired hemophilia to guide treatment choices. MEDLINE and EMBASE search from 1985–2008 retrieved 32 studies with detailed clinical information on five or more patients with acquired hemophilia. Univariate and multivariate analysis of the effects of age, sex, underlying condition, inhibitor titer, and treatment regimen were evaluated with regards to complete remission and death. A total of 32 studies containing 359 patients with acquired hemophilia were included in the analysis. The all-cause mortality rate in this cohort was 21%. Multivariate analyses revealed that patients more likely to die are the elderly [odds ratio (OR) 2.4, 95% confidence interval (CI) 1.32–4.36] and those with underlying malignancy (OR 2.76, CI 1.38–5.50). Early achievement of complete remission resulted in improved survival. Complete remission occurred in 94% of patients receiving combination chemotherapy, 82% receiving dual therapy, and 68% receiving steroids alone. Patients receiving immunosuppression had reduced odds of persistent hemophilia, with combination chemotherapy being the most efficacious (OR 0.04, CI 0.01–0.23) and steroid therapy alone being the least (OR 0.38, CI 0.14–0.94). In acquired hemophilia, increased age, underlying malignancy, and lack of complete remission are risk factors for death. Although the included studies were not randomized, patients treated with combination chemotherapy had the greatest odds of remission and the lowest odds of death.

Supervising the Supervisors—Procedural Training and Supervision in Internal Medicine Residency

BACKGROUNDAt teaching hospitals, bedside procedures (paracentesis, thoracentesis, lumbar puncture, arthrocentesis and central venous catheter insertion) are performed by junior residents and supervised by senior peers. Residents’ perceptions about supervision or how often peer supervision produces unsafe clinical situations are unknown.OBJECTIVETo examine the experience and practice patterns of residents performing bedside procedures.DESIGN AND PARTICIPANTSCross-sectional e-mail survey of 653 internal medicine (IM) residents at seven California teaching hospitals.MEASUREMENTSSurveys asked questions in three areas: (1) resident experience performing procedures: numbers of procedures performed and whether they received other (e.g., simulator) training; (2) resident comfort performing and supervising procedures; (3) resident reports of their current level of supervision doing procedures, experience with complications as well as perceptions of factors that may have contributed to complications.RESULTSThree hundred sixty-seven (56%) of the residents responded. Most PGY1 residents had performed fewer than five of any of the procedures, but most PGY-3 residents had performed at least ten by the end of their training. Resident comfort for each procedure increased with the number of procedures performed (p < 0.001). Although residents reported that peer supervision happened often, they also reported high rates of supervising a procedure before feeling comfortable with proper technique. The majority of residents (64%) reported at least one complication and did not feel supervision would have prevented complications, even though many reported complications represented technique- or preparation-related problems.CONCLUSIONSResidents report low levels of comfort and experience with procedures, and frequently report supervising prior to feeling comfortable. Our findings suggest a need to examine best practices for procedural supervision of trainees.

Off-label use of rituximab in a multipayer insurance system.

PURPOSE Off-label prescribing in oncology is common and unregulated. The aim of this study was to describe the off-label use of rituximab, a novel anti-CD20 antibody, among patients from a large proprietary insurance database to understand how frequently and appropriately off-label prescribing occurs for this medication. PATIENTS AND METHODS In this descriptive study, 11,232,642 patients were enrolled in the D2 Hawkeye commercial insurance database between 2001 and 2007, and 2,782 patients received rituximab. The main outcome measures were quantity and type of off-label usage, and expenditures for off-label usage. RESULTS Seven hundred five (25.3%) patients received rituximab for off-label indications, and of those, 332 (47.1%) received rituximab for uncertain or inadequate evidence-based diagnoses. Expenditures for off-label indications were 17.1% of expenditures for rituximab usage. CONCLUSION The frequent use of rituximab for off-label indications should lead to improved postapproval surveillance of biologics by the US Food and Drug Administration, so that use can be adequately studied. This will also facilitate improved regulatory mechanisms to ensure evidence-based use.

Psychotropic drug prescribing for hospitalized patients with acquired immunodeficiency syndrome

PURPOSE The purpose of this study was to investigate the prescribing practices for the use of psychoactive medication in treating hospitalized patients with the acquired immunodeficiency syndrome (AIDS). PATIENTS AND METHODS The medical charts were studied for all patients admitted to a 20-bed AIDS inpatient ward from July through December 1986. One hundred and fifty-one patient-admissions comprised the sample. The average age of the patients was 37 years, and the average length of the hospital stay was 13 days. Retrospective chart review collected demographic data, length of stay, medical diagnosis, psychiatric history, and mental status on admission. Data on psychoactive drugs included the reasons for use, maximum daily dose and duration, and side effects ascribed to the drugs. RESULTS Psychoactive drugs were used in 89% of the cases. Anxiolytics accounted for 49% of the psychoactive prescriptions, and hypnotics made up 43% of these prescriptions. The five most frequently used psychoactive drugs were benzodiazepines. The most frequent reasons for psychotropic prescriptions were insomnia (39%), psychologic distress (20%), and nausea (16%). The most frequently used anxiolytic and anti-psychotic medications were used in moderate dosage, while the most frequently used antidepressant was prescribed in low dosage. CONCLUSIONS Hospitalized AIDS patients are highly likely to be prescribed a psychotropic medication, especially an anxiolytic or hypnotic. Anxiolytics are used for several purposes, including reduction of nausea associated with the use of antibiotics. Antidepressants and antipsychotics are rarely used. Practitioners must be sensitive to the presence of psychiatric conditions presenting as insomnia or disturbed mood so that the most specific and appropriate treatments can be used.

The Impact of a “Low-Intensity” Versus “High-Intensity” Medical Intensive Care Unit on Patient Outcomes in Critically Ill Veterans

Objective: To determine whether a low-intensity versus high-intensity medical intensive care unit (MICU) format in a Veterans Affairs (VA) hospital setting improves patient outcomes, as measured by duration of mechanical ventilation (MV), ventilator-free days (VFDs), and hospital mortality. Design: Retrospective cohort study. Setting: Medical intensive care unit at the San Francisco Veterans Affairs Medical Center (SFVAMC). Patients: On July 1, 2004, the SFVAMC transitioned from a low-intensity MICU to a high-intensity MICU. All patients admitted to the MICU who required MV for 18 months before (n = 96) and 18 months after (n = 131) the transition were included in the analysis. Measurements: We prospectively defined the primary outcome measure as the difference in the median duration of MV between groups. Secondary outcomes included VFDs and hospital mortality. Continuous variables were compared using the Wilcoxon rank sum test; dichotomous variables were compared using Fisher exact test. Main results: The low-intensity and high-intensity MICU groups were similar in age, gender, weight, and admitting diagnosis (P > .27 in all cases). Acute Physiology and Chronic Health Evaluation II (APACHE II) scores were 22.0 in the low-intensity era and 20.0 in the high-intensity era (P = .048). Median duration of MV was significantly lower in the high-intensity MICU format compared to the low-intensity MICU format (102 vs 61 hours, P for log-rank test = .0052). After controlling for covariates, there were 4.2 more VFDs in the high-intensity era (95% CI 1.9 to 6.6 days). The high-intensity era was associated with a reduced hospital mortality rate (27% vs 40%) and an adjusted odds ratio of 0.34 (95% CI 0.15 to 0.74). Conclusions: For critically ill veterans admitted to an MICU requiring MV, a high-intensity ICU structure is associated with more favorable mechanical ventilatory outcomes and lower mortality.

Does Research Training During Residency Promote Scholarship and Influence Career Choice? A Cross-Sectional Analysis of a 10-Year Cohort of the UCSF–PRIME Internal Medicine Residency Program

ABSTRACT Problem: The Association of Program Directors in Internal Medicine, the Accreditation Council for Graduate Medical Education, the Alliance for Academic Internal Medicine, and the Carnegie Foundation report on medical education recommend creating individualized learning pathways during medical training so that learners can experience broader professional roles beyond patient care. Little data exist to support the success of these specialized pathways in graduate medical education. Intervention: We present the 10-year experience of the Primary Care Medicine Education (PRIME) track, a clinical-outcomes research pathway for internal medicine residents at the University of California San Francisco (UCSF). We hypothesized that participation in an individualized learning track, PRIME, would lead to a greater likelihood of publishing research from residency and accessing adequate career mentorship and would be influential on subsequent alumni careers. Context: We performed a cross-sectional survey of internal medicine residency alumni from UCSF who graduated in 2001 through 2010. We compared responses of PRIME and non-PRIME categorical alumni. We used Pearsons chi-square and Students t test to compare PRIME and non-PRIME alumni on categorical and continuous variables. Outcome: Sixty-six percent (211/319) of alumni responded to the survey. A higher percentage of PRIME alumni published residency research projects compared to non-PRIME alumni (64% vs. 40%; p = .002). The number of PRIME alumni identifying research as their primary career role was not significantly different from non-PRIME internal medicine residency graduates (35% of PRIME vs. 29% non-PRIME). Process measures that could explain these findings include adequate access to mentors (M 4.4 for PRIME vs. 3.6 for non-PRIME alumni, p < .001, on a 5-point Likert scale) and agreeing that mentoring relationships affected career choice (M 4.2 for PRIME vs. 3.7 for categorical alumni, p = .001). Finally, 63% of PRIME alumni agreed that their research experience during residency influenced their subsequent career choice versus 46% of non-PRIME alumni (p = .023). Lessons Learned: Our results support the concept that providing residents with an individualized learning pathway focusing on clinical outcomes research during residency enables them to successfully publish manuscripts and access mentorship, and may influence subsequent career choice. Implementation of individualized residency program tracks that nurture academic interests along with clinical skills can support career development within medicine residency programs.

The Horrible Taste of Nectar and Honey—Inappropriate Use of Thickened Liquids in Dementia: A Teachable Moment

Story From the Front Lines A woman in her 90s with advanced dementia was admitted for stage IV pressure ulcers. She lived at home with her son, who was her primary caregiver and surrogate decision maker. After a hip fracture 1 year prior, she became completely dependent on her son. Because of her dementia, she was started on honey-thickened liquids for aspiration prevention. During the first week of hospitalization, she refused nearly all food and thickened liquids, turning her head when nurses attempted one-on-one feeding. By the end of the week, only her son was able to feed her small amounts of familiar foods. In consultation with the palliative care service, the primary team and son transitioned the patient to inpatient hospice. Her diet was liberalized to regular liquids and her favorite foods. Her intake improved. Two months later, she had gained 8 kg without enduring a clinically significant aspiration event.

Introducing Exercises in Clinical Reasoning

Clinical reasoning is a quintessential skill of the internist, the cornerstone of safe, effective medical care. While new technologies and accumulating evidence continue to change practice, they supplement rather than supplant clinical judgment in the care of the individual patient. Multiple studies show that transparency and instruction in clinical reasoning are what trainees seek most from their attending physicians. And while we know good reasoning when we see it, we often have difficulty describing it in specific terms. With this issue, JGIM introduces readers to the lexicon of clinical reasoning with a new feature called Exercises in Clinical Reasoning. These exercises build on the successful format of other series that intersperse iterative case presentations with extemporaneous expert clinician commentary. We allow the reader to follow along as an experienced clinician approaches an unknown, challenging patient care dilemma. But we add another layer of commentary that peers into the clinician’s mind (as much as we can) to highlight the normal, insightful, and sometimes errant judgment displayed in attempting to solve each case. We hope that this running diagnostic reasoning commentary and concluding discussion will help our readers become more facile with the science of clinical reasoning, enabling them to become more cognizant of their own judgments and more effectively teach reasoning skills to others. In this issue, “Doing What Comes Naturally” begins the Exercises with a relatively straightforward case involving a common clinical problem (hypertension) to introduce basic clinical reasoning concepts. In upcoming issues we hope to present cases with more twists, turns, blind alleys, or errors that highlight the more nuanced aspects of clinical reasoning. We invite clinicians to submit their interesting cases and to provide any suggestions or critiques on this new series.

A Curriculum for Diagnostic Reasoning: JGIM’s Exercises in Clinical Reasoning

University ofCalifornia, San Francisco, San Francisco,CA, USA; Veterans AffairsMedicalCenter, San Francisco, San Francisco,CA, USA; University of Washington, Seattle, WA, USA; University of Colorado School of Medicine, Aurora, CO, USA; Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH, USA; Johns Hopkins University School of Medicine, Baltimore, MD, USA.