By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Thyroid papillary microcarcinoma (TPMC) is an incidentally discovered papillary carcinoma that measures ≤1.0 cm in size. Most TPMCs are indolent, whereas some behave aggressively. The objective of the study was to evaluate whether the combination of v-raf murine sarcoma viral oncogene homolog B1 (BRAF) mutation and specific histopathologic features allows risk stratification of TPMC.
A group aggressive TPMCs was selected based on the presence of lymph node metastasis or tumor recurrence. Another group of nonaggressive tumors included TPMCs matched with the first group for age, sex, and tumor size, but with no extrathyroid spread. A molecular analysis was performed, and histologic slides were scored for multiple histopathologic criteria. A separate validation cohort of 40 TPMCs was evaluated.
BRAF mutations were detected in 77% of aggressive TPMCs and in 32% of nonaggressive tumors (P = .001). Several histopathologic features differed significantly between the groups. By using multivariate regression analysis, a molecular-pathologic (MP) score was developed that included BRAF status and 3 histopathologic features: superficial tumor location, intraglandular tumor spread/multifocality, and tumor fibrosis. By adding the histologic criteria to BRAF status, sensitivity was increased from 77% to 96%, and specificity was increased from 68% to 80%. In the independent validation cohort, the MP score stratified tumors into low-risk, moderate-risk, and high-risk groups with the probability of lymph node metastases or tumor recurrence in 0%, 20%, and 60% of patients, respectively.
Thyroid papillary microcarcinoma (TPMC) is a type of thyroid cancer defined by the World Health Organization Histological Classification of Tumors by 2 criteria: 1) ≤1 cm in size and 2) an incidental finding on histopathologic examination after lobectomy or total thyroidectomy performed for a larger adenoma or nodular hyperplasia.1 TPMCs are very common. In autopsy series, they are identified in 3% to 9% of thyroid glands in many regions of North America, Europe, and South America,2-9 and they appear with even higher incidence in systematically performed thyroid gland examinations in Japan (28%) and Finland (36%).10, 11 In surgical series, TPMC is reported in 5% to 17% of patients who undergo thyroidectomy for benign lesions.12-14 Its prevalence continues to rise. In fact, TPMC represents the fastest growing type of thyroid cancer.15 On the basis of Surveillance, Epidemiology, and End Results (SEER) data, papillary carcinoma ≤1 cm constituted about 30% of all papillary carcinoma in 1988 and about 40% in 2003,16 making it the most common variant of papillary carcinoma in the United States. Similar trends have been observed in France17 and in many other countries around the world (for review, see Roti et al18).
The clinical management of patients with TPMC remains nonstandardized. These tumors generally are considered clinically innocuous, although some have an aggressive clinical behavior. A meta-analysis encompassing >4000 papillary carcinomas ≤1 cm indicated that 28% of tumors had lymph node metastasis, 0.6% of tumors had distant metastasis, 3.3% of patients experienced disease recurrence, and tumor-related mortality was 0.3%.19 Those results suggest that, as an entity, TPMC includes at least 2 biologically distinct subpopulations: 1) indolent tumors with minimal or no potential for progression and 2) tumors with the propensity for aggressive behavior and dissemination. The ability to stratify those relatively few patients with aggressive TPMC from the vast majority who are low risk is crucial to offer the most appropriate clinical management.
The risk factors associated with aggressive behavior of TPMCs are not well defined. One of the more consistently reported features associated with the risk of tumor recurrence or metastasis is tumor multifocality or bilaterality.20-22 However, tumor multifocality is common and is present in 30% to 40% of all TPMCs.19, 23 Therefore, multifocality cannot be used by itself as an accurate marker of tumor virulence. Studies have demonstrated that extensive fibrosis in TPMC is associated with more frequent lymph node or distant metastases.22, 24 In some studies, the metastatic spread of TPMC was limited to tumors >8 mm18 or ≥5 mm21 in size, although other observers have indicated that tumor size was not correlated with aggressive disease.25 One study indicated that familial TPMCs were more aggressive than sporadic tumors,26 although other reports did not confirm that observation.18, 22 A recent meta-analysis of 17 studies revealed that recurrence of TPMC was associated with younger age (<45 years), tumor multifocality, and lymph node metastasis at presentation; whereas no association with sex, tumor size, or extrathyroid extension was identified.27
More recently, genetic markers have been explored to assess tumor behavior in papillary thyroid carcinoma, including TPMC. Many studies have demonstrated the association between the v-raf murine sarcoma viral oncogene homolog B1 (BRAF) valine-to-glutamic acid mutation at codon 600 (V600E) and aggressive histopathologic features of papillary carcinoma, risk of tumor recurrence, and cancer-related death (for review, see Xing28). Several recent studies of TPMC described a correlation between the BRAF V600E mutation and extrathyroid extension,29-32 advanced stage at presentation,30-32 lymph node metastasis,31, 32 tumor size >5 mm,30 and multifocality.32 However, in those studies, a BRAF mutation was present in 24% to 63% of TPMCs,29-32 and it is unlikely that such a large proportion of these tumors would have aggressive behavior. Therefore, the presence of a BRAF mutation in any given tumor is not an absolute predictor of tumor aggressiveness.
The objective of the current study was to evaluate the role of BRAF mutational status together with several specific histopathologic tumor characteristics in the identification of a TPMC subset that demonstrates more aggressive behavior, such as extrathyroid spread. We also propose a relatively simple molecular-pathologic score for risk stratification and report on a test of its performance in an independent validation cohort of TPMCs.
MATERIALS AND METHODS
Selection Criteria and Cohort Assignment
With the approval of the University of Pittsburgh Institutional Review Board, deidentified data from 2377 patients who were treated at the hospitals of our university system from 1991 to 2004 with an inpatient or outpatient International Classification of Diseases, Ninth Edition, Clinical Modification diagnosis code of 193 (thyroid malignancy) were retrieved. Among them, 888 patients had surgical pathology reports available electronically to confirm both tumor diagnosis and tumor size. Of those, 403 patients (45%) had papillary cancer ≤1.0 cm without evidence of other thyroid malignancy and were used to form the original cohort for the study.
Of 403 patients with confirmed TPMC, 41 patients (10%) had cervical lymph node metastasis identified during initial pathologic examination or during subsequent clinical follow-up. These patients were matched for age, sex, and tumor size to 41 patients with TPMC who did not have either local metastasis or tumor recurrence. Patients from both groups who could not be studied further because of unavailability of paraffin blocks or disappearance of tumor tissue in additional sections taken for molecular analysis were excluded, leaving 29 patients in the group with clinically aggressive TPMCs, subsequently referred to as Group A of the original cohort. After similar exclusions, 30 of the original 41 patients who were selected as matched controls without aggressive features remained and are referred to subsequently to as Group B. Groups A and B comprise the original cohort used in this analysis. There was no significant difference in the extent of initial surgery (total thyroidectomy vs lobectomy) between the 2 groups, although completion of thyroidectomy was more common in Group A (52% vs 17%). Because none of the tumors in either group were diagnosed preoperatively, they represented true TPMC defined according to strict World Health Organization criteria.
In Group A, 27 of 29 patients had lymph node metastasis detected at surgery. The follow-up for patients in this cohort ranged from 0 to 11.6 years (mean follow-up, 5.3 years). Six patients had a recurrence in the regional lymph nodes that was diagnosed on average 4.3 years after initial surgery. Among these 6 patients, 4 had lymph node metastases detected during initial surgery, and 2 had no positive lymph nodes at presentation. In Group B, patients were followed for 0 to 16.9 years (mean follow-up, 4.8 years). During surveillance, none had evidence of recurrence based on serum thyroglobulin measurements, neck ultrasonography, and clinical examination.
A validation cohort (Group C) was selected from cases retrieved from the year 2008 to avoid overlap with Groups A and B. The validation cohort consisted of 40 patients with TPMC, including 39 consecutive patients and 1 additional patient with aggressive TPMC who was known to the authors and who developed a thyroid bed soft tissue recurrence 14 months after surgery. The follow-up in this cohort ranged from 0 to 2.8 years (mean follow-up, 1.5 years). Four patients had local lymph node metastasis detected at surgery, and 1 patient was diagnosed with pulmonary metastases 2.8 years after initial surgery.
For all patients, surgical pathology reports were examined, and histologic slides were reviewed independently by 2 pathologists (L.A.N. and Y.E.N.) to evaluate and score the following microscopic features: anatomic location of TPMC (left lobe, right lobe, or isthmus), tumor size, tumor location with respect to thyroid surface/capsule (superficial vs intrathyroid, as described below), status of surgical margins, presence of infiltrative tumor border (as opposed to smooth pushing border), tumor growth pattern (classic papillary, follicular, tall cell), intraglandular spread or tumor multifocality (IGS/MF), extrathyroid extension, degree of fibrosis (as defined below), presence of psammoma bodies, and presence of significant lymphocytic thyroiditis in non-neoplastic thyroid tissue consistent with Hashimoto thyroiditis. The distance from the edge of tumor to the closest margin of resection was measured.
Superficial/subcapsular tumor location was defined as location immediately adjacent to perithyroid adipose tissue, ie, with no benign thyroid tissue between the tumor and the extrathyroid soft tissues. IGS/MF was defined as the presence of either 1) ≥2 separate tumors, or 2) smaller tumor aggregates located close to the main tumor separated by a layer of normal thyroid parenchyma, or 3) tumor cells within lymphatic channels, or 4) isolated psammoma bodies located in thyroid stroma outside the tumor. Tumor fibrosis was scored as none, 1+, and 2+ and was defined as follows: 1+, fibrosis (mild fibrosis with the presence of few inconspicuous, delicate fibrous areas within or at the periphery of the tumor nodule); and 2+, fibrosis (moderate/extensive fibrosis that was clearly recognizable, with multiple fibrotic bands within and at the periphery of the tumor). Fibrotic tumor capsule alone, ie, without significant fibrosis within the tumor, was not sufficient to score tumor fibrosis as 2+.
For each tumor case, 5 unstained tumor slides were obtained and microdissected to collect tumor tissue. Molecular analysis was performed to test for BRAF V600E mutation and for the 3 most common RAS mutations, including those at neuroblastoma RAS viral oncogene homolog (NRAS) codon 61, v-Ha-ras Harvey rat sarcoma viral oncogene homolog (HRAS) codon 61, and v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) codons 12/13. Detection of these mutations was performed using real-time polymerase chain reaction (PCR) and fluorescence melting curve analysis (FMCA) from DNA as reported previously.33, 34 Briefly, a pair of oligonucleotide primers flanking the mutation site was designed together with 2 fluorescent probes. The reaction mixture was subjected to 45 cycles of rapid PCR amplification. Postamplification FMCA was performed by gradual heating of samples at a rate of 0.2°C per second from 45°C to 95°C. All PCR products that deviated from the wild-type (placental DNA) melting peak were sequenced to verify the presence of mutation.
The chi-square test was used to estimate difference in accuracy between different approaches. The difference between the frequency of each feature in the 2 study cohorts was compared using a univariate logistic regression model for continuous variables and the Fisher exact test for categorical variables; P values < .05 were considered statistically significant without multiple comparison correction.
Multivariate analysis and logistic regression were used to model the data, and outcomes were analyzed by study cohort, ie, by defined factors of clinical aggression. Whether a tumor was assigned to the aggressive group or the nonaggressive group was considered the outcome. All evaluated factors (eg, BRAF, tumor location) were considered predictors. The statistical package R was used to perform statistical computing (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org/; Accessed April 1, 2011). When building the multivariate regression model, patient age, sex, and tumor size were excluded from the analysis, because the groups were built initially by matching these parameters. Step-wise variable selection in the regression model was performed manually. Parameters were selected based on: 1) predictability of a feature determined by its P value in the univariate model (only variables with P values < .01 were selected as candidates for the multivariate model), 2) overall performance of the multivariate model, and 3) the P value of each parameter estimate in the multivariate model.
The histopathologic and molecular features of the 29 patients with TPMC in Group A (tumors with aggressive features) and the 30 control patients in Group B (tumors without aggressive features) are summarized in Table 1.
Table 1. Histopathologic and Molecular Features of Thyroid Papillary Microcarcinomas in 2 Groups of the Original Cohort
Seven histopathologic features differed significantly between the groups. These features included superficial (subcapsular) tumor location, extrathyroid extension, infiltrative border, follicular growth pattern morphology, tall cell features, tumor fibrosis, and IGS/MF. The combined category of IGS/MF differed between the 2 cohorts with a high degree of statistical significance, although the frequency of multifocality alone did not differ between the groups, probably because of the criteria applied to distinguish between tumor multifocality and intraglandular spread.
Molecular analysis was informative in 26 aggressive TPMCs (Group A) and 25 nonaggressive TPMCs (Group B) and was uninformative for 8 tumors, which measured < 3 mm and yielded insufficient amounts of DNA for testing. BRAF V600E mutations were detected in 20 aggressive TPMCs (77%) compared with 8 nonaggressive TPMCs (32%; P = .001) (Fig. 1). RAS mutations were present in 2 TPMCs from Group B (1 NRAS codon 61 mutation and 1 HRAS codon 61 mutation) and were not observed in any aggressive tumors.
Stepwise regression analysis was used to identify the smallest set of parameters that provided the best separation between the 2 cohorts (Table 2). The set included 4 parameters: BRAF mutation status, tumor location (superficial), significant (2+) fibrosis, and IGS/MF (Figs. 2-4). The inclusion of additional evaluated histopathologic parameters did not improve the separation between the cohorts. Table 2 indicates that each of the 4 predictive parameters, when positive, contributed differently to separation of the 2 study cohorts. For practical use, the coefficients were rounded to the nearest integer for each included factor. The resulting weighted molecular-pathologic (MPW) scoring algorithm was as follows: MPW score = 4 × superficial tumor location + 3 × BRAF(+) + 3 × IGS/MF(+) + 2 × fibrosis(2+).
Table 2. The Multivariate Regression Model Parameters, Coefficients Estimates, and Corresponding P Values
Abbreviations: BRAF, v-raf murine sarcoma viral oncogene homolog B1; SE, standard error.
This P value was statistically significant on multivariate analysis.
Scores from 0 to 7 and scores from 8 to 12 correctly identified nonaggressive and aggressive TPMCs, respectively, with 96% sensitivity and 80% specificity. Scores from 0 to 7 reflected the presence of 0 to 2 of the 4 features evaluated. Score of 8 to 12 corresponded to the presence of 3 to 4 features. This allowed a simplified, unweighted version of the molecular-pathologic (MPU) score using a sum of features: MPU score = superficial tumor location + BRAF(+) + IGS/MF(+) + fibrosis(2+). The simpler MPU score achieved sensitivity (96%) and specificity (80%) identical to the MPW score.
The combined MP score provided significantly better accuracy for the detection of TPMCs with aggressive features than BRAF status alone (P = .04). The sensitivity for identifying tumors with aggressive features was 77% based on the presence of BRAF V600E but increased to 96% when the full MP model was applied. Similarly, specificity increased in the original cohort from 68% when BRAF only was considered to 80% when tumors were evaluated with the combined MP score.
The performance of the MP score was validated in an independent set of 40 TPMCs. Among tumors in this validation cohort, 5 tumors had aggressive features: four had lymph node metastasis at presentation, including 1 patient who also developed pulmonary metastases 34 months later, and 1 recurred in the thyroid bed 14 months after surgery. An MPW score of 8 to 12 was able to detect aggressive tumors with 100% sensitivity and 71% specificity. Identical sensitivity and specificity were obtained again when the simplified MPU score was applied. Two aggressive tumors in this set had no BRAF mutation, but both had aggressive histology and, thus, were identified correctly by the MP analysis. Specificity of the MP score in the validation cohort was lower than in the original cohort (71% vs 77%). This resulted from an increased number of tumors with higher MP scores that lacked overt clinically aggressive behavior or recurrence. This is likely because of the significantly shorter follow-up in the validation cohort compared with the original cohort. It is conceivable that some of the tumors with high MP scores that did not demonstrate clinically evident disease eventually would recur with longer follow-up, which already has happened in 1 patient.
The MP score distribution in the validation cohort of tumors suggested that scoring could be subdivided further to resolve 3 distinct risk groups, ie, low risk, moderate risk, and high risk. Using either version of the MP score, the 3-tiered probability of aggressive behavior, ie, lymph node metastasis or tumor recurrence, rose stepwise from 0% to 20%, and 60%, respectively (Table 3). The risk stratification remained quite accurate for predicting the risk of tumor recurrence alone. Indeed, among 7 tumors with documented anatomic recurrence on follow-up (6 from the original cohort and 1 from the validation cohort), 6 tumors had an MPW score of 12 and an equivalent MPU score of 4, falling into a high-risk group; and 1 tumor had an MPW score of 10 and an MPU score of 3, corresponding to a moderate-risk tier.
Table 3. Molecular-Pathologic Scores in the Validation Cohort of Thyroid Papillary Microcarcinomas and the Risk of More Aggressive Tumor Behaviora
The MP score included v-raf murine sarcoma viral oncogene homolog B1 BRAF status and 3 histopathologic features (superficial tumor location, intraglandular tumor spread/multifocality, and tumor fibrosis).
Probability of extrathyroid spread or recurrence, %
The validation cohort allowed an estimation of the proportion of TPMCs that would fall into each risk group based on the MP score. On the basis of this cohort, 64% of tumors would be placed in the low-risk category (MPU score, 0-2), 26% would be placed in the intermediate-risk category (MPU score, 3), and 10% would be placed in the high-risk category (MPU score, 4).
In this study, we identify several molecular and histopathologic features that correlate with more aggressive behavior of TPMC and offer a practical and simple scoring system to assess the clinical behavior of this common type of thyroid cancer. The scoring system relies on BRAF mutation status and 3 histopathologic features to assign tumors to 1 of 3 risk categories. Neither mutational status for BRAF V600E nor histologic features alone were adequate to provide accurate risk stratification.
The BRAF V600E mutation has been correlated in many studies with more aggressive behavior of thyroid papillary carcinoma.28 Although the reported data on TPMCs are more limited, this mutation has been correlated with more aggressive histopathologic characteristics of these tumors in several recent studies.29-32 In the current study, we observed that the BRAF V600E mutation was highly prevalent in TPMCs with metastatic spread to lymph nodes or tumor recurrence compared with a matched control cohort of nonrecurrent TPMCs without metastatic spread. The difference between the 2 cohorts was significant on univariate and multivariate analyses. Consequently, BRAF mutation status was included as an important component of the combined MP scoring algorithm. However, BRAF mutation status alone was not sufficient to reliably identify all TPMCs with aggressive characteristics. Specifically, among aggressive tumors in the original and validation cohorts, 8 tumors had no BRAF mutation, but all had aggressive histology and would not have been identified by molecular analysis alone. Conversely, the histopathologic features alone were not adequate to identify all tumors that demonstrated aggressive clinical features. In the original cohort, among tumors with aggressive clinical features, 20 tumors were BRAF positive. However, only 13 of these tumors also demonstrated all 3 aggressive histologic features that would be sufficient to render an aggressive MP score. The remaining 7 tumors had only 2 aggressive histologic features and would have been missed without molecular assessment for BRAF mutation.
Extrathyroid tumor extension, multifocality, and significant tumor fibrosis have been associated with aggressive behavior of TPMCs in previous studies.20-22, 24 Our series confirm these findings. However, we observed not only that extrathyroid extension was associated with aggressive features of TPMCs but also that the simple presence of tumor located superficially at the surface of the thyroid gland had a strong correlation with the risk of tumor spread to regional lymph nodes and/or recurrence. Tumor location at the thyroid surface obviously predisposes to extrathyroid tumor extension, which was observed in 75% of these tumors. It is likely that, even in the absence of detectable extrathyroid extension, superficial tumor location may facilitate tumor spread into lymphatic channels and regional lymph nodes, explaining the increased risk associated with this parameter.
Significant tumor fibrosis also was a predictor of risk in this study. We previously observed the association between significant, sclerotic type fibrosis in TPMC and lymph node or distant metastasis.24 This association also has been observed by others22 and was confirmed in the current study. The biologic basis for this finding requires further investigation. However, tumor desmoplasia is a well known feature of invasive neoplastic growth, and it is plausible that the increased aggressiveness of these TPMCs may be related to their propensity for matrix formation and angiogenesis.
We treated intraglandular tumor spread and tumor multifocality as a combined feature (IGS/MF) because of the inherently subjective manner in which, in some patients, 2 tumors are designated as synchronous primaries or as a tumor with intraglandular spread. The combined parameter, IGS/MF, retained retained statistical significant on multivariate analysis and allowed its use as a reproducible predictor of risk in this study.
Tall cell morphology also has been reported as a risk factor in TPMC but was not part of our final MP scoring algorithm, because it did not improve its performance further. In the original cohort, 9 tumors had tall cell features, and 8 of those were from the aggressive tumor cohort. All 8 of these tumors were positive for BRAF mutation and, thus, were included indirectly in our score.
The MP score can be calculated using the same 4 parameters that are weighted differently (MPW score) or in a simplified, unweighted way (MPU score). It is conceivable that, in a very large data set, the MPW score may provide additional prognostic information. For example, the intermediate-risk category, which corresponds to an MPU score of 3, encompasses tumors with an MPW score in the range of 8 to 10, and it remains unknown whether or not an incrementally higher score within this category would confer a progressively higher risk.
The proposed scoring system assigns 65% of all TPMCs to the low-risk category, ie, tumors with virtually no risk of aggressive behavior; assigns 25% of all TPMCs to the intermediate-risk category; and assigns only 10% of all TPMCs to the high-risk category. Therefore, our model is consistent with the expected biologic behavior of TPMC (ie, the majority are indolent, nonaggressive tumors) and allows clinically relevant risk stratification that can further inform treatment decision-making. The revised American Thyroid Association management guidelines for patients with thyroid nodules released in 2009 recommend against completion thyroidectomy and subsequent radioiodine ablation for tumors <1 cm when tumors are unifocal, intrathyroid, lymph node-negative, and lack other high-risk features, such as certain histologic subtypes (tall cell variant, poorly differentiated carcinoma).35 In our validation cohort of 39 consecutive TPMCs, 19 tumors (48%) were either multifocal or had extrathyroid extension or tall cell morphology. Consequently, almost 50% of TPMCs would have been candidates for more aggressive treatment based on the current American Thyroid Association guidelines. On the basis of our analysis, only 10% of these TPMCs were at notably high risk for recurrence, and that figure increased to only 35% if the TPMCs of indeterminate risk were included. In addition, our model provides a means of assigning proportional risk based on the number of positive features present.
How exactly the proposed scoring system should be applied to the management of patients with these tumors requires further investigation. It seems reasonable to suggest that patients with low-risk tumors should be followed conservatively, and patients with high-risk tumors probably would benefit from total thyroidectomy, possibly with central compartment or lateral neck lymph node sampling. It is likely that patients who have moderate-risk and high-risk TPMC would benefit from more intensive postoperative follow-up surveillance. The role of radioactive iodine ablation in the treatment of patients with high-risk TPMCs remain unclear.
This study has several limitations that must be considered. The aggressiveness of TPMC in both tumor sets was defined primarily based on lymph node metastasis identified at presentation. Therefore, the presence of malignant adenopathy remained the primary outcome measure that defined aggressive tumor behavior in this study. Few patients demonstrated tumor recurrence in either cervical lymph nodes or soft tissues of the thyroid bed. Distant metastasis and tumor-related death are exceedingly rare in TPMC and were not observed in any patients in the current series. However, it is important to note that lymph node involvement at presentation has a strong correlation with recurrence of TPMC27; therefore, we believe that it can be used reliably to identify a given tumor as more aggressive. Additional validation will be required to determine how well our model can predict the discrete outcomes of tumor recurrence and mortality associated with TPMC.
In summary, in this report, we describe several clinical and histopathologic characteristics that correlate with the risk of extrathyroid spread and recurrence of TPMC, and we offer a simple MP score that can be used to assess tumor aggressiveness. The combined MP score allows for a more robust risk stratification than either molecular or histopathologic evaluation alone and may be helpful in the clinical management of patients with this common type of thyroid tumor.
This study was supported in part by a grant from the National Institutes of Health (grant R01 CA88041) and by the Carrie L. Hughes Endocrine Genetics Research Fund of the University of Pittsburgh School of Medicine.