ARTICLE IN PRESS

Pathology – Research and Practice 204 (2008) 867–873

www.elsevier.de/prp

ORIGINAL ARTICLE
Clinico-pathological discrepancies in the diagnoses of solid malignancies
Lea Tenenholz Grinberga,b,, Luiz Fernando Ferraz da Silvaa,
Ana Carolina Galtarossa Xavierb, Paulo Hilário Nascimento Saldivaa, Thais Mauada
a
Department of Pathology, Sao Paulo University Medical School, Sao Paulo, Brazil
b
Albert Einstein Research and Education Institute, Albert Einstein Hospital, São Paulo, Brazil

Received 24 September 2007; accepted 11 July 2008

Abstract
Autopsy is a valuable tool in evaluating diagnostic accuracy. Solid malignancies may have a protracted presentation,
and diagnosis frequently requires imaging and deep-sited biopsies; clinical and postmortem diagnosis discrepancies
may occur in a high rate in these diseases. Here, we analyzed the occurrence of clinico-pathological discrepancies in the
diagnoses of solid malignancies in a Brazilian academic hospital. We reviewed charts and autopsy reports of the
patients that died from 2001 to 2003 with at least one solid neoplasm. Patients were classified in concordant and
discordant cases regarding cancer diagnosis. Discordant cases were categorized in undiagnosed cases (no suspicion of
cancer) and in misdiagnosed cases (clinical suspicion of cancer but incompletely diagnosed). Among the 264 patients
with a single non-incidental solid neoplasm, the clinico-pathological discrepancy rate was 37.1%. Liver (22.5%), lung
(19.4%), and pancreatic cancer (15.3%) were the most frequent malignancies in the discordant group. Misdiagnosis
category comprised 68% of the discordant cases, i.e., there was no correct knowledge about the tumor primary site
and/or the histological type during life. Our data show that a high rate of discrepancies occurs in solid malignancies.
Autopsies may provide the basis for a better understanding of diagnostic deficiencies in different circumstances.
r 2008 Elsevier GmbH. All rights reserved.

Keywords: Autopsy; Discrepancy; Cancer; Diagnoses and examinations; Diagnostic errors

Introduction approximately 6% [5]. This decline has been attributed

Many studies have validated autopsy as an indis-
pensable tool in verifying clinical diagnosis and quality
of health care [11,20,21]. Despite this, the number of
autopsies decreased significantly in the last decades
worldwide. For example, in the USA, the autopsy rates
performed in adults have decreased from 50% in the
1950s to 10–20% in the 1980s [7,15] to currently
Corresponding author at: Department of Pathology, Faculdade de
Medicina da Universidade de São Paulo, Avenida Dr. Arnaldo, 455
1st floor, Room 1351 – Zipcode 01246-903, São Paulo, SP, Brazil.
Tel.: +55 11 30618249; fax: +55 11 30642744.
E-mail address: leagrinberg@usp.br (L.T. Grinberg).
to an increase in invasive diagnostic techniques, better
accuracy in imaging techniques prior to death, a
decrease in consent rates from relatives, litigation due
to medical errors, and decreased reimbursement for such
procedures [6,10,17].
Several published studies comparing pre- and post-
mortem diagnoses have shown that in 0–29% of cases,
patient management would have been different if the
diagnosis had been made prior to death [21]. However,
the frequency of important missed diagnoses has been
shown to be higher among patients with severe and/or
rare disorders [18]. Patients presenting solid neoplasms
quite often present protracted clinical symptoms
and may require multiple diagnostic steps to complete

0344-0338/$ - see front matter r 2008 Elsevier GmbH. All rights reserved.
doi:10.1016/j.prp.2008.07.001
 ARTICLE IN PRESS
868 L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873
work-up and confirm the diagnosis. As a consequence of
that, it is possible that the occurrence of missed
diagnosis is still high among patients who died with
solid neoplasms.
Indeed, several published studies have addressed the
role of autopsies in detecting pre- and postmortem
discrepant diagnosis of neoplasms, with discrepancy
rates varying from 16.6% to 50% [2,8,9,12,24]. How-
ever, few studies have specifically analyzed clinico-
pathological discrepancy rates of solid neoplasms in
developing countries, where infectious diseases are
usually more prevalent and are likely to be first
considered and addressed during the care of the patient,
and where socio-economic limitations may increase the
risk of the patient to have a misdiagnosis [19].
In this study, we estimated the frequency of clinico-
pathological diagnostic discrepancies among patients
with solid neoplasms in a teaching hospital in Sao
Paulo, Brazil.
Materials and methods
Study population
This study was performed in the Department of
Pathology at the University of Sao Paulo Medical
School, which is linked to Hospital das Clinicas (HC).
HC is the largest tertiary care University Hospital
in Brazil. Services include multidisciplinary inpatient
and outpatient care from basic health care to high-
complexity procedures, such as transplantations. The
HC had an average annual autopsy rate of 50% in the
last 5 years of the study period. The study protocol
was approved by the Internal Review Board of the
Hospital.
We reviewed the autopsy reports performed between
January 2001 and December 2003 from patients who
died at least 48 h after admission to the hospital,
excluding those cases with insufficient time for proper
initial evaluation. All cases in which at least one
malignant solid neoplasm was found at autopsy were
included. Central nervous system and hematological
neoplasms were excluded from the analysis.
Of those cases included, information regarding the
primary site and histological subtype of the tumor was
extracted from the autopsy records. At least one
pathologist reviewed the microscopic findings for each
case using hematoxylin–eosin stain. Special stains and
immunohistochemistry were used as needed. Patient
age, gender, and information regarding primary site and
histopathological subtype were collected from the
medical charts if the diagnosis of the malignancy had
been made or suspected before death. Unavailable,
incomplete, or unreadable medical records were ex-
cluded from analysis.
Data collected from both autopsy records and
medical charts were compared, and cases were subse-
quently grouped in two main categories: concordant, if
the malignant tumor had been histologically diagnosed
with the correct primary site and histological type prior
to death and had been confirmed by autopsy; or
discordant, if the malignant tumor detected at autopsy
had not been fully diagnosed or suspected prior to
death. Classification of concordant vs. discordant
categories was performed by two pathologists and the
discrepancies were discussed for consensus.
Discrepancy analysis
The discordant diagnoses were further divided in two
categories as follows: (1) no clinical suspicion of
neoplasm prior to death (undiagnosed cancer); and
(2) clinical suspicion of a neoplasm, but the primary
site was incorrect or unknown or the primary site
was correct or not, but the histological type of the
neoplasm was not defined (misdiagnosed cancer). For
instance, if a patient is suspected of gastric cancer but
biopsy was not obtained pre-mortem and the diagnosis
was confirmed by autopsy, the case was classified as
discrepant, because the treatment is dependent on the
histological type. The primary site was determined by
the autopsy macroscopic finding and histological
evaluation.
For discordant cases, information regarding length of
the inpatient care and unit of care (emergency room,
intensive care unit, medical or surgical ward) was also
extracted from the clinical charts. In addition, we
determined the presence of metastases at autopsy. For
each discrepant category, we estimated the proportion
of patients and ranked the most common malignant
neoplasms.
Non-metastatic small (o1 cm) neoplasms detected at
autopsy and not related to medical symptoms or to the
death of the patient were considered as incidental and
were excluded from discrepancy analysis [22]. Cases
presenting multiple primary neoplasms at autopsy were
also excluded from the analysis.
Statistical analysis
The Statistical Package for the Social Sciences (SPSS,
Chicago, IL, USA) was employed in the statistical
processing of data. Patient age was expressed as median
and interquartiles. For statistical analysis, concordant
and discordant categories were compared for age at death
and gender in univariate analysis. Discrepant subcate-
gories (un- and misdiagnosis) were further compared for
the presence of metastasis, length of inpatient care, unit
of care, and whether the immediate death cause was
directly related to the neoplasm in univariate analysis.
 ARTICLE IN PRESS
L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873 869

The Chi-square and the Mann–Whitney tests were
used to compare the variables. We used multivariate
logistic regression in these parameters to assess the
occurrence of an independently associated factor related
to the occurrence of misdiagnosis. Two-sided P values of
less than 0.05 were considered to indicate statistical
significance.
Results
264 cases, 166 (62.9%) were classified as concordant and
98 (37.1%) were considered as discordant.
In the concordant group, there were 98 males (59%):
their median age was 61.5 (23) years; in the discordant
group, there were 56 males (57%): their median age was
61.5 (21) years (Fig. 1). No statistically significant
differences between concordant and discordant cate-
gories regarding age (p ¼ 0.687) and gender (p ¼ 0.763)
were found.
The most common neoplasms in the concordant and
discrepant groups are presented in Tables 1 and 2.

Population
During the 3-year period covered by the present
study, 3531 autopsies of natural deaths were performed
in our institution, corresponding to 47.5% of all
hospital deaths occurring in that period.
There were 289 patients hospitalized for more than
48 h who presented with at least one solid neoplasm
detected at autopsy; however, two cases were further
excluded of the analysis, because the complete medical
records were not available.
Two hundred and ninety-eight (298) malignant
neoplasms were identified after autopsy among
287 patients. Ten patients had two or more neoplasms
detected at autopsy and were excluded from analysis.
Out of 277 cases, 13 presented tumors considered
as incidental and were also excluded (Fig. 1). Out of
Discrepant cases
Among the 98 discordant cases, in 31 patients, the
neoplasms were not suspected prior to death (31.6%,
31/98), whereas in 67 (68.4%, 67/98), the diagnostic
work-up was incomplete (neoplasm site and/or neo-
plasm type).
Within the misdiagnosed cases, for 30 of 67 (44.7%)
patients, neither the primary site nor histological type
was known prior to death. In 11 out of 67 cases (16.4%),
the neoplasm was clinically and histologically detected,
but the site of the primary tumor was unknown or
mistaken. On the other hand, in 26 out of 67 cases
(38.8%), the site of the primary tumor was clinically
detected, but the histological type was only defined by
the autopsy (Fig. 2).

Neoplasm = 287 (7.8%)

(41.1%f)
62y (22)

Single neoplasm = 277 Two or more neoplasms = 10 (3.5%)

(96.5%) 35%f
41.5%f 69y (18)
62y (22)

Incidental neoplasm = 13 (4.5%) Non-incidental neoplasm = 264 (95.5%)

38.5% f 41.7 %f
62y (21) 61y (22)

Concordant cases= 166 (62.9%) Discrepant cases = 98 (37.1%)

41%f 42.9%f
61.5y (23) 61.5y (21)

Non-suspicion prior to death

= 31(31.6%) Suspicion prior to death
35.5%f = 67 (68.4%)
58y (21) 46.2 %f
62y (19)

Fig. 1. Distribution of the studied population within the diagnostic categories. The total number of autopsies was 3531. The first
line in each box indicates the number of patients, the second line gives the gender proportion, and the third line indicates the median
(interquartiles) age.
 ARTICLE IN PRESS
870 L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873

Table 1. Frequency distribution of patients with a concordant and discordant single non-incidental neoplasm (n ¼ 264) by site,
gender, and category

Neoplasm primary site Discordant, N ¼ 98 Concordant, N ¼ 166 Total, N ¼ 264

a
Liver, n (%) 22 (22.5) 23 (13.9) 45 (17)
Lung, n (%) 19 (19.4) 16 (9.6) 35 (13.3)
Pancreas, n (%) 15 (15.3) 6 (3.6) 21 (7.9)
Stomach (%) 7 (7.2) 13 (7.8) 20 (7.6)
Urinary bladder, n (%) 7 (7.2) 4 (2.4) 11 (4.2)
Prostate, n (%) 4 (4) 10 (6) 14 (5.3)
Colon, n (%) 3 (3) 24 (14.5) 27 (10)
Breast, n (%) 0 13 (7.8) 13 (4.9)
Upper respiratory tract, n (%) 0 22 (13.3) 22 (8.3)
Others, n (%) 21 (21.4) 35 (21.1) 56 (21.2)
a
% within each category.

Table 2. Frequency distribution of patients with a misdiagnosed and undiagnosed single non-incidental cancer (n ¼ 98) by site,
gender, and category

Neoplasm primary site Undiagnosed, N ¼ 31 Misdiagnosed, N ¼ 67 Total, N ¼ 98

Liver, n (%)a 11 (35.5) 11 (16.4) 22 (22.5)

Lung, n (%) 4 (12.9) 15 (22.4) 19 (19.4)
Pancreas, n (%) 3 (9.7) 12 (17.9) 15 (15.3)
Stomach (%) 1 (3.2) 6 (9) 7 (7.2)
Urinary bladder, n (%) 3 (9.7) 4 (6) 7 (7.2)
Prostate, n (%) 3 (9.7) 1 (1.4) 4 (4)
Colon, n (%) 1 (3.2) 2 (3) 3 (3)
Breast, n (%) 0 0 0
Upper respiratory tract, n (%) 0 0 0
Others, n (%) 5 (16.1) 16 (23.9) 21 (21.4)
a
% within each category.

Fig. 2. Distribution of patients (n ¼ 67) according to primary tumor site and histology within category 2 (misdiagnosis).

Discrepant cases subanalysis
Undiagnosed  misdiagnosed cases
Table 3 shows data related to age, gender, hospita-
lization length and unit of care, as well the presence of
metastasis.
Chi-square analysis showed that the frequency of
metastasis was higher in the misdiagnosed category
(p ¼ 0.002). Using multivariate logistic regression ana-
lysis with the categorized clinical parameters, age (o19;
19–59; 460 years), hospitalization length (o72 h, 72 h-
10 days, 410 days), gender, hospital ward, and the
presence of metastasis confirmed that the only statisti-
cally significant, independently associated factor related
to the occurrence of undiagnoses was the presence of
metastasis (p ¼ 0.017, odds ratio ¼ 0.34).
Additional findings
Interestingly, in the undiagnosed neoplasms category,
9/11 (81.8%) of the liver neoplasms were associated with
cirrhosis; and in two of four (50%) lung neoplasms, the
clinical suspicion was tuberculosis. Within the misdiagnosis

L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873
Table 3. Characteristics of the patients included in the discrepant subcategories
Category
No of patients
Women, n (%)
Age (year)a
Length of hospitalization (days)a
Metastasis, n (%)
Medical ward, ERb, n (%)
Medical ward, ICUc, n (%)
Medical ward, CSWd, n (%)
a
Values expressed as medians (interquartiles).
b
Emergency room (ER).
c
Intensive care unit (ICU).
d
Clinical/surgical wards (CSW).
category, 27.2% of the patients with liver neoplasms and
50% of the patients with lung neoplasms were waiting
for biopsy results.
Discussion
In this study, we reported a 37.1% discrepancy rate of
the clinico-pathological diagnoses related to solid
neoplasms in a large, tertiary care hospital in Brazil,
which had a high autopsy rate during the study period.
We have previously reported high rates of overall
clinico-pathological discrepancies in hematological neo-
plasms [26]. Here, we were specifically interested in the
solid neoplasms, because these neoplasms may have less
evident symptoms in the early phases and usually
require image techniques and deep-sited biopsies to be
definitively diagnosed, possibly indicating a disease
category where clinico-pathological discrepancies occur
at high rates.
The three most frequent sites of unsuspected or
incompletely misdiagnosed cancer neoplasm in our
study were the liver, lungs, and pancreas. These findings
are in line with other studies, as reviewed by Avgerinos
et al. [2]. As expected, neoplasms in more accessible
sites, such as in the breast and the upper respiratory
system, have the highest rates of a correct clinical
diagnosis [14].
A high frequency of discrepancies was found in
patients with primary liver neoplasms. Certainly, the
presence of coexisting cirrhosis was a confusing factor
for clinical diagnoses, because most of the liver
neoplasms in this series occurred in cirrhotic patients.
In fact, 50% of the liver neoplasms diagnosed at autopsy
in this study were unsuspected in life. It is noteworthy
that more than 80% of the hepatocellular carcinomas
(HCC) arise in developing countries [25], and a previous
local study showed that HCC in Brazil has intermediate
epidemiological characteristics concerning incidence,
ARTICLE IN PRESS
871
Undiagnosed (1) Misdiagnosed (2) P-value
31 67
11 (35.5) 28 (42.2) 0.661
58 (21) 62 (19) 0.354
5 (12) 9 (12) 0.930
10 (32.3) 44 (65.6) 0.002
15 (48.4) 28 (41.8)
8 (25.8) 24 (35.8) 0.617
8 (25.8) 15 (22.4)
which can explain the relatively high number of hepatic
cancers in our cases [13]. A previous study in Sweden
revealed that the majority of the HCC in this low
endemic area were diagnosed only at autopsy, reinfor-
cing the value of autopsy for proper assessment of
epidemiological rates for this disease [19].
Lung cancer accounted for the second most frequent
site of discrepancy, in accordance with other studies [2].
In developing countries, tuberculosis may be a source of
confusion in the differential diagnosis of lung cancer,
causing delay in the suspicion of malignancy, as it
occurred in some of our patients. Contrary to the liver
neoplasms, 78% of the lung cancers in this study were
suspected during life, but ended up with no complete
diagnosis work-up, possibly reflecting the intrinsic
diagnostic difficulties of this neoplasm or an advanced
disease stage at admission.
Pancreatic cancer was the third most common cancer
presenting as a discrepant clinico-pathological diagno-
sis. The retroperitoneal localization of the pancreas with
the limitations of diagnostic procedures assessing this
area and the late clinical presentation [1] may explain
the high rate of un- and misdiagnosis. In fact, together
with lung, tumors of the pancreaticobiliary system
represent the most frequent cancers, presenting as
unknown primary in some autopsy series [1,4].
The few studies that have specifically analyzed
diagnostic discrepancies in neoplasms included both
the solid and the hematological types [2,5,8]. Similar
studies described higher discrepancy rates in the
diagnosis of neoplasms to older age [2], lower duration
of hospitalization [8], and not being admitted to a
specialized unit [26]. In our study, we could not find
differences regarding age and sex between concordant
vs. discordant categories.
The first category of discrepancies in the present study
considers those cases related to a non-clinical suspicion
of a malignancy. We found a rate of 11.7% (or 15.9% if
the incidental cases were considered) in this category.
Burton et al. related a 103/225 (45.8%) discrepancy
 ARTICLE IN PRESS
872 L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873
rate [5], and another study [8] found a 228/538 (42.3%)
discrepancy rate after reviewing 1036 autopsies.
Although we have not included the hematological and
central nervous system cancers in our study, the
discrepancy rates of undiagnosed cancers were not
higher than those found in other studies. Despite all
comprehensive medical work-ups, a significant propor-
tion of malignancies remains undiagnosed, reinforcing
the importance of autopsy as an accuracy tool for
epidemiological data [2,16]. Di Furia et al. stressed that
in such situations, the diagnostic attitude of the clinician
is also under evaluation, because many of the described
discrepancies could have been diagnosed with the
available diagnostic techniques [9].
A second category of discrepancy was related to
patients with misdiagnosed solid neoplasms, without a
complete or correct diagnostic follow-up. They repre-
sented 68.4% of the total discrepancies or 25.4% of the
patients with single non-incidental solid neoplasms.
These numbers are higher than those presented by other
studies. De Pangher-Manzini et al. [8], Gobatto et al.
[12], and Burton et al. [5] found a 10%, 19%, and 3.5%
rate, respectively. Patients within the misdiagnosed
category presented more increased frequency of metas-
tasis than the undiagnosed group. This is predictable,
because the presence of metastasis may indicate the
presence of cancer. We could not identify other risk
factors related to the occurrence of a misdiagnosis in our
study.
There might be some explanations for the high
discrepancy rates found in this study. The mean
hospitalization duration was approximately 11 days
(range 2–62): perhaps not enough to fully work out the
diagnosis of a neoplasm. However, it is difficult to define
what would be a cut-off in terms of days needed to
perform a cancer diagnosis. Also, similar discrepancy
studies in cancer did not take into account the
hospitalization duration to categorize their data [3,8,9].
We have also excluded the patients with incidental and
dual/triple multiple primary neoplasms from the
study that could have contributed to an even higher
diagnostic discrepancy rate in the discordant category
of misdiagnosis. Our data may also reflect deficiencies
in the primary care health system related to neoplasm
diagnosis, a subset of diseases that require complex
diagnostic steps. Therefore, some of the patients
might arrive at a specialized treatment center with
untreated advanced disease and poor clinical status and
die before the diagnostic procedures can be completed.
In fact, 43.8% of these patients died in the emergency
rooms, many of them with signs of advanced metastatic
disease. Vadillo et al. described that 12.4% of the
autopsies with clinical–pathological discrepancies in
their 266 general autopsies study could be explained
by patients with poor clinical status and irreversibility of
the disease [23].
Certainly, our study has limitations. Possibly, the
autopsied cases were those with high likelihood of
discrepancies, overestimating the discrepancy rates.
However, this is an inherent problem to studies with
similar design and, we believe, was minimized in this
study by the high autopsy rate in our institution. The
hospital is a highly specialized teaching facility, so it is
unlikely that constraints in diagnostic technologies or
therapeutics would negatively affect patient care and,
consequently, occurrence of autopsy discrepancies.
Indeed, a previous study of unselected 300 autopsies in
our institution shows overall discrepancy rates of 7%
(Dr. Paulo H.N. Saldiva, unpublished data, personal
communication), in accordance with other studies done
in general hospitals [2,3,21].
In summary, liver, lung, and pancreatic neoplastic
sites presented the highest rate of clinico-pathological
diagnostic discrepancy. Most of the inconsistencies were
related to an incomplete diagnostic work-up prior to
death. We confirm that autopsy remains an extremely
valuable tool for checking diagnostic accuracy, and may
further provide the basis for a better understanding of
diagnostic deficiencies in different circumstances.
Acknowledgments
The authors thank the Conselho Nacional de
Desenvolvimento Cientı́fico e Tecnológico (CNPq) and
LIM05-HCFMUSP for the financial support.
References
[1] N. Al-Brahim, C. Ross, B. Carter, K. Chorneyko, The
value of postmortem examination in cases of metastasis of
unknown origin-20-year retrospective data from a tertiary
care center, Ann. Diagn. Pathol. 9 (2005) 77–80.
[2] D.V. Avgerinos, J. Bjornsson, Malignant neoplasms:
discordance between clinical diagnoses and autopsy
findings in 3118 cases, APMIS 109 (2001) 774–780.
[3] F.W. Bauer, S.L. Robbins, An autopsy study of cancer
patients. I. Accuracy of the clinical diagnoses (1955 to
1965) Boston City Hospital, JAMA 221 (1972) 1471–1474.
[4] H. Blaszyk, A. Hartmann, J. Bjornsson, Cancer of
unknown primary: clinicopathologic correlations, APMIS
111 (2003) 1089–1094.
[5] E.C. Burton, D.A. Troxclair, W.P. Newman III, Autopsy
diagnoses of malignant neoplasms: how often are clinical
diagnoses incorrect?, JAMA 280 (1998) 1245–1248.
[6] A. Coradazzi, A. Morganti, M. Montenegro, Discrepan-
cies between clinical diagnosis and autopsy findings,
Brazil. J. Med. Biol. Res. 36 (2003) 385–391.
[7] Council of scientific affairs. Autopsy, A comprehensive
review of current issues, JAMA 258 (1987) 364–369.
[8] V. de Pangher Manzini, M.G. Revignas, A. Brollo,
Diagnosis of malignant tumor: comparison between
 ARTICLE IN PRESS
L.T. Grinberg et al. / Pathology – Research and Practice 204 (2008) 867–873
clinical and autopsy diagnoses, Hum. Pathol. 26 (1995)
280–283.
[9] L. Di Furia, A. Piga, S. Marmili, et al., The value of
necropsy in oncology, Eur. J. Cancer 27 (1991) 559–561.
[10] H. Friederici, Reflections on the post mortem audit,
JAMA 260 (1988) 3461–3465.
[11] L. Gabrscek, Should we confirm our clinical diagnosis
certainly by autopsy?, Intensive Care Med. 27 (2001)
1750–1755.
[12] F. Gobbato, F. Vecchiet, D. Barbierato, M. Melato, R.
Manconi, Inaccuracy of death certificate diagnoses in
malignancy: an analysis of 1,405 autopsied cases, Hum.
Pathol. 13 (1982) 1036–1038.
[13] C.S. Goncalves, F.E. Pereira, L.C. Gayotto, Hepatocel-
lular carcinoma in Brazil: report of a national survey,
Rev. Inst. Med. Trop. Sao Paulo 39 (1997) 165–170.
[14] H. Grosse, Positive and negative false diagnosis of cancer
detected by autopsy (author’s translation), Arch. Gesch-
wulstforsch 50 (1980) 794–800.
[15] P. Guariglia, C. Abrahams, The impact of autopsy data
on DRG reimbursement, Hum. Pathol. 16 (1985)
1184–1186.
[16] K. Inoue, T. Yoshikawa, Discrepancies between clinical
and autopsy diagnoses in cases of malignant neoplasm,
Chest 121 (2002) 309.
[17] P. Nemetz, C. Beard, D. Ballard, et al., Resurrecting the
autopsy: benefits and recommendations, Mayo Clin.
Proc. 64 (2007) 1065–1076.
[18] S.M. Pastores, A. Dulu, L. Voigt, N. Raoof, M. Alicea,
N.A. Halpern, Premortem clinical diagnoses and post-
873
mortem autopsy findings: discrepancies in critically ill
cancer patients, Crit. Care 11 (2007) R48.
[19] V.R. Sarode, B.N. Datta, A.K. Banerjee, et al., Autopsy
findings and clinical diagnoses: a review of 1000 cases,
Hum. Pathol. 24 (1993) 194–198.
[20] K.G. Shojania, E.C. Burton, K.M. McDonald, L. Gold-
man, The autopsy as an outcome and performance measure,
Evid. Rep. Technol. Assess. (Summ.) 58 (2002) 1–5.
[21] K.G. Shojania, E.C. Burton, K.M. McDonald, L. Gold-
man, Changes in rates of autopsy-detected diagnostic
errors over time: a systematic review, JAMA 289 (2003)
2849–2856.
[22] K.C. Suen, L.L. Lau, V. Yermakov, Cancer and old age.
An autopsy study of 3535 patients over 65 years old,
Cancer 33 (1974) 1164–1168.
[23] M. Vadillo, R.P. Pujol, X. Corbella, T. Gorriz, P. Rabasa,
R. Bernat, Improvements in clinical diagnostic accuracy
after a 5-year systematic analysis of clinical and autopsy
discrepancies, Arch. Pathol. Lab. Med. 130 (2006)
1261–1262.
[24] B. Veress, I. Alafuzoff, A retrospective analysis of clinical
diagnoses and autopsy findings in 3,042 cases during two
different time periods, Hum. Pathol. 25 (1994) 140–145.
[25] C.P. Wild, A.J. Hall, Primary prevention of hepatocel-
lular carcinoma in developing countries, Mutat. Res. 462
(2000) 381–393.
[26] A.C. Xavier, S.A. Siqueira, L.J. Costa, T. Mauad, P.H.
Nascimento Saldiva, Missed diagnosis in hematological
patients—an autopsy study, Virchows Arch. 446 (2005)
225–231.