Conventional Autopsy versus Minimally Invasive Autopsy with Postmortem MRI, CT, and CT-guided Biopsy: Comparison of Diagnostic Performance

Introduction

Conventional autopsy is a valuable tool, particularly for quality control in health care (1–3). Nevertheless, autopsy rates have been rapidly decreasing for various reasons, such as lack of interest of clinicians and next of kin due to overconfidence in premortem diagnostics, reluctance of family members to provide consent to autopsy because of the invasiveness of the procedure, reluctance of pathologists to perform autopsies, budgetary issues, and ideological opposition to postmortem investigation (4–11). Hence, noninvasive or minimally invasive alternative autopsy methods are being developed (12).

The entire body can be visualized with postmortem CT and MRI (13–15), and imaging-guided biopsy can be performed to obtain tissue for histologic examination (16). In addition, CT angiography can be performed (17–23). Some of these methods are already used to support or even substitute for the forensic autopsy (24–26). In the clinical setting, noninvasive or minimally invasive autopsies of fetuses, newborns, and infants (27) have gained acceptance with parents and physicians along with political and public interest (28). However, they are still rarely applied in adult patients.

The main purpose of this study was to compare immediate cause of death established with conventional autopsy and minimally invasive autopsy in adults who died in a clinical setting, under the a priori hypothesis that the performance of minimally invasive autopsy and conventional autopsy would not be significantly different.

Materials and Methods

Study Design

The institutional review board approved this study before data collection. Written informed consent was obtained from the next of kin for all cadavers.

Our prospective, single-center, cross-sectional study in an academic hospital was performed to compare the diagnostic performance of minimally invasive autopsy with that of conventional autopsy. Minimally invasive autopsy was followed by conventional autopsy. Those involved in minimally invasive autopsy and conventional autopsy were blinded to each other’s findings. The primary outcome measure was performance in establishing the immediate cause of death with findings from minimally invasive autopsy and conventional autopsy. The secondary outcome measures were diagnostic yield for all diagnoses, major diagnoses, and grouped major diagnoses, frequency of clinically unsuspected diagnoses and causes of death, and the percentage of answered clinical questions.

Cadavers

All 2197 patients aged 18 years and older who died at the Erasmus University Medical Center in 2012–2014 were eligible for this study if written informed consent was obtained from next of kin for minimally invasive autopsy and conventional autopsy.

Exclusion criteria were as follows: suspected unnatural cause of death, body size exceeding height of 16 inches in supine position (limitation for MRI), known or suspected high-risk transmittable disease (eg, human immunodeficiency virus, tuberculosis, hepatitis B, and hepatitis C), and open abdominal wounds. Pathologic examination of the brain was not compulsory for inclusion.

The size of the cohort was determined by the available 3 years for inclusion, with the goal of assembling a substantially larger cohort than in our earlier study (16).

Preparation for Autopsy Procedures

For each cadaver, the clinically assumed immediate cause of death, specific clinical questions (both entered on the autopsy request forms), and a comprehensive medical history of the deceased patient, including premortem imaging findings, were collected and made available to the teams performing minimally invasive autopsy and conventional autopsy.

Minimally Invasive Autopsy Procedure

MR images and CT scans were obtained according to standardized protocols (Tables 1, 2).

A board-certified radiologist (A.C.W., with 10 years of experience in postmortem imaging) performed the initial reading of the MR images and CT scans, according to protocol (Appendix E1 [online]), compared these images to available premortem images, and marked suspected pathologic lesions for biopsy.

The minimally invasive autopsy researchers (B.M.B., resident in pathology; I.M.W., who recently obtained Doctor of Medicine degree; and J.W.O., with 36 years of experience), who had a brief training in performing CT-guided biopsies, obtained samples for biopsy with a reusable biopsy gun (Bard Magnum; Bard Biopsy Systems, Tempe, Ariz) (12-gauge needle) according to protocol. Samples were obtained from the heart, lungs, liver, kidneys, and spleen in all cases, even if there were no suspicious imaging findings. If lesions were suspected in these organs or elsewhere in the body, they were also sampled.

Stereotactic biopsy was performed in the brain (Appendix E1 [online]) with assistance from a board-certified neurosurgeon (R.D., with 9 years of experience in neurosurgery) or a resident in neurosurgery (up to 5 years of training) (Appendix E1 [online]). The pathologist (J.W.O.) and researcher (B.M.B.) from the minimally invasive autopsy team examined the microscopic slides from the biopsies; when in doubt, they consulted pathologists with specific expertise to reach a conclusion.

A board-certified general radiologist (N.S.R., with 10 years of experience in postmortem imaging) independently performed a second reading (within 4 weeks of the first reading) of the CT and MR images. A board-certified cardiovascular radiologist (A.P., with 10 years of experience in cardiovascular radiology) performed a second reading of the MR images of the heart. Both readers were blinded to the findings of the initial radiologic reading. In case of disagreement between the initial and second readings, consensus was reached in joint sessions.

Radiologic and histologic findings were combined in the minimally invasive autopsy report, which included clinical history, postmortem diagnoses, proposed cause of death, and answers to clinical questions.

Conventional Autopsy Procedure

The day after minimally invasive autopsy, a resident in pathology (experience: varying from starting to 5 years of training), supervised by the attending pathologist (rotation among all board-certified staff pathologists, with 5–35 years of experience), performed conventional autopsy according to departmental protocol, as published previously (16) (Appendix E1 [online]).

The conventional autopsy report included clinical history, postmortem diagnoses, the proposed cause of death, and answers to clinical questions and was authorized by the attending pathologist.

Data Extraction and Comparison of Autopsy Procedures

Agreement on cause of death.—Immediate cause of death determined by means of minimally invasive autopsy and conventional autopsy were compared in three successive reviews by independent experts, as described in Appendix E2 (online).

Diagnoses.—A researcher from the minimally invasive autopsy team (B.M.B.) extracted and coded all different postmortem diagnoses from the final minimally invasive autopsy and conventional autopsy reports according to the International Classification of Diseases, 10th revision (29). Diagnoses were sorted by 20 organ and tissue categories and one category of general diagnoses.

Per case, a researcher on the minimally invasive autopsy team (I.M.W.) extracted and coded all premortem diagnoses known from the clinical evaluation, using the prepared International Classification of Diseases, 10th revision, list. The minimally invasive autopsy pathologist and radiologist jointly scored all postmortem diagnoses per minimally invasive autopsy, and two independent pathologists together scored all postmortem diagnoses per conventional autopsy. Postmortem diagnoses were scored as certain according to established radiologic and/or pathologic criteria or as probable if there was any uncertainty. Furthermore, a postmortem diagnosis was classified as major if it was directly related to the proposed immediate cause of death determined with minimally invasive autopsy or conventional autopsy.

After the cause of death was established by consensus (“consensus cause of death”), related major diagnoses were retrospectively combined to grouped major diagnoses (eg, necrosis of the lungs, plus infection of the lungs, plus acute pneumonia).

The diagnostic errors at minimally invasive autopsy and conventional autopsy were retrospectively classified. A perceptual error was defined as an abnormality that, though present, was not reported or as an abnormality that, although absent, was reported (30–32). A cognitive error was defined as an abnormality that, although reported correctly, was not correctly interpreted. An error was defined as a sampling error when a biopsy of a suspected radiologic finding was negative but findings of conventional autopsy were positive.

Clinical questions.—All clinical questions asked by the treating physicians before the start of the minimally invasive autopsy and conventional autopsy were extracted from the autopsy request forms. When the autopsy reports had been finalized, we scored whether the questions had been answered.

Statistical Analysis

All data were sealed in an electronic database (OpenClinica Community, version 3.1.3.1). Agreement on immediate cause of death between minimally invasive autopsy and conventional autopsy, and the percentage in which immediate cause of death was classified as correct, were calculated for each method. The clinically assumed cause of death, extracted from the autopsy request forms, was compared with the consensus cause of death.

Diagnostic yield; the contribution of minimally invasive autopsy and conventional autopsy to all diagnoses, major diagnoses, and grouped major diagnoses; and the percentage of clinically unsuspected diagnoses were calculated by using software (SPSS Statistics, version 21.0 [IBM, Armonk, NY]; Excel, Microsoft Office 2011 [Microsoft, Redmond, Wash]). To test for differences in the detection of cause of death, we used the McNemar test. P < .05 was considered indicative of a statistically significant difference. For grouped major diagnoses, we also analyzed overlap between minimally invasive autopsy, conventional autopsy, and the premortem clinical evaluation.

The performance of CT and MRI in establishing grouped major diagnoses and immediate cause of death, and the added value of the imaging-guided biopsy, were retrospectively analyzed. Interobserver agreement for major postmortem diagnoses made with CT and MRI was calculated with the κ statistic.

We also calculated the percentage of clinical questions that were answered with the minimally invasive autopsy and conventional autopsy procedures. The χ² test was used to test for differences in these proportions. P < .05 was considered indicative of a statistically significant difference.

Previous Studies in the Same Cohort

From 24 cases, tissues were collected for analysis of RNA quality (33). All 99 cases were investigated for CT and MRI features of postmortem change (34) and for the accuracy of myocardial CT, MRI, and biopsies in the detection of acute and chronic ischemia (35). Only the latter study overlaps with the present one as far as myocardial ischemia is concerned.

Results

Inclusion and Data Acquisition

From January 2012 to December 2014, conventional autopsy was performed in 295 of 2197 deceased adults (13.4%). Of those 295 cadavers, consent for minimally invasive autopsy was obtained for 139 (47.1%). Both minimally invasive autopsy and conventional autopsy were performed in 99 of the 295 cadavers (33.6%); these cases were included in our study. Twenty of the 99 cadavers (20%) underwent both stereotactic brain biopsy and brain autopsy, four (4.0%) underwent brain biopsy alone, and 18 (18%) underwent brain autopsy only. A flowchart of the included cases, along with demographic data, is shown in Figure 1.

Figure 1:

Download as PowerPointOpen in Image Viewer

The mean time (±standard deviation) between death and minimally invasive autopsy was 23.2 hours ± 15.6. The mean procedure time for the minimally invasive autopsy, including transportation, was 6.28 hours ± 1.07. The mean time between minimally invasive autopsy and conventional autopsy was 9.47 hours ± 1.06.

At minimally invasive autopsy, 1574 biopsy samples were obtained. Of those 1574 samples, 338 (21.5%) were targeted at a suspected abnormality seen at imaging. In 22 of the 99 cases (22%), pleural effusion, ascites, and/or cerebrospinal fluid were obtained for cytologic examination.

Agreement on Cause of Death

The immediate cause of death determined with minimally invasive autopsy agreed with that determined with conventional autopsy in 91 of the 99 cadavers (92%) (Table 3). Of the eight discordant cases, the consensus cause of death was determined with minimally invasive autopsy in five cases and with conventional autopsy in three (Table 4), resulting in a correct immediate cause of death in 96 (97%) minimally invasive autopsies and 94 (95%) conventional autopsies (P = .73).

The performance of imaging (CT and MRI) alone in establishing the immediate cause of death is shown in Table 5. In 11 minimally invasive autopsies, the immediate cause of death could have been established without the need for biopsy. Those causes included tension pneumothorax, massive air embolus (Fig 2), type A aortic dissection, esophagopleural fistula, ruptured aneurysm of abdominal aorta, repeat bleeding of cerebral arteriovenous malformation, and acute subdural and intracerebral hemorrhages with compression and cerebral ischemia.

Figure 2:

Download as PowerPointOpen in Image Viewer

Correlation with Premortem Clinical Evaluation

In 65 of the 99 cadavers (66%), the clinically presumed cause of death was the same as the consensus cause of death found with minimally invasive autopsy and/or conventional autopsy. In an additional 17 cadavers (17%), the consensus cause of death was mentioned in the clinical differential diagnosis. In the remaining 17 cases (17%), the cause of death was not suspected clinically. The latter causes were pneumonia (n = 5), myocardial infarction (n = 2), type A aortic dissection (n = 2), and tension pneumothorax, massive air embolus, multiple organ failure (due to disseminated bladder cancer), acute cellular (grade A2) lung rejection, mesenteric ischemia, sepsis (due to primary biliary cirrhosis with hepatocellular carcinoma), disseminated intravascular coagulation (on clinical grounds probably due to viral infection), and subdural hematoma (n = 1 each). Three of the 17 clinically unsuspected immediate causes of death were found only with minimally invasive autopsy (tension pneumothorax, massive air embolus, type A dissection), and two were found only with conventional autopsy (severe coronary atherosclerosis causing acute coronary syndrome, mesenteric ischemia).

Diagnoses

Within our study population, there were 347 different postmortem International Classification of Diseases diagnoses; of these, 230 (66.3%) were classified as major in at least one case (Table 6). From the minimally invasive autopsy and conventional autopsy reports of the 99 included cadavers, a total of 3097 postmortem diagnoses were extracted (Table E1 [online]); minimally invasive autopsy helped identify 2448 of the 3097 postmortem diagnoses (79.0%) and conventional autopsy helped identify 1421 (45.9%). Of the 3097 postmortem diagnoses, 1372 (44.3%) were classified as major. The 1372 major diagnoses were detected with minimally invasive autopsy in 992 cases (72.3%) and with conventional autopsy in 919 (67.0%).

After retrospective grouping, 85 different grouped major diagnoses remained. The postmortem techniques together scored 283 grouped major diagnoses. In addition, there were five certain diagnoses at premortem clinical evaluation that were not scored as certain with either postmortem method but were nevertheless classified as a major diagnosis because they were related to immediate cause of death and therefore added to the list of grouped major diagnoses. Of the 288 grouped major diagnoses, 259 (89.9%) were found with minimally invasive autopsy and 224 (77.8%) with conventional autopsy; 200 of the 288 grouped major diagnoses (69.4%) were found with both minimally invasive autopsy and conventional autopsy.

Of the 288 grouped major diagnoses, 124 (43.1%) were clinically unsuspected, with 111 of the 124 (89.5%) found with minimally invasive autopsy and 92 (74.2%) found with conventional autopsy. Seventy-nine grouped major diagnoses were detected with both methods, 32 were detected only with minimally invasive autopsy, and 13 were detected only with conventional autopsy. Agreement between premortem clinical evaluation, minimally invasive autopsy, and conventional autopsy for all grouped major diagnoses and for grouped major diagnoses in lungs and for all neoplastic diseases is illustrated in Venn diagrams (Fig 3).

Figure 3:

Download as PowerPointOpen in Image Viewer

The performance of CT, MRI, biopsies, and conventional autopsy separately in the detection of grouped major diagnoses per category is shown in Table E2 (online). Interobserver agreement (κ value) for radiologic detection of grouped major diagnoses was excellent: 0.91 for CT and 0.80 for MRI.

Clinical Questions

Apart from the cause of death, 219 additional clinical questions were asked by the referring physicians in 86 of the 99 autopsy request forms (87%). Of these 219 questions, 189 (86.3%) were answered with minimally invasive autopsy and 182 (83.1%) with conventional autopsy (P = .35).

Errors

There were 16 perceptual errors with minimally invasive autopsy: 12 at imaging and four at microscopic examination. There were seven cognitive errors, all at microscopy. Four diagnoses were missed due to sampling error. There were 26 perceptual errors with conventional autopsy: nine at gross examination and 17 at microscopic examination. There were six cognitive errors.

Discussion

In our prospective study of a cohort of adult patients who died in-hospital, minimally invasive autopsy combining MRI, CT, and image-guided biopsies and conventional autopsy did not show a significant difference in identifying immediate cause of death and answering clinical questions. Minimally invasive autopsy had a higher yield than conventional autopsy for postmortem diagnoses, many of which were clinically unsuspected. The methods we used for minimally invasive autopsy appeared adequate and are technically feasible for most hospitals. Our technique of minimally invasive autopsy was improved compared with an earlier study, in which there were fewer biopsies and/or random biopsies guided with US (16).

Apart from our current study, to our knowledge the study by Weustink et al (16) is the only one reporting on the diagnostic performance of minimally invasive autopsy in patients who died in-hospital using the combination of CT, MRI, and postmortem biopsy. In a cohort of 182 coroner’s cases, Roberts et al (13) compared CT and MRI to conventional autopsy and found an agreement for cause of death of 70% (95% confidence interval: 62.6%, 76.4%). Most often missed were ischemic heart disease, pulmonary embolism, pneumonia, and intra-abdominal lesions. Westphal et al (14), who investigated the feasibility of minimally invasive autopsy with CT in only 29 cases, reported accuracy for cause of death of 68% and a positive predictive value of 75%. In agreement with these studies, we found that CT and MRI alone could not with certainty help diagnose common causes of death such as pneumonia, myocardial infarction, peripheral pulmonary emboli, gastrointestinal ischemia, and sepsis without biopsy confirmation.

Recent studies investigated the additional value of CT angiography. The study by Westphal et al (19) achieved an accuracy of 80% for cardiac causes of death (19). In a selected group of 50 cases, Wichmann et al (20) compared diagnoses (not cause of death) identified with CT angiography or conventional autopsy. They found 16 new major diagnoses, comparable to grouped major diagnoses in our study; 93.8% of these were identified with CT angiography and 87.5% with conventional autopsy. These figures are comparable to the clinically unsuspected grouped major diagnoses in our study. Most recently, Rutty et al (17) established a correct cause of death in 86% of cases using CT angiography when compared with a consensus cause of death. Bolliger et al (23) and Ross et al (21) combined CT, CT angiography, and biopsies. In a systematic review, the pooled sensitivity for cause of death was 90.9% (95% confidence interval: 74.5%, 97.6%) (12).

Postmortem angiography, an important technical advancement of postmortem imaging (17,20,21), appears not necessary for establishing immediate cause of death if cardiac MRI protocols are used in combination with targeted myocardial biopsy. For subclinical arterial stenoses and localization of the origin of bleeding, angiography is indispensable. However, the logistics of postmortem angiography require expertise not yet available in general hospitals.

Minimally invasive autopsy provides a permanent auditable record of the entire body that can be referred to by pathologists, radiologists, clinicians, scientists, and next of kin (12,13). For minimally invasive autopsy to become a routine procedure, cultural adaptation will be required, especially for pathologists. Minimally invasive autopsy requires dedicated radiologists and pathologists who know the strengths and limitations of both disciplines and are willing to jointly answer the questions asked by the clinicians. In view of the expensive imaging equipment, it seems sensible to centralize minimally invasive autopsy and/or postmortem imaging in regional autopsy facilities, where state-of-the-art imaging units and adequately trained specialists are available. These centers can take the lead in developing standardized consent forms, triaging cases, establishing efficient operating procedures, and managing quality. A centralized approach may maximize the use of the equipment, thereby reducing depreciation and running costs, and minimize interference with clinical radiology. Finally, adoption of these methods also requires awareness from both medical professionals and the lay public.

Our study had some limitations. The teams performing minimally invasive autopsy and conventional autopsy were blinded to each other; however, for the pathologist performing conventional autopsy, the biopsy sites could potentially lead to increased suspicion of a pathologic condition in the biopsied areas. We used consensus cause of death as a reference standard instead of conventional autopsy. However, errors are not uncommon with conventional autopsy alone (12,16,36). Errors in the interpretation of histologic findings obtained with conventional autopsy were corrected to compensate for this bias. Furthermore, when abnormal air collections were present, imaging appears more useful than conventional autopsy.

We did not calculate sensitivity and specificity of the two autopsy procedures for each diagnosis but scored their respective diagnostic yields. Specificity cannot be calculated in a meaningful way because it directly depends on the total number of diagnoses in the entire study population, since the number of true-negative diagnoses per case increases when more diagnoses are encountered in the population.

Only a small proportion of the 2197 patients who died during our study period underwent autopsy, and even fewer underwent minimally invasive autopsy. Furthermore, a number of minimally invasive autopsies for which we obtained consent could not be executed for logistical reasons. However, the immediate causes of death in our cohort were representative of the findings with routine autopsies in Western countries (37). Consent for histologic sampling from the brain was obtained in a small proportion of minimally invasive autopsies (24 of 99 cases) and conventional autopsies (38 of 99 cases). We also did not address costs and benefits of the minimally invasive autopsy and supplementary postmortem imaging.

In summary, there was excellent agreement between minimally invasive autopsy and conventional autopsy as to immediate cause of death. Importantly, minimally invasive autopsy resulted in a higher yield of diagnoses. Seventeen percent of immediate causes of death and 43% of grouped major diagnoses were not suspected at clinical examination, which illustrates the lasting importance of postmortem examination for quality control in health care. Because minimally invasive autopsy and conventional autopsy have their own strengths and weaknesses, future postmortem examination in the clinic will probably use combinations of the two approaches.

Disclosures of Conflicts of Interest: B.M.B. disclosed no relevant relationships. A.C.W. disclosed no relevant relationships. I.M.W. disclosed no relevant relationships. J.H.v.d.T. disclosed no relevant relationships. A.P. disclosed no relevant relationships. R.D. disclosed no relevant relationships. J.B. disclosed no relevant relationships. N.S.R. disclosed no relevant relationships. M.A.d.B. disclosed no relevant relationships. F.J.v.K. disclosed no relevant relationships. G.P.K. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: institution receives money for board membership from Quantib B.V.; institution receives money for consultancy activities from Bracco; institution has grants/grants pending with GE Healthcare, Bayer, Bracco, and Siemens. Other relationships: disclosed no relevant relationships. M.G.M.H. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: receives royalties from Cambridge University Press. Other relationships: disclosed no relevant relationships. J.W.O. disclosed no relevant relationships.

Author Contributions

Author contributions: Guarantors of integrity of entire study, B.M.B., R.D., N.S.R., J.W.O.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, B.M.B., A.C.W., J.B., N.S.R., J.W.O.; clinical studies, B.M.B., A.C.W., I.M.W., J.H.v.d.T., R.D., J.B., N.S.R., J.W.O.; experimental studies, A.P., N.S.R., M.A.d.B., J.W.O.; statistical analysis, B.M.B., I.M.W., N.S.R., M.G.M.H., J.W.O.; and manuscript editing, B.M.B., A.C.W., I.M.W., J.H.v.d.T., R.D., J.B., N.S.R., M.A.d.B., G.P.K., M.G.M.H., J.W.O.