The paper by Turnbull and colleagues (1) on the decline of the adult hospital autopsy rate in UK prompted me to review and extend the Norwich data that they kindly quoted (2). I calculated the adult hospital autopsy rate since our publication (including that relating to the first five months of this year) using the method that we both employed. The modest improvement that we reported in the adult hospital autopsy rate in Norwich between 2003 and 2005 was not subsequently sustained and the overall trend continued downward. During the period since 1 January 2006, the mean adult hospital autopsy rate in Norwich was 1.1% (range 0.1%-1.7%). The rate in 2013 was 0.6%, in line with the data presented by Turnbull et al. (1).

During 2006-2014, there has been a rising trend here in the annual number of autopsies undertaken for HM Coroner (mean 1191; range 1037- 1355). Most such cases represented deaths in the community or in the A&E Department, but some were patients who died in the wards after admission. In such cases the attending clinicians were unable to complete a certificate of the cause of death. While some such cases followed trauma or other unnatural events, many were natural deaths in which the clinical course was difficult to understand, sometimes despite detailed examination and testing; no confident diagnostic label was attached before death. Such Coroner's cases arising from within the hospital patient population represent medicolegal quasi-hospital autopsies and provide a partial means of alleviating the decline in consented adult hospital autopsies. It is my experience that many such cases provide valuable opportunities to study disease and the efficacy of diagnosis and treatment in the hospital setting.

The continuing decline of the adult hospital autopsy is of great concern. It is likely to multifactorial and to be influenced by the attitudes of clinicians, pathologists and bereaved families (3,4). It does have important implications for clinical audit, diagnostics and medical education (3). While such matters can, at least in part, be addressed by the coronial autopsies that are undertaken on certain hospital deaths, such cases are relatively few and do not necessarily reflect the normal population of patients who die in hospital; they are likely to provide a skewed experience. We know little about the numbers of such cases or what stimulates a clinician to discuss a seemingly natural death with the Coroner, rather than to use clinical judgement and suggest the most likely cause of death and complete a death certificate. It may be that clinicians wish to have a greater degree of certainty than such an approach allows.

While medicolegal quasi-hospital autopsies are an imperfect substitute for properly conducted hospital autopsies, they are better than nothing and should, perhaps, be considered when assessing adult hospital autopsy rates. The whole matter clearly requires further research. Autopsy practice in general is threatened by the decline in the numbers of histopathologists who are prepared to undertake any such examinations. That itself is a matter of concern and requires deeper understanding and rectification. It chimes with Byard's view that the decline in the autopsy is, at least in part, the responsibility of histopathologists (4).

References

1. Turnbull A, Osborn M, Nicholas N. Hospital autopsy: Endangered or extinct? J Clin Pathol Published Online First: [15 06 2015] doi:10.1136/jclinpath-2014-202700

2. Limacher E, Carr U, Bowker L, et al. Reversing the slow death of the clinical necropsy: developing the post of the Pathology Liaison Nurse. J Clin Pathol 2007; 60:1129-34.

3. Carr U, Bowker L, Ball RY. The slow death of the clinical post- mortem examination: implications for clinical audit, diagnostics and medical education. Clin Med 2004; 4:417-23.

4. Byard RW. Who's killing the autopsy? A new tool for assessing the causes of falling autopsy rates. Med J Aust 2005; 183:654-5.

Conflict of Interest:

None declared

Dear Sir / Madam,

We are happy to address the points raised by Lee et al in their commentary on our paper [1] and thank them for their interest in it.

Lee et al correctly note that our re-assessment was of the mitotic count in these specimens. To clarify, these were carried out by either of two observers (CAD, JL) blinded to the original core and excision grading. The other elements of the tumour grade (tubule formation and nuclear score) were not re-assessed, but were as per the original report, that is to say, as reported prospectively by the original reporting pathologist. Similar to Nottingham, in Edinburgh five pathologists with a specialist or subspecialist interest in breast pathology were involved in reporting these cases. The mitotic counts of cases in which our count on review would have assigned the case to a different mitotic score (M) category were reviewed jointly by both authors and a consensus count agreed. These represented just 7% of our cases. It is thus very unlikely that this had any significant effect on the assessment of grade between core and excision.

Edinburgh is one of the largest breast cancer centres in the UK. We follow the NHSBSP guidelines in our handling of specimens and are aware of the potential difficulties of poor fixation. We do not however have this problem, as evidenced by the grade distribution of tumours in our department which is comparable to other centres on national audit.[2]

Lee's main concerns however appear to be that our series is selective (presumably relating to our inclusion of breast conserving surgery tumours only) and the low proportion of grade 3 tumours (on excision) in our series, suggesting it would lower the potential impact of undergrading on core biopsy.

We did not include tumours treated by mastectomy principally because the majority of these (117/170) would have been excluded on the basis of multifocal disease (56), having received neoadjuvant chemotherapy / systemic therapy (31), having a past history of breast cancer (20) or having had their core biopsy performed in another hospital and not available for review (10). Had the remaining 53 cases been included they would not have made a material difference to the outcome or conclusions of our study and they would only have increased our proportion of grade 3 cancers to 29% (120/412).

More importantly in relation to the potential impact of the proportion of grade 3 tumours in our series Lee at al fail to note that of our symptomatic cases 48% (72/150) were grade 3 on excision. Neither the whole symptomatic group nor the subset of 41 cases scored as (T3, N3, M1) within it showed benefit from application of the modification of the mitotic scores. Of the (T3, N3, M1) subset 14 would have been reclassified as grade 3* on core biopsy, only 6 of which were grade 3 on excision. As for all other subsets, in our experience there is a marginal increase in the sensitivity counterbalanced by a larger reduction in the specificity of cases designated as grade 3 on core biopsy using the modified scores.

Finally, we do not seek to disprove the findings of Lee's group. They have shown their utility in their own population.[3] We demonstrate that they do not work in ours. Our conclusion remains that the current recommended mitotic count thresholds are appropriate and should be maintained. We agree that further studies could contribute to the debate, but note that the treatment of the breast cancer population is changing to include more cases receiving neoadjuvant therapy which may make repetition more difficult.

References:

1. Dhaliwal CA, Graham, C, Loane J. Grading of breast cancer on needle core biopsy; does a reduction in mitotic count threshold improve agreement with grade on excised specimens? J Clin Pathol 2014; 67:1106-8

2. NHS breast screening programme and Association of Breast Surgery: An audit of screen detected cancers for the year of screening April 2012 to March 2013; Public Health England, May 2014.

3. Lee A, Rakha E, Hodi Z et al. Re-audit of revised method for assessing the mitotic component of the histological grade in needle core biopsies of invasive carcinoma of the breast. Histopathology 2012; 60: 1166-7

Conflict of Interest:

None declared

To the Editor: Without doubt the hospital-based autopsy is an effective quality assurance and learning tool. The study by Kuijpers et al. supports this.[1] However, autopsy is a time-consuming and expensive procedure which may sometimes cause distress to the deceased patient's family and be associated with complex consent issues. It is therefore important to ask how far reaching, beyond the pathologist and the referring clinician, is the learning impact of each individual autopsy?

The clinician requesting the autopsy is interested in the bullet- point preliminary summary (based on the macroscopic findings) that is signed out immediately. But how often is the final detailed multi-page report followed up? Autopsies and the associated laboratory tissue- processing usually do not take priority in a busy diagnostic department. Final reports may not be verified for some time after the procedure, decreasing their learning impact.

Frequency of minor diagnostic errors, as measured by the autopsy gold standard, is increasing over time.[1] Improper use of imaging investigations is contributing significantly to incorrect pre-mortem diagnosis.[1] Could it be that the main problem is that there has been a huge increase in the number of diagnostic investigations that can be requested during life? More investigations at our fingertips do not necessarily make things better for the doctor, or for the patient.[1] How do we learn how to optimally use these investigations to avoid misdiagnosis or alternatively to optimally learn from the inappropriate use of past investigations, so that errors are not repeated?

Some of the current issues with use of diagnostic investigations are as follows: 1. Understanding the tests: a. Wrong test selected for the situation - whether it be imaging or diagnostic pathology. This results in false positives or false negatives; b. Sensitivity, specificity and positive/negative predictive value - these parameters of a particular imaging modality or pathology test may not be appreciated. Acting on the false result whether it is negative or positive may have significant detrimental clinical impact; c. With complex histopathology or imaging investigations, are the nuances in the body of the report skipped or overlooked by the treating clinician? 2. Communication: a. With immensely busy clinical loads, how often is the suitability of the investigation or the unexpected final result discussed with the radiologist or pathologist? In other words, how often do we practice in a vacuum because of time constraints? 3. Clinico-pathological and radio-pathological correlation: a. Were the radiologist or pathologist given any clinical information, and if not did they seek such information before reporting?

Obviously selection and interpretation of diagnostic tests is often not clear-cut in the real world. Furthermore, with increasing numbers and complexity of diagnostic tests come increasing workloads for pathologists and radiologists. There is increased expectation from both patients and referring clinicians. Faced sometimes with extraordinary numbers of cases to report each day, the pathologist may fail to look for or may miss a vital component of the history that was provided to them by the radiologist or clinician. This is easy to do in an electronic era where most reports and referrals are scanned and it takes time to search. There will also always be cases where only a proportion of slides or images were viewed before issuing a report. Errors are part of the human condition. The key is to learn from errors and to not repeat them.

In summary, autopsy has always been an excellent tool for quality assessment in diagnostic accuracy. But is it a teaching and quality assurance procedure that is time and cost-effective, with results that are easy to disseminate with maximum learning benefit? Focus on development and delivery of high impact and time-efficient continuing education modules (particularly online) regarding quality assurance errors and diagnostic and investigative medicine has already been demonstrated to be of value,[2] so such modules may represent an alternative solution to these issues.

References 1. Kuijpers C.C.H.J., Fronczek J., van de Goot F.R.W., et al. J Clin Pathol 2014;67:512-519 doi:10.1136/jclinpath-2013-202122 2. Ritchie A., Jureidini E. and Kumar R.K. Educating Junior Doctors to Reduce Requests for Laboratory Investigations: Opportunities and Challenges. Med.Sci.Educ. 2014;24:161-163 DOI 10.1007/s40670-014-0041-2

Conflict of Interest:

None declared

Dear Editor,

I entirely agree with the authors of 'A Survey of reporting of colorectal cancer in Scotland: compliance with guidelines and effect of proforma reporting'1 that proforma reporting should be standard across Scotland for reporting colorectal caner excision specimens. Although obvious, I feel it should be stated the 2007 RCPath dataset standards 2 were issued for a symptomatic population and, as stated in the study, these audited cases were both screening and symptomatic. Also, in their conclusions they allude to the potential impact that neo-adjuvant therapy could play in Serosal Involvement (SI), Lymph node retrieval and extramural venous invasion (EMVI) but state that this aspect could not be accurately assessed in this study. All reported cases in our Health Board are staged using TNM5 and, as such, will be prefixed with 'y' to identify them as having had neo-adjuvant therapy. If this was not known prior to reporting, a supplementary report will be issued after the case has undergone MDT discussion when the patient history is reviewed. This may be unique to our health board but, as it is part of the RCPath dataset, it should also be recorded.

I do find the wording used in 'What this paper adds', where it states that "....this is likely to have serious adverse consequences for patient care." hard to extract from the data presented and this claim is not reported in the article by the authors themselves. The data suggests that there may be a group of patients that are falsely node 'negative' due to insufficient nodal sampling and a further group in whom the EMVI or SI is not identified. However, it does not state the cross over between these groups or if the ones 'missing' the EMVI/SI are node positive patients. Taken together this suggests that a small percentage of patients may have been excluded the option of adjuvant therapy, but without looking at specific patient outcomes and the case slides is it fair to label this as 'serious adverse consequences for patient care'?

I do look forward to the results of the repeat data collection which will, hopefully, show proformas being used across all NHS Scotland boards as well as an uplift in the percentage of boards reaching all the standards. Further investigation of a national electronic dataset would also be welcomed especially if the data required for such audits can be extracted easily and possibly centrally from this. Given the existence of one in Norway maybe we should be moving to access that and use it throughout Scotland? I am sure the follow up audit will also take into account the influences of neo-adjuvant therapies on the audited adverse factors in the rectum, a treatment which is an established local practice, but also the emerging use of neo-adjuvant therapy in advanced colonic cancers. These data could be collated to allow reporting of all cases together and in different cohorts (Treatment naive V's neo-adjuvant) to try to identify the changes attributable to therapy.

Conflict of Interest:

I report GI resection specimens in a health board in Scotland that provided raw data for this survey. (Our board median nodal count is above 17 for both colonic and rectal excisions including post treatment cases)