Does including histological outcomes in a multiparametric MRI report improve prostate biopsy decision making?

Abstract

Objectives:

To see if inserting audited histological outcome data for each Likert score into prostate mpMRI reports was helpful for clinicians counselling patients and influenced prostate biopsy uptake.

Methods:

A single radiologist reported 791 mpMRI scans for query prostate cancer between 2017 and 2019. A structured template which included histological outcome data from this cohort was devised and included in 207 mpMRI reports between January and June 2021. The outcomes of the new cohort were compared with the historical cohort, and with 160 contemporaneous reports without histological outcome data from the four other radiologists in the department. The opinion of this template was sought from referring clinicians who counselled patients.

Results:

The proportion of patients biopsied fell from 58.0 to 32.9% overall between the n = 791 cohort and the n = 207 cohort. This was most noticeable in those scoring Likert 3, where the proportion biopsied fell from 78.4 to 42.9%. This reduction was also seen when comparing the biopsy rates of patients scored Likert 3 by other reporters in a contemporaneous n = 160 cohort without the audit information (65.2%) with the n = 207 cohort (42.9%). 100% of counselling clinicians were in favour and 66.7% said it gave them greater confidence to advise the patient when they did not need a biopsy.

Conclusion:

Fewer low-risk patients choose unnecessary biopsies when audited histological outcomes for the radiologist’s Likert scores are included in mpMRI reports.

Advances in knowledge:

Clinicians welcome reporter-specific audit information in mpMRI reports which could result in fewer biopsies.

Introduction

The 2020 UK National Institute for Health and Care Excellence (NICE) guidelines advise all UK radiologists to report prostate MRI using a 5-point Likert scale which quantifies the risk of significant prostate cancer (csPCa). ^1,2 Prostates are scored as either highly unlikely (1), unlikely (2), equivocal (3), likely (4), or highly likely (5) to contain csPCa.

Biopsies should be minimised in low-risk patients. Prostate biopsies can cause complications including sepsis, bleeding and urinary retention. ¹ Diagnosis of a clinically insignificant cancer can also have significant psychological impact on a patient, despite being unlikely to need treatment or shorten the patient’s life. ^3–6 It is therefore important to identify patients at high risk of csPCa for biopsy, while minimising biopsies in low-risk patients. NICE recommends that patients scoring Likert 1 and 2 (low risk) should not routinely be referred for prostate biopsy, while those scoring Likert 4 or 5 (high risk) should be. ¹

Likert scores vary between reporters. This is partly because Likert scoring lacks the standardisation associated with PI-RADS scoring, which benefits from widely accepted and accredited guidelines. ⁷ There is also little clarity about which clinical factors (e.g. age, PSA density, digital rectal examination findings, family history) should be taken into consideration when assigning the Likert score, and different reporters may take different factors into account to varying degrees depending on availability of clinical information, personal preference and departmental policy. ^8–10 This makes Likert scores susceptible to a greater degree of inter-reporter variability than PI-RADS. Likert score discordance of up to 54% has been reported between low- and high-volume centres ¹¹ and 20% even between experienced reporters. ¹²

Because of the variability inherent to the more subjective Likert scores, the importance of radiologists being aware of their own individual outcomes and predictive accuracy has been recognised. Radiologists have been advised by recent consensus reports ^13,14 to audit their patients’ results, allowing each radiologist to demonstrate the histologically validated outcomes of their own Likert scores.

Despite these recommendations, the authors were not aware of any peer reviewed publications in which individual radiologist-audited histological outcome data for Likert scores were routinely being added to mpMRI reports. The authors agreed with the recommendations and hypothesised that adding context in this way could be of value for referring clinicians and help patients make a more informed decision regarding biopsy.

This paper aimed to evaluate the impact of including such data as part of a structured report for the first time; assessing whether it was helpful to the clinicians counselling patients, if it facilitated patient discussion, and if it influenced the number of patients electing for a prostate biopsy.

Methods and materials

A single radiologist reported 791 patients between 2017 and 2019. The patients were all new referrals to a large quaternary cancer centre in the UK with query prostate cancer, and those with a prior diagnosis of prostate cancer were excluded. Patients were examined on one of three 1.5T MRI models; Aero, Avanto or Sola (Siemens Healthineers, Erlangen, Germany), and settings and protocols were in line with PI-RADS (v. 2) guidance. ¹⁵ These mpMRI reports were used to inform the discussions between the clinical team and the patients when deciding whether or not their risk of csPCa was high enough to proceed to a prostate biopsy.

Several authors suggest that prostate MRI reporters should be core members of prostate multidisciplinary team meetings, should have experience and training in prostate imaging, and should report over 100 scans per year. ^9,13,16 The Likert and PI-RADS scores in the n = 791 audit were prospectively allocated by a single radiologist with 15 years of experience who met these recommendations. This reporter used a standard departmental reporting template. For allocating PI-RADS Scores, the most up-to-date PI-RADS guidance available at the time was used. ^7,15 For Likert scores, the additional clinical factors used were derived from clinic letters and the Patient Pathway Manager (PPM) system. Results are described per patient not per lesion, as this study is looking at patient outcomes in terms of decision to biopsy. For any patient with more than one lesion, the highest scoring lesion was scored.

The histological outcomes of these patients were audited retrospectively using the PPM system, multidisciplinary meeting electronic records, pathology reports and patient electronic notes. An ISUP grade as defined by Epstein et al. ¹⁷ was recorded.

The audit data were added to subsequent radiology reports. Between January and June 2021, a further 207 mpMRI scans were reported by the same radiologist. This time, each PIRADs score allocated was quantified using a standardised report template which included the historical audit data from the previous 791 patients for that radiologist. When these mpMRI reports were used during the discussions between the clinical team and the patients, the risk of csPCa for the Likert score they were given was defined in the context of the audited outcomes of previous patients who had been given that same Likert score by that same radiologist in the past.

The impact of the added audit data was analysed. The proportion of patients in this cohort of 207 who elected to proceed to biopsy was determined and compared with the historical cohort of 791 patients for whom this information had not been available in the reports. Cancer services including prostate biopsy services had continued at this institution despite the COVID-19 pandemic, but it was recognised that any change in uptake in the 2021 cohort due to COVID-19 might affect the results. Comparison was therefore also made with 160 contemporaneous reports without histological outcome data from the four other radiologists who reported prostate mpMRI in the department between January and June 2021. They used the same departmental template for reporting. If a reduction in biopsy uptake had occurred only because of the COVID-19 pandemic, there should be minimal difference between two contemporaneous groups.

There were seven referring clinicians in the department who gave patients the results of their MRI scans and counselled them regarding subsequent prostate biopsy. These clinicians were directly asked by an emailed SurveyMonkey ¹⁸ questionnaire if they felt the inclusion of the audit data in the mpMRI reports changed their practice, and the biopsy uptake rates of patients. Of the seven clinicians, four were consultant urologists, one was a urology nurse specialist and two were advanced clinical practitioners (ACPs). Six of the seven counselling clinicians stated they had seen the histology outcome data in the reports. The one who had never seen a report containing these data were excluded from the analysis.

Results

The demographics of the patients in each cohort can be seen in Table 1. This is followed below by the single reporter histological audit data results of the n = 791 cohort of 2017–2019. The impact of adding these audit results upon on biopsy uptake can be seen in Figure 1. The opinion of the clinicians who counselled patients can be seen in Tables 2 and 3.

Table 1.

Patient demographics of each cohort.

	Cohort
	n = 791	n = 207	n = 160	All
Median age (years)	66	68	65	66
Median PSA (ng/mL)	6.20	5.80	5.83	6.10
Median prostate volume (mL)	48	53	47	49
Median PSAD (ng/mL2)	0.12	0.11	0.12	0.12

PSA, prostate-specific antigen.

The n = 791 cohort is the original cohort reported from 2017 to 2019 by a single reporter which formed the basis of the historical audit. The n = 207 cohort was reported by the same radiologist between January and June 2021 and were the only reports containing the audit results. The n = 160 cohort was reported contemporaneously (January to June 2021) by the four other prostate MRI reporters in the department, and did not contain audit results.

Figure 1.

Differences in biopsy rates in the three cohorts.

Table 2.

Reasons clinicians provided for being in favour of added histological information in radiology reports

	Frequency	Percentage
I find this information useful when discussing the MRI results with patients	6	100%
I find this information has an impact on the patient's decision on whether to proceed with biopsy	2	33%
This information gave me greater confidence to advise the patient when they did not need a biopsy	4	67%
This information gave the patient greater confidence in deciding not to proceed to biopsy	2	33%
I think this information empowers the patient to decide about proceeding to biopsy	3	50%
I prefer to have a consultation with patients when I have this information in the MRI report	1	17%
This information has given me facts about outcomes that I was not previously aware of	4	67%
I may not have given the patient as good advice without this information	1	17%
I share this information with the patient verbally	2	33%

Table 3.

Lesions clinicians felt less likely to refer for biopsy after being provided histological information

	Frequency	Percentage
Less likely in all cases	0	0%
Less likely in Likert 1 and 2	4	67%
Less likely in Likert 3	3	50%
Less likely in Likert 4 and 5	0	0%
Neither less nor more likely	0	0%
More likely to refer a patient for biopsy	0	0%

Histological audit data added to mpMRI reports

The following histological audit data reflect the results of the outcomes of scans for query prostate cancer reported by the single reporter between 2017 and 2019 (n = 791) and was added to the end of each of that reporter’s subsequent mpMRI reports from January to June 2021 (n = 207). The audit data added to the scan report depended on the Likert score the patient had been given, and it was presented exactly as shown below. This information was not added to the reports of mpMRIs reported by other radiologists in the department during the same period of January to June 2021 (n = 160).

Likert 1

This patient has been given a Likert score of 1 which means “very low level of suspicion of prostate cancer on MRI”.

In 2 years, this radiologist scored the scans of 791 patients of which 49 (6.1%) were Likert 1. 5/49 (10%) of these patients had a biopsy and none of the patients biopsied were diagnosed with prostate cancer. When we followed up all the patients who had not had a biopsy over 1–3 years, we could find none who re-presented with cancer.

Likert 2

This patient has been given a Likert score of 2 which means “low level of suspicion of prostate cancer on MRI”.

In 2 years, this radiologist scored the scans of 791 patients of which 411 (52%) were Likert 2. 155 (37%) of these patients had a biopsy. Of these 155 patients 51 (32.9%) had cancer when their biopsies were analysed but only 1 (0.6%) had ISUP Grade 3,4 or 5*. When we followed up all 256 patients who had not had a biopsy over 1–3 years, we found only 2 patients (0.8%) who re-presented with cancer—both with low grade ISUP 2. Please see supplementary materials.

Likert 3

This patient has been given a Likert score of 3 which means “equivocal level of suspicion (neither high nor low) of prostate cancer on MRI”.

In 2 years, this radiologist scored the scans of 791 patients of which 102 (13%) were Likert 3. 80 (78%) of these patients had a biopsy. Of these 80 patients, 53 (66%) had cancer when their biopsies were analysed but only 7 (8.8%) had ISUP Grade 3,4 or 5*.

Likert 4

This patient has been given a Likert score of 4 which means “high level of suspicion of prostate cancer on MRI”.

In 2 years, this radiologist scored the scans of 791 patients of which 63 (8%) were Likert 4. 57 (90%) of these patients had a biopsy. Of these 57 patients, 45 (79%) had cancer when their biopsies were analysed but only 10 (18%) had ISUP Grade 3,4 or 5*.

Likert 5

This patient has been given a Likert score of 5 which means “very high level of suspicion of prostate cancer on MRI”.

In 2 years, this radiologist scored the scans of 791 patients of which 166 (21%) were Likert 5. 161 (97%) of these patients had a biopsy. Of these 161 patients 152 (94%) had cancer when their biopsies were analysed but only 89 (55%) had ISUP Grade 3,4 or 5*.

*The rest of these patients either had no cancer or had a ISUP Grade 1 or 2 diagnosis. ISUP Grade 1 and 2 are still technically prostate cancers but have a very low risk of progressing to cause symptoms or needing treatment within 10 years, so are less likely to be “clinically significant”.

Fewer patients were biopsied overall in the new n = 207 cohort with audit data vs the old n = 791 cohort without audit data reported by the same radiologist (32.9% vs  58%). The most significant decrease in biopsy uptake between the n = 207 cohort and the n = 791 cohort was in lower risk and equivocal patients (0.0% vs  10.2% of Likert 1 cases; 3.3 vs 38.0% of Likert 2 cases; 42.9 vs 78.4% of Likert 3 cases). Higher risk patients retained high biopsy rates (92.6% vs  97.0% of Likert 5 cases).

Of the 460 patients who scored either Likert 1 or 2 in the initial n = 791 cohort with no audit data in their reports, only one was diagnosed with ISUP 3/4/5. Despite this, 34.8% elected for a biopsy. After the inclusion of the audit data, only 1.7% patients with a Likert score of 1 or 2 in the study cohort elected for a biopsy.

When comparing the two contemporaneous cohorts, fewer biopsies were carried out overall in the n = 207 cohort which included the audit information vs the n = 160 cohort which did not (32.9% vs  36.9%). Again, the largest difference in biopsy rate between the n = 207 cohort and the n = 160 cohort was identified in lower risk and equivocal patients (3.3% vs  4.5% of Likert 2 cases; 42.9 vs 65.2% of Likert 3 cases). By contrast, the biopsy uptake rates of the patients with a Likert score of 5 was almost identical in these two groups (92.6% vs  92.0%).

Of the equivocal Likert 3 patients, 78.4% were biopsied in the historical n = 791 cohort. 65.2% were biopsied in the contemporaneous n = 160 cohort without audit information, and 42.9% were biopsied in the n = 207 cohort with the audit data.

The clinicians gave two suggestions for improvement. One suggested that there was too much information, and that the resulting radiology report was too long. Another suggested that the detail provided could confuse patients.

Discussion

Differences in biopsy rates

NICE guidance recommends avoiding prostate biopsy in patients with a Likert score of 1 or 2. ¹ The results show a significant reduction in the number of low-risk patients (Likert 1 or 2) choosing to have potentially unnecessary biopsies (from 34.8 to 1.7%) after audit information was included in the reports.

NICE recommends that risks and benefits should be discussed with patients scoring Likert 3. ¹ Of the 102 patients allocated a Likert score of 3 from the n = 791 cohort, 78.4% were biopsied. However, only 7/102 received a diagnosis of ISUP 3/4/5. Because of this, the reduction in biopsy uptake from 78.4 to 42.9% seen after inclusion of audit data was also likely to benefit many patients scored Likert 3 while missing few cancers.

Finally, NICE advocates that patients with high Likert scores of 4 or 5 are referred for biopsy. ¹ Of the 166 patients allocated a Likert score of 5 from the n = 791 cohort, cohort, 161 (97%) elected to proceed to biopsy, and a far greater proportion of these—89/161 (55%) received a diagnosis of ISUP 3/4/5 than did the patients scored Likert 3 or lower. Appropriately, the number of high-risk Likert 5 patients who elected to proceed to biopsy remained high with the addition of the audit data in the n = 207 cohort: 92.6%. The predominant reason for the remaining patients not proceeding to biopsy was their poor fitness and significant comorbidity.

In essence, although the biopsy rates were much reduced in the patients with the audit results in their MRI reports, this was not because of a global reduction in biopsy take up. Rather, this was due to patients making more informed and appropriate decisions for their particular score. After inclusion of the audit information, the low risk patients were less inclined to have a biopsy, whereas the rates of biopsy uptake for the high risk patients remained similar. It appeared that the previous low cancer outcomes in the audit data reassured the low risk patients that a biopsy was not necessary. The reduction in biopsies in low-risk groups saves these patients the potential complications and anxiety associated with the investigation. It also reduces cost to the healthcare provider; allowing greater proportion of departmental capacity and resource to be concentrated on the management of high-risk lesions. Conversely, patients with a high Likert score would have had the benefits of prostate biopsy highlighted by the data and hence retained high uptake.

Ultimately, even when accounting for “COVID effect” by looking at contemporaneous reports, proportionally fewer biopsies were carried out overall in the n = 207 cohort which included audit information vs the n = 160 cohort which did not. Again, those patients spared biopsy were more likely to have a low Likert score. This suggests that for this reporter, inclusion of the audit data reduced the unnecessary biopsies in the low-risk groups, but did not significantly affect the uptake of biopsies in the highest risk group. Although this general trend may be transferrable to other centres, the exact impact on biopsy uptake will vary for the individual reporter concerned. There is scope for future work involving more than one radiologist to help quantify this variability and to further demonstrate the known inter-reporter variability in both Likert and PI-RAD scores. This was not undertaken in this study which aimed to assess the effect of a single reporter including their individual audit results in their reports.

Opinion of referring clinicians

The impression of the counselling clinicians was that the audit information was useful (100%) and gave them greater confidence to advise the patient when they did not require a biopsy (Likert 1 or 2: 67%, Likert 3: 50%). This was borne out in the results which show a markedly lower biopsy uptake in Likert 3 patients in the n = 207 cohort with the audit data (42.9%), vs the historical n = 791 cohort (78.4%) or the contemporaneous n = 160 cohort (65.2%) without audit information.

Suggestions for improvement were sought. The feedback was overwhelmingly positive, which suggests other clinicians are also likely to welcome these changes. However, there were two comments suggesting improvement. We intend to seek ongoing feedback, ensuring any change in reports is useful for the clinical team and patients at which it is targeted.

Limitations

This is a single reporter study by design, as the intention was to provide individualised reporter specific audit results. When using other reporters in the second control group, the cases were not randomised and the contemporaneous reporters had different levels of experience and reporting numbers. In addition, there was concern that once the counselling clinician had seen the audit data in the author’s reports, they might apply that knowledge to the counselling of subsequent patients even if their reports contained no such data.

Conclusions

Personal audit is a useful part of every conscientious doctor’s practice, but often this time-consuming work cannot be demonstrated to have had a direct beneficial effect on patient care. Auditing reporter-specific histological outcomes for query prostate cancer mpMRI scans has previously been recommended due to inter-reporter variability, and this study shows that adding these results to subsequent reports is seen as helpful by counselling clinicians, and can spare low risk patients unnecessary biopsies. This simple intervention could be put in place by any individual radiologist in any department reporting a sufficient volume. By doing this, the radiologist can deliver a more individualised and meaningful assessment of the risk of clinically significant prostate cancer to the referring team and patients.