Surgical margins for borderline and malignant phyllodes tumours

Abstract

Background

Phyllodes tumours represent less than 1% of all UK breast neoplasms. Histological features allow classification into benign, borderline or malignant, which has a significant impact on prognosis and recurrence. Currently, there is no consensus for the optimal surgical excision margin. This systematic review aims to provide a comparative summary of outcomes (local recurrence, metastasis and survival) for borderline and malignant phyllodes tumours resected with either ≥1cm or <1cm margins.

Methods

MEDLINE and Embase were systematically searched (1990 to July 2019), in line with PRISMA guidelines. Study quality was assessed using the Newcastle–Ottawa scale.

Results

Ten retrospective studies were included (Newcastle–Ottawa scale mean score: 5.6, range: 8–4). Nine reported local recurrence rates, four reported distant metastasis and four reported survival. Meta-analysis pooling demonstrated no statistically significant difference between <1cm and ≥1cm margins in terms of local recurrence rates (relative risk [RR] 1.43, 95% confidence interval [95% CI] 0.70 – 2.93; p=0.33, n=456), distant metastasis (RR 1.93, 95% CI 0.35 – 10.63; p=0.45, n=72) or mortality (RR 1.93, 95% CI 0.42 – 8.77; p=0.40, n=58) for borderline and malignant tumours. Additionally, two studies demonstrated no significant difference in local recurrence for borderline tumours excised with <0.1cm margins compared to ≥1cm.

Conclusion

Current evidence suggests that margins <1cm may provide adequate tumour excision. This could enable breast conservation in patients with smaller breast-to-tumour volume ratios, with improved cosmetic outcomes and patient satisfaction. A prospective, multi-institutional trial would be appropriate to further elucidate the safety of smaller margins.

Introduction

Phyllodes tumours are uncommon fibroepithelial neoplasms of the breast. They make up less than 1% of all female breast neoplasms and typically affect women aged 35–55 years.^1,2 Their presentation is classically as a rapidly growing painless lump, sometimes in a lesion that had been stable for years.³ However, the presentation and behaviour of these tumours can vary greatly between subtypes (benign, borderline and malignant), which are defined by their histological features (Table 1).⁴ In particular, borderline and malignant subtypes exhibit an especially recurrent and metastatic nature, with a reported survival time for patients with malignant phyllodes tumours as low as two years.^5,6 Effective management guidelines to ensure adequate excision and optimal disease-free survival for these subtypes are therefore paramount.

Table 1.

Histological classification of phyllodes tumours.

	Benign	Borderline	Malignanta
Mitoses/10 HPF	<5	5–9	≥10
Tumour borders	Well defined	Well-defined, may be focally infiltrative	Infiltrative
Stromal atypia	None to mild	Mild to moderate	Marked
Stromal overgrowth	Absent	Absent or very focal	Often present
Stromal cellularity	Cellular, usually mild, may be non-uniform or diffuse	Cellular, usually moderate, may be non-uniform or diffuse	Cellular, usually marked or diffuse
Malignant heterologous stromal elements	Absent	Absent	May be present

^a For the diagnosis of a malignant phyllodes tumour, all of the malignant features should be met. Malignant tumours do not always have malignant heterologous stromal elements, but their presence is sufficient for a diagnosis of a malignant tumour, irrespective of other features.

Nevertheless, an optimal strategy for the surgical management of borderline and malignant phyllodes tumours remains a subject of debate. The National Comprehensive Cancer Network and MD Anderson guidelines recommend wide local excision of phyllodes tumours, aiming for margins 1cm or greater (≥1cm).^7,8 A margin of this size is considerably larger than the National Institute for Health and Care Excellence’s stated minimum margin for invasive or in situ breast cancer.⁹ Several studies have questioned the need for such large margins, suggesting that margins less than 1cm (<1cm) are sufficient, with no difference in local or distant recurrence.^10–12 However, these studies combined data for all three subtypes, with benign tumours forming majority of the numbers, and whilst a 2017 review by Shabaan and Barthelmes demonstrated no difference in outcomes when aiming for (but not histologically confirmed) margins 0.1cm compared with 1cm in benign phyllodes tumours, there is a clear paucity of data relating specifically to borderline and malignant subtypes.¹³

This systematic review summarises current research data comparing outcomes (in terms of local recurrence, distant metastasis, disease-free and overall survival) of women with borderline or malignant phyllodes tumours surgically resected with either <1cm or ≥1cm on postoperative histology. The findings could have significant implications on the surgical management of phyllodes tumours.

Methods

Search strategy and selection criteria

A systematic review protocol was developed in accordance with the Preferred Reporting for Items for Systematic Reviews and Meta-Analyses-Protocols (PRISMA-P) guidance,¹⁴ and registered with PROSPERO (reference: CRD42018115969). A title and abstract search of Embase and MEDLINE was conducted between January 1990 and July 2019. Search strategies were formulated using the medical search heading terms: phyllodes, margin, borderline, malignant, recurrence, survival and metastasis. Duplicates were removed and two authors conducted a screening of titles and abstracts independently, according to inclusion and exclusion criteria (Table 2).

Table 2.

Inclusion and exclusion criteria.

Inclusion	Exclusion
Studies published in English only  Outcomes (local recurrence, distant metastasis or survival rates) of surgical resection of phyllodes tumours with margin sizes ≥1cm or <1cm  Phyllodes tumour subtypes investigated include borderline or malignant types only	Margins not histologically confirmed postoperatively  Studies reporting outcomes exclusively for benign phyllodes tumours or mix benign phyllodes tumour data with malignant and borderline subtypesa  Case reports, case reviews below 10 patients, literature reviews

^a For studies that mixed benign phyllodes tumour data with borderline and malignant phyllodes tumours, if specifically borderline or malignant could not be extracted from the literature article and authors could not be contacted for raw data, the study was excluded.

Where the eligibility of a study could not be determined based on an abstract alone, the full text of the article was retrieved. For studies that included benign phyllodes tumour data, attempts were made to extract results of borderline and malignant phyllodes tumours from the article. If this was not possible, authors of these studies were contacted for raw data. If data relating strictly to borderline and malignant phyllodes tumours could not be obtained, the articles were excluded to ensure that data were not distorted by the presence of benign phyllodes tumours. Disagreements between reviewers were resolved by discussion.

After the initial screen, full texts of the remaining studies were retrieved and data extracted. Two authors independently extracted data from included studies using a standardised data extraction form developed for the purpose of this study. Demographic follow-up time and tumour characteristics (number of each tumour subtype, mean tumour size and mean age of cohort) were recorded. For analysis purposes, the study data were grouped according to reported outcomes of interest (local recurrence, distant metastasis, disease-free and overall survival) and tumour subtypes. In cases where data descriptors were not published, the authors were contacted for raw data. If data could not be obtained, p-values calculated by authors were recorded and these articles were excluded from the meta-analysis. Studies were assessed for bias using the Newcastle–Ottawa scale for non-randomised cohort studies.¹⁵ Publication bias was assessed using an inverted funnel plot.

Statistics

All statistical analysis was performed using Review Manager 5.3 software (Cochrane Collaboration). Results data extracted were used to calculate relative risk (RR) plus 95% confidence interval (CI) for discrete data and mean difference plus 95% CI for continuous data. Data were pooled with random effects analysis. Sensitivity analysis was conducted by excluding one study at a time from the meta-analysis, observing the effect on the pooled outcome. Statistical heterogeneity between studies was assessed using the Chi² test and quantified using the I² statistic. A p-value less than 0.05 was considered significant for statistical analyses.

Findings

The search identified 378 studies, including 126 duplicates (Fig 1). After abstract screening, 115 remained for full-text review. Of the 137 papers excluded, 59 were not relevant to the surgical management of phyllodes tumours (investigated the role of radiotherapy, imaging techniques, immunohistochemistry, histopathology or genetic features of phyllodes tumours), 41 case studies, 20 reviews, 9 further duplicates and 8 not written in English. Full-text review excluded an additional 105 papers (85 did not report margin sizes, 16 reported outcomes for benign tumours only or mixed benign tumour data with borderline and malignant results, 4 measured margins but did not report outcomes according to margin size to allow for comparison), leaving 10 articles for data extraction.^16–25

Figure 1.

PRISMA flowchart for the selection and exclusion of studies.

Study characteristics

Included studies were published between 2001 and 2019 from North America, Europe, the Middle East and Asia (Table 3). All studies were retrospective, with Borhani-Khomani et al²⁰ and Choi et al¹⁷ being multi-centre and the rest single-centre. The most commonly used grading criteria was the World Health Organization histological classification. Lin et al,²² Asoglu et al²⁴ and Kapiris et al²⁵ used alternative histological classification methods, while the remaining two (Ganesh et al¹⁹ and Borhani-Khomani et al²⁰) did not report the criteria by which subtypes were identified. Half of the included studies (Mitus et al,¹⁶ Sevinç et al,¹⁸ Onkendi et al,²¹ Lin et al²² and Kapiris et al²⁵) re-examined histological slides for reporting margin sizes, though none of which detailed the methodology behind the pathological assessment of margins, to allow comparison to current methods or guidelines. Borhani-Khomani et al²⁰ obtained margin histology data from the Danish Pathology Register database. The remaining four (Choi et al,¹⁷ Ganesh et al,¹⁹ Abdalla et al²³ and Asoglu et al²⁴) did not state how margin sizes were obtained.

Table 3.

Included studies.

Study	Study type and design	Country	Patients with PT (n)		Surgery types	Median follow-up in months (range)	Mean PT size in cm (range)	Mean age in years (range)
			Borderline	Malignant
Mitus 201915	Single-centre, retrospective cohort	Poland	40	113	BCS, M	Mean: 144 (n/a)	6 (2–40)	51 (19–89)
Choi 201816	Multi-centre, retrospective cohort	Korea	127	235	BCS, M	60 (24–372)	6 (0–38)	Median: 43 (13–75)
Sevinc 201817	Single-centre, retrospective cohort	Turkey	14	0	BCS	133 (36–240)	2 (1–4)	51 (n/a)
Ganesh 201818	Single-centre, retrospective cohort	Canada	17	53	BCS, M	Bor: 66 (1–130) Mal: 50 (2–131)	n/a	Bor: 49 (31–70) Mal: 50 (24–88)
Borhani-Khomani 201619	Multi-centre, retrospective cohort	Denmark	89	0	LE, WLE, M	Mean: 84 (1–180)	3 (1–20)	44 (18–85)
Onkendi 201420	Single-centre, retrospective cohort	USA	15	52	WLE, M	120 (12–408)	8 (n/a)	47 (14–86)
Lin 201321	Single-centre, retrospective cohort	Taiwan	13	12	M, PM	n/a	n/a	n/a
Abdalla 200622	Single-centre, retrospective cohort	Cairo	27	21	LE, WLE, M	n/a	n/a	n/a
Asoglu 200423	Single-centre, retrospective cohort	USA	3	31	WLE, M	n/a	n/a	n/a
Kapiris 200124	Single-centre, retrospective cohort	UK	0	48	LE, WLE, M	108 (5–336)	8 (2–20)	47 (21–85)

BCS, breast conserving surgery; Bor, Borderline phyllodes tumour; LE, local excision; M, mastectomy; Mal, Malignant phyllodes tumour; n/a, data not available; PM, partial mastectomy; PT, phyllodes tumour; WLE, wide local excision.

Follow-up, mean tumour size and mean patient age relating specifically to borderline and malignant tumours only could not be obtained from Lin et al,²² Abdalla et al²³ and Asoglu et al,²⁴ as these authors included benign tumour data that could not be separated from borderline and malignant tumour data in their reporting of demographics and study characteristics.

Studies were evaluated for quality using the Newcastle–Ottawa scale (Table 4), demonstrating a moderate risk of bias in all studies included (mean score 5.6, range 8–4). The majority of the studies lost points on three common categories: 1) failure to demonstrate that the outcome was not present at the start of the study, for example, the presence of distant metastasis at the time of diagnosis; 2) no comparability of cohorts on the basis of the design or no controls for confounders in the study analysis; and 3) not stating the completeness of patient follow-up.

Table 4.

Newcastle–Ottawa scale (modified for cohort studies) for assessment of quality of included studies.

Study	Selection score	Comparability score	Outcome score	Total score
Mitus 201915	***		***	6
Choi 201816	****		**	6
Sevinc 201817	***		**	5
Ganesh 201818	**	*	*	4
Borhani-Khomani 201619	***	*	***	8
Onkendi 201420	***		**	5
Lin 201321	***	*	**	6
Abdalla 200622	***		**	5
Asoglu 200423	***		**	5
Kapiris 200124	**	*	***	6

An asterisk (*) denotes fulfilment of each criterion.

Local recurrence

There were nine studies that investigated local recurrence rates for margins ≥1cm and <1cm (Fig 2), including Asoglu et al,²⁴ who only reported p-values. Asoglu et al²⁴ and Kapiris et al²⁵ were the only authors to report superior local control of margins ≥1cm with statistical significance (p=0.01 and p=0.02, respectively) for malignant phyllodes tumours. Pooling of data demonstrated no statistically significant difference in recurrence rates between <1cm and ≥1cm margins for both borderline and malignant subtypes combined or when analysed separately (Fig 2).

Figure 2.

Pooled results of local recurrence rates for borderline and malignant phyllodes tumours (PTs) excised with less than 1cm compared with 1cm or greater margins. An asterisk (*) denotes authors with statistically significant data.

Exclusion of Lin et al,²² as part of the sensitivity analysis, achieved statistical significance in favour of ≥1cm margins in terms of local recurrence rates of both subtypes combined (RR 2.02, 95% CI 1.19–3.42; p=0.009, n=431) and malignant phyllodes tumours alone (RR 2.3, 95% CI 1.13–4.68; p=0.02, n=32) but not borderline phyllodes tumours. Exclusion of other studies did not change the outcome of pooled analysis. There was evidence of moderate heterogeneity between the local recurrence rates of malignant phyllodes tumours only (Chi² 8.17, p=0.04, I² 63%). However, there was no evidence of significant heterogeneity for the pooled results of both subtypes together and borderline phyllodes tumours only (Chi² 10.1, p=0.12 and Chi² 2.4, p=0.49, respectively). No significant publication bias was demonstrated by the funnel plot analysis (Fig 3), although there are a limited number of studies.

Figure 3.

Funnel plot for assessment of publication bias.

For smaller resection margins, Sevinç et al¹⁸ and Borhani-Khomani et al²⁰ reported local recurrence rates of borderline phyllodes tumours with resection margins of <0.1cm. Neither reported any statistically significant data when compared with margins ≥1cm (p=0.61 and 0.15, respectively).

Distant metastasis

All four studies that reported outcomes in terms of distant metastasis (Asoglu et al,²⁴ Abdalla et al,²³ Sevinç et al¹⁸ and Kapiris et al²⁵) did not identify any statistically significant difference for borderline or malignant phyllodes tumours resected with a margin <1cm compared with a margin ≥1cm (Fig 4). Data descriptors of metastasis rates were not available from Asoglu et al,²⁴ but the authors reported a p-value of 0.36 for borderline and malignant tumours. No statistically significant difference was identified from the pooled data, which excluded the findings of Asoglu et al.²⁴ In addition, there was no evidence of statistically significant heterogeneity between the studies (Chi² 1.89, p=0.17).

Figure 4.

Pooled results of distant metastasis and mortality rates for borderline and malignant phyllodes tumours (PTs) excised with less than 1cm compared with 1cm or greater margins. An asterisk (*) denotes authors with statistically significant data.

Overall and disease-free survival

Both Asoglu et al²⁴ and Kapiris et al²⁵ identified worse overall survival in borderline and malignant phyllodes tumours resected with <1cm margins (p=0.03 and 0.013, respectively), while data from Onkendi and Abdalla et al were not statistically significant for these subtypes. The combined data did not reveal any statistical difference in terms of overall survival (Fig 4). There was no evidence of significant study heterogeneity (Chi²=1.79, p=0.18).

Ganesh et al¹⁹ and Onkendi et al²¹ investigated the effect of margins size on disease-free survival. Data from Onkendi et al allowed comparison between <1cm and ≥1cm margins for borderline and malignant phyllodes tumours, reporting a p-value of 0.48. Smaller margins of <0.1cm and ≥0.1cm for malignant phyllodes tumours was compared by Ganesh et al,¹⁹ also not significant (p=0.42).

Discussion

Local recurrence, distant metastasis, overall and disease-free survival were compared for malignant and borderline phyllodes tumours excised with pathologically confirmed <1cm and ≥1cm. Our meta-analysis of current literature revealed that a larger surgical margin of 1cm or greater, compared with less than 1cm, does not confer any statistically significant advantage in terms of local control, distant metastasis or overall survival for borderline and malignant phyllodes tumours. Furthermore, Onkendi et al²¹ revealed no significant difference in disease-free survival.¹⁰ Two studies included in this systematic review even demonstrated no difference in local control between <0.1cm compared with ≥1cm for borderline tumours.^15,19

Together, these data suggest that the recommended margins of 1cm or greater may be redundant for the surgical management of malignant and borderline phyllodes tumours. These are undoubtedly important findings, as smaller margins lead to less breast deficit and deformity, and can make breast-conserving surgery technically possible. In addition, this can be of relevance in situations where it is difficult to achieve margins of 1cm or more¹⁰; for example, due to proximity of the chest wall or skin. What is more, the need for re-excisional surgery can be reduced as smaller clearance margins can be accepted, resulting in reduced cost, patient inconvenience and risk of surgical complications, as well as improved cosmetic outcomes.²⁶

One hypothesis is that histopathological margin status (representing the presence or absence of tumour cells at the edge of excised tissue on postoperative histology), rather than margin size, plays a key role in recurrence. Indeed, in a multicentre study of 362 patients with borderline or malignant phyllodes tumours by Choi et al,¹⁷ margin positivity (ie the presence of tumour cells at the edge of the excised tissue) was identified to be an independent risk factor for recurrence, while negative margins of <1cm was not.¹⁶ This may also account for the contradictory findings of Kapiris et al²⁵ and Asoglu et al,²⁴ who demonstrated poorer outcomes for margins <1cm. Both authors, unlike other studies included in this review, may have included positive margins in the <1cm margin group. Kapiris et al,²⁵ for example, defined margins as ‘involved’ (<1cm) or ‘not involved’ (≥1cm), but it was not reported how many of the ‘involved’ cases had positive margins.

It is possible that, provided that the margins are negative, the tumour margin size, no matter how small, may not be relevant. Drawing parallels to breast cancer, a 2014 meta-analysis by the Society of Surgical Oncology and American Society for Radiation Oncology revealed that margin negativity reduces the risk of local recurrence by twofold in stage I and II breast cancer, compared with positive margins.²⁷ Yet there was no evidence that wider negative margins reduced the rate of recurrence, regardless of factors such as patient age, tumour biology or the presence of an extensive intraductal component. The multidisciplinary consensus panel concluded that the presence of ‘no ink on tumour’ was sufficient as an adequate margin for invasive cancer. In spite of this, from a practical point of view, it would be logical to assume that less extensive surgery would result in an increased risk of positive margins; an important factor to consider for phyllodes tumours with infiltrative borders or nodules, which exhibit increased recurrence rates compared with pushing borders and tumours without nodules,^11,28 respectively, possibly due to the greater susceptibility of incomplete excision. Therefore, in the case of borderline or malignant phyllodes tumours resected with less than 1cm margins, and particularly if there are unfavourable tumour borders, it would be prudent to ensure a more detailed margin assessment and consideration of a wider margin prescription to ensure a negative margin status. The role of local radiotherapy may also be of benefit in these cases, however it is yet to be fully evaluated.^8,29,30

Despite the promising findings, one must proceed with caution before drawing firm conclusions from current literature data. At times, the histological distinction between benign, borderline and malignant phyllodes tumours can be hampered by interobserver variation.³¹ Consequently, the inevitable presence of lower-grade phyllodes tumours within the borderline and malignant subtype datasets could have introduced bias. In terms of the included studies, all were retrospective and non-randomised, with small sample sizes, due to the inherently rare nature of borderline and malignant subtypes. The possibility that larger sample sizes may have revealed significant differences cannot be excluded. Furthermore, several non-English papers were excluded and data descriptors from Asoglu et al²⁴ could not be obtained, which may have had an effect on the outcome of the meta-analysis. Adjuvant therapy, in most of the included studies, was either not used or was reserved for a small portion of patients with positive margins. In spite of this, for Choi et al¹⁷and Ganesh et al¹⁹ the use of adjuvant therapy was more frequent, with its use in 8.5% and 46% of cases, respectively.^16,17 The authors did not publish whether chemotherapy and radiotherapy use was comparable between <1cm and ≥1cm margins, and although the value of adjuvant therapy in phyllodes tumours has not been fully established, it could represent a potential confounder. Another possible issue is selection bias: less extensive surgery and thus smaller margins may have been reserved for less aggressive tumours with better prognostic indicators. Although no studies directly compared the prognostic factors between the two margin sizes to confirm this, there were several studies that had differences in tumour prognostic factors among surgical procedures. Onkendi et al,²¹ for example, identified that patients who underwent mastectomy were significantly more likely to harbour phyllodes tumours with adverse characteristics, compared with those who underwent breast-conserving surgery, which may account for the worse survival rates of these patients.²⁰ On the other hand, phyllodes tumours with a worse prognosis, such as large tumours,²⁴ may be at higher risk of achieving smaller margins, pushing the results for margins <1cm towards worse outcomes. Of course, this is merely speculation, but future studies should control for confounders to ensure an accurate understanding of the true effect of margin size on outcome.

Conclusion

The available evidence regarding whether surgical margins less than 1cm are adequate for the management of borderline and malignant phyllodes tumours is promising. Smaller margins offer the benefit of improved cosmetic outcomes and reduce the need for further surgery when pathological margins of 1cm or greater are not achieved. Evidence is clear for the relationship between positive margin status and adverse outcomes, so a wider excision should be considered if negative margins are not obtained. Last, while the role of adjuvant radiotherapy has yet to be determined, current literature indicates that it could help boost outcomes, at least in terms of local recurrence. Admittedly, data thus far are limited, owing to the inherent limitations of small single-centre retrospective studies, and it would be inappropriate to make definitive recommendations based on these data alone. A multi-institutional trial, with careful attention to the documentation of margin size and status, would certainly be warranted to further elucidate the safety of smaller tumour-free margins and their role in the development of local recurrence and distant metastases. Given the rarity of borderline and malignant subtypes, a UK-wide phyllodes tumour database of prospectively collected data would certainly be of benefit, similar to the national Danish Pathology Register, which has been updated daily since 1999.¹⁹