IJC International Journal of Cancer
Sentinel node biopsy performance after neoadjuvant chemotherapy in locally advanced breast cancer: A systematic review and meta-analysis Simone Mocellin, Elena Goldin, Alberto Marchet and Donato Nitti Surgery Branch, Department of Surgery, Oncology and Gastroenterology, University of Padova, Padova, Italy
Cancer Therapy and Prevention
The use of sentinel node biopsy (SNB) after neoadjuvant chemotherapy (NAC) in patients with locally advanced breast cancer is debated. Our aim was to quantitatively review the available evidence on the performance of SNB after NAC in patients with locally advanced breast cancer. We performed a systematic review (by searching the PubMed, Cochrane and Scopus databases) and random effects meta-analysis to investigate on the feasibility and accuracy of SNB in these patients. The two outcomes of interest were the sentinel node identification rate (SIR) and the false negative rate (FNR). Sensitivity analysis and meta-regression were used to investigate the potential sources of between-study heterogeneity. We retrieved 72 eligible studies enrolling 7,451 patients. Upon meta-analysis, summary SIR resulted 89.6% [95% confidence interval (CI): 87.8–91.2; heterogeneity I2: 76.9%], which poorly compares with the 95% SIR observed in some recent series of early breast cancer. The summary FNR resulted 14.2% (CI: 12.5–16.0; heterogeneity I2: 29.1%), which was significantly higher than the 8–10% reference value. Considering an average post-NAC lymph node positivity rate of 50%, the downstaging due to false negative SNB would occur in 7/100 patients (with an excess error rate of 2–3/100 as compared to the early-stage setting). No plausible source of between-study heterogeneity was found. Based on the largest series of studies ever meta-analyzed, our findings highlight the limits of SNB performance in this population, where the impact of SNB on patient survival is still to be defined.
Sentinel node biopsy (SNB) has revolutionized the surgical approach to axillary lymph nodes in patients with early breast cancer,1 and current international guidelines recommend no completion lymph node dissection not only in case of lack of metastatic disease but also in selected cases of positive sentinel node.2 While the use of SNB is well established in patients with early-stage breast cancer, no consensus exists on its role in patients with locally advanced disease.3–9 In the latter setting, the standard management of axillary lymph nodes remains radical lymph node dissection in most centers: this is due to the scattered evidence on SNB performance as well as to the lack of knowledge on the survival effects of a SNB-guided axillary node surgery in these patients. It has been hypothesized that colonization of lymphatic vessels by malignant cells during disease progression might alter the Key words: breast cancer, sentinel node biopsy, neoadjuvant chemotherapy, meta-analysis Additional Supporting Information may be found in the online version of this article. Grant sponsor: Cariparo Foundation (Italy) DOI: 10.1002/ijc.29644 History: Received 11 Apr 2015; Accepted 12 June 2015; Online 17 June 2015 Correspondence to: Simone Mocellin, Surgery Branch, Department of Surgery, Oncology and Gastroenterology, University of Padova, Via Giustiniani 2, 35128 Padova, Italy, Tel.: 139-049-8211851, Fax: 139-049-651891, E-mail: [email protected]
C 2015 UICC Int. J. Cancer: 138, 472–480 (2016) V
lymphatic flow: this would undermine the reliability of lymphatic mapping and ultimately the ability to identify the SN. The relatively recent implementation of neoadjuvant chemotherapy (NAC) might make the interpretation of these issues even more complex. NAC—although associated with survival outcomes comparable to those observed with adjuvant chemotherapy regimens—increases the rate of conservative surgery of the primary tumor (downstaging effect) and represents an in vivo chemosensitivity test (which helps modulating subsequent drug regimens).10–13 However, NAC might impair the performance of SNB in two ways: first, it might contribute to the above mentioned alteration of lymphatic flow by inducing tissue fibrosis; second, NAC might eliminate tumor deposits in the SN but not in non-SNs, which would ultimately lead to an increased rate of false negative cases. To provide readers with a comprehensive and quantitative overview of the existing evidence in this field, we performed a systematic review and meta-analysis of the available literature regarding the performance of SNB after NAC in patients with locally advanced breast cancer.
Material and Methods Search strategy
We performed a comprehensive search of the international literature to identify studies (either retrospective or prospective) that examined the ability of SNB to assess the axillary lymph node status in patients with locally advanced breast cancer after NAC. Locally advanced disease was defined by a
Mocellin et al.
473
What’s new? The use of sentinel node biopsy (SNB) to determine whether a tumor has spread into the lymphatic system has revolutionized the surgical management of early breast cancer. But whether SNB also benefits patients with locally advanced breast cancer undergoing neoadjuvant chemotherapy remains unknown. Here, meta-analysis of existing evidence shows that SNB performs less well and is less accurate in the setting of locally advanced disease. Potential misdiagnosis with SNB in this setting could result in incomplete lymph node dissection and incorrect downstaging of disease. More information is needed, however, to fully assess SNB in locally advanced disease.
C 2015 UICC Int. J. Cancer: 138, 472–480 (2016) V
aimed at identifying the risk of bias (classified as high, low or unclear) that would undermine the reliability of the evidence: 1. Patient selection: to verify that the spectrum of patients enrolled is representative of the patient population that will receive the index test in practice (selection bias). 2. Index test (in the present case: SNB): to assess if the index test results were interpreted without knowledge of the results of the reference standard, and if a positivity threshold (if any) was prespecified (and thus if the conduct or interpretation of the index test have introduced bias). 3. Reference test (pathology evaluation of the lymph node dissection following SNB): to assess if the reference standard utilized is likely to correctly classify the target condition (i.e., axilla lymph node status) and if the reference standard results were interpreted without knowledge of the results of the index test (and thus, if the reference standard, its conduct or its interpretation have introduced bias). 4. Flow and timing: to verify if there was an appropriate interval between index test and reference standard (to make sure the target condition did not change between the two tests), if all patients received the same reference standard (differential verification bias) and if all patients were included in the analysis (partial verification bias, reporting bias). Summary estimates of performance parameters were calculated by random effects meta-analysis using the inverse variance method. Meta-analysis included also evaluation of between-study heterogeneity, sensitivity analysis and examination for bias. Heterogeneity (true variance of effect size across studies) was quantified by means of I2 statistic [which indicates the percentage of the variability in effect estimates due to true heterogeneity rather than sampling error (within study variance)] and formally tested by means of the v2-based Cochran’s Q-test (to assess whether observed variance exceeds expected variance). Potential sources of heterogeneity were searched by random effects meta-regression; to this aim, the following covariates were considered: publication year, study design features (prospective vs. retrospective; monocentric vs. multicentric), patient mean age, primary tumor stage at presentation, clinical lymph node status before and after NAC, SIR (for FNR), pathological axillary lymph node positivity rate.
Cancer Therapy and Prevention
primary tumor not amenable to conservative surgery (due to unfavorable tumor/breast size ratio, as per surgeon’s judgment) and/or by clinically positive axillary lymph nodes. We systematically searched the PubMed, Cochrane and Scopus databases for full-text articles published until December 2014, using the following key words: “breast,” “cancer” (or “carcinoma”), “sentinel node biopsy,” “neoadjuvant” (or “pre-operative,” “induction”) and “chemotherapy.” No language restriction was applied. We searched for additional references by cross-checking bibliographies of retrieved full-text articles and by scrutinizing relevant review articles and meta-analyses. The systematic review was carried out using established methods14 and following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines.15 Inclusion of studies, data extraction and quality assessment were carried out independently by two authors (SM and EG) using dedicated data extraction forms. Disagreements were resolved by consensus and/or arbitration involving a third author (AM). A two-stage process was used: we first screened titles and abstracts; then, for all references categorized as “eligible” or “uncertain,” the full text was retrieved and final inclusion decisions were made after reading the full paper. The two performance parameters we considered were the sentinel node identification rate (SIR, defined as the number of cases with successfully identified SN over the number of SNB procedures attempted) and the false negative rate (FNR, defined as the number of false negative cases over the number of cases with positive axilla at the SNB and/or the completion lymph node dissection). To define the latter, SNB must be followed by completion lymph node dissection, so to define the false negative cases (i.e., patients with negative SNB but with one or more metastatic lymph nodes at completion lymph node dissection). A case was classified as positive (metastatic axilla) when the pathological examination of the SN and/or of the lymph nodes harvested at completion lymph node dissection revealed metastatic disease. Studies enrolling fewer than ten patients were excluded. We also excluded studies with possible overlap with the selected series (duplicate studies). Each study quality was assessed using the QUADAS-2 tool, a well-established guideline for studies on diagnostic test accuracy.16 This tool consists of the following four domains
474
The extent to which the combined risk estimate might be affected by individual studies (leading study effect) was assessed by consecutively omitting every study from the meta-analysis (leave-one-out sensitivity analysis). Sensitivity analysis was also used to verify the impact of low-quality studies (if any) on summary estimates. Funnel plots were used to detect the so-called small study effect. Publication and selection biases in meta-analysis are more likely to affect small studies, which also tend to be of lower methodological quality: this may lead to small study effect, where the smaller studies in a meta-analysis show larger treatment effects. Funnel plot asymmetry was formally investigated with the Egger linear regression approach. The trim-and-fill method was adopted to adjust summary estimates for potentially missing (unpublished) studies.17 We formally compared summary estimates of SIR and FNR obtained with the meta-analysis to reference values (90– 95% and 8–10%, respectively) derived from the international experience with SNB in patients with early-stage breast cancer1–4: to this aim, we considered the summary estimate significantly different from the reference value if the 95% confidence interval (CI) of the former does not cross the latter. The difference between summary values obtained in two subgroups was formally tested by means of the Cochran Q-test. All analyses were performed using the Stata SE 11.2 software (Stata Corp LP, College Station, TX).
Results
Cancer Therapy and Prevention
Characteristics of included studies
Literature search yielded 410 articles. Screening by title and abstract reading led to the identification of 112 potentially relevant articles whose full text was retrieved: ultimately, 72 studies published between 2000 and 2014 met the inclusion criteria and thus were included for both qualitative and quantitative analyses (Fig. 1). Overall, 7451 patients were enrolled, with a mean of 103 patients per study (range: 14– 689). The main characteristics of these studies are presented in Tables 1 and 2. With regard to study quality, no high risk of bias was detected in any of the included study. SNB performance
SIR was investigated in 71 studies (n 5 7,368) and ranged between 58.8 and 100%: upon meta-analysis, the summary SIR resulted 89.6% (95% CI: 87.8–91.2), which poorly compared with the 95% reference value. Between-study heterogeneity was high (I2: 76.9%) and statistically significant (Q-test p values:
Sentinel node biopsy performance after neoadjuvant chemotherapy in locally advanced breast cancer: A systematic review and meta-analysis Simone Mocellin, Elena Goldin, Alberto Marchet and Donato Nitti Surgery Branch, Department of Surgery, Oncology and Gastroenterology, University of Padova, Padova, Italy
Cancer Therapy and Prevention
The use of sentinel node biopsy (SNB) after neoadjuvant chemotherapy (NAC) in patients with locally advanced breast cancer is debated. Our aim was to quantitatively review the available evidence on the performance of SNB after NAC in patients with locally advanced breast cancer. We performed a systematic review (by searching the PubMed, Cochrane and Scopus databases) and random effects meta-analysis to investigate on the feasibility and accuracy of SNB in these patients. The two outcomes of interest were the sentinel node identification rate (SIR) and the false negative rate (FNR). Sensitivity analysis and meta-regression were used to investigate the potential sources of between-study heterogeneity. We retrieved 72 eligible studies enrolling 7,451 patients. Upon meta-analysis, summary SIR resulted 89.6% [95% confidence interval (CI): 87.8–91.2; heterogeneity I2: 76.9%], which poorly compares with the 95% SIR observed in some recent series of early breast cancer. The summary FNR resulted 14.2% (CI: 12.5–16.0; heterogeneity I2: 29.1%), which was significantly higher than the 8–10% reference value. Considering an average post-NAC lymph node positivity rate of 50%, the downstaging due to false negative SNB would occur in 7/100 patients (with an excess error rate of 2–3/100 as compared to the early-stage setting). No plausible source of between-study heterogeneity was found. Based on the largest series of studies ever meta-analyzed, our findings highlight the limits of SNB performance in this population, where the impact of SNB on patient survival is still to be defined.
Sentinel node biopsy (SNB) has revolutionized the surgical approach to axillary lymph nodes in patients with early breast cancer,1 and current international guidelines recommend no completion lymph node dissection not only in case of lack of metastatic disease but also in selected cases of positive sentinel node.2 While the use of SNB is well established in patients with early-stage breast cancer, no consensus exists on its role in patients with locally advanced disease.3–9 In the latter setting, the standard management of axillary lymph nodes remains radical lymph node dissection in most centers: this is due to the scattered evidence on SNB performance as well as to the lack of knowledge on the survival effects of a SNB-guided axillary node surgery in these patients. It has been hypothesized that colonization of lymphatic vessels by malignant cells during disease progression might alter the Key words: breast cancer, sentinel node biopsy, neoadjuvant chemotherapy, meta-analysis Additional Supporting Information may be found in the online version of this article. Grant sponsor: Cariparo Foundation (Italy) DOI: 10.1002/ijc.29644 History: Received 11 Apr 2015; Accepted 12 June 2015; Online 17 June 2015 Correspondence to: Simone Mocellin, Surgery Branch, Department of Surgery, Oncology and Gastroenterology, University of Padova, Via Giustiniani 2, 35128 Padova, Italy, Tel.: 139-049-8211851, Fax: 139-049-651891, E-mail: [email protected]
C 2015 UICC Int. J. Cancer: 138, 472–480 (2016) V
lymphatic flow: this would undermine the reliability of lymphatic mapping and ultimately the ability to identify the SN. The relatively recent implementation of neoadjuvant chemotherapy (NAC) might make the interpretation of these issues even more complex. NAC—although associated with survival outcomes comparable to those observed with adjuvant chemotherapy regimens—increases the rate of conservative surgery of the primary tumor (downstaging effect) and represents an in vivo chemosensitivity test (which helps modulating subsequent drug regimens).10–13 However, NAC might impair the performance of SNB in two ways: first, it might contribute to the above mentioned alteration of lymphatic flow by inducing tissue fibrosis; second, NAC might eliminate tumor deposits in the SN but not in non-SNs, which would ultimately lead to an increased rate of false negative cases. To provide readers with a comprehensive and quantitative overview of the existing evidence in this field, we performed a systematic review and meta-analysis of the available literature regarding the performance of SNB after NAC in patients with locally advanced breast cancer.
Material and Methods Search strategy
We performed a comprehensive search of the international literature to identify studies (either retrospective or prospective) that examined the ability of SNB to assess the axillary lymph node status in patients with locally advanced breast cancer after NAC. Locally advanced disease was defined by a
Mocellin et al.
473
What’s new? The use of sentinel node biopsy (SNB) to determine whether a tumor has spread into the lymphatic system has revolutionized the surgical management of early breast cancer. But whether SNB also benefits patients with locally advanced breast cancer undergoing neoadjuvant chemotherapy remains unknown. Here, meta-analysis of existing evidence shows that SNB performs less well and is less accurate in the setting of locally advanced disease. Potential misdiagnosis with SNB in this setting could result in incomplete lymph node dissection and incorrect downstaging of disease. More information is needed, however, to fully assess SNB in locally advanced disease.
C 2015 UICC Int. J. Cancer: 138, 472–480 (2016) V
aimed at identifying the risk of bias (classified as high, low or unclear) that would undermine the reliability of the evidence: 1. Patient selection: to verify that the spectrum of patients enrolled is representative of the patient population that will receive the index test in practice (selection bias). 2. Index test (in the present case: SNB): to assess if the index test results were interpreted without knowledge of the results of the reference standard, and if a positivity threshold (if any) was prespecified (and thus if the conduct or interpretation of the index test have introduced bias). 3. Reference test (pathology evaluation of the lymph node dissection following SNB): to assess if the reference standard utilized is likely to correctly classify the target condition (i.e., axilla lymph node status) and if the reference standard results were interpreted without knowledge of the results of the index test (and thus, if the reference standard, its conduct or its interpretation have introduced bias). 4. Flow and timing: to verify if there was an appropriate interval between index test and reference standard (to make sure the target condition did not change between the two tests), if all patients received the same reference standard (differential verification bias) and if all patients were included in the analysis (partial verification bias, reporting bias). Summary estimates of performance parameters were calculated by random effects meta-analysis using the inverse variance method. Meta-analysis included also evaluation of between-study heterogeneity, sensitivity analysis and examination for bias. Heterogeneity (true variance of effect size across studies) was quantified by means of I2 statistic [which indicates the percentage of the variability in effect estimates due to true heterogeneity rather than sampling error (within study variance)] and formally tested by means of the v2-based Cochran’s Q-test (to assess whether observed variance exceeds expected variance). Potential sources of heterogeneity were searched by random effects meta-regression; to this aim, the following covariates were considered: publication year, study design features (prospective vs. retrospective; monocentric vs. multicentric), patient mean age, primary tumor stage at presentation, clinical lymph node status before and after NAC, SIR (for FNR), pathological axillary lymph node positivity rate.
Cancer Therapy and Prevention
primary tumor not amenable to conservative surgery (due to unfavorable tumor/breast size ratio, as per surgeon’s judgment) and/or by clinically positive axillary lymph nodes. We systematically searched the PubMed, Cochrane and Scopus databases for full-text articles published until December 2014, using the following key words: “breast,” “cancer” (or “carcinoma”), “sentinel node biopsy,” “neoadjuvant” (or “pre-operative,” “induction”) and “chemotherapy.” No language restriction was applied. We searched for additional references by cross-checking bibliographies of retrieved full-text articles and by scrutinizing relevant review articles and meta-analyses. The systematic review was carried out using established methods14 and following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines.15 Inclusion of studies, data extraction and quality assessment were carried out independently by two authors (SM and EG) using dedicated data extraction forms. Disagreements were resolved by consensus and/or arbitration involving a third author (AM). A two-stage process was used: we first screened titles and abstracts; then, for all references categorized as “eligible” or “uncertain,” the full text was retrieved and final inclusion decisions were made after reading the full paper. The two performance parameters we considered were the sentinel node identification rate (SIR, defined as the number of cases with successfully identified SN over the number of SNB procedures attempted) and the false negative rate (FNR, defined as the number of false negative cases over the number of cases with positive axilla at the SNB and/or the completion lymph node dissection). To define the latter, SNB must be followed by completion lymph node dissection, so to define the false negative cases (i.e., patients with negative SNB but with one or more metastatic lymph nodes at completion lymph node dissection). A case was classified as positive (metastatic axilla) when the pathological examination of the SN and/or of the lymph nodes harvested at completion lymph node dissection revealed metastatic disease. Studies enrolling fewer than ten patients were excluded. We also excluded studies with possible overlap with the selected series (duplicate studies). Each study quality was assessed using the QUADAS-2 tool, a well-established guideline for studies on diagnostic test accuracy.16 This tool consists of the following four domains
474
The extent to which the combined risk estimate might be affected by individual studies (leading study effect) was assessed by consecutively omitting every study from the meta-analysis (leave-one-out sensitivity analysis). Sensitivity analysis was also used to verify the impact of low-quality studies (if any) on summary estimates. Funnel plots were used to detect the so-called small study effect. Publication and selection biases in meta-analysis are more likely to affect small studies, which also tend to be of lower methodological quality: this may lead to small study effect, where the smaller studies in a meta-analysis show larger treatment effects. Funnel plot asymmetry was formally investigated with the Egger linear regression approach. The trim-and-fill method was adopted to adjust summary estimates for potentially missing (unpublished) studies.17 We formally compared summary estimates of SIR and FNR obtained with the meta-analysis to reference values (90– 95% and 8–10%, respectively) derived from the international experience with SNB in patients with early-stage breast cancer1–4: to this aim, we considered the summary estimate significantly different from the reference value if the 95% confidence interval (CI) of the former does not cross the latter. The difference between summary values obtained in two subgroups was formally tested by means of the Cochran Q-test. All analyses were performed using the Stata SE 11.2 software (Stata Corp LP, College Station, TX).
Results
Cancer Therapy and Prevention
Characteristics of included studies
Literature search yielded 410 articles. Screening by title and abstract reading led to the identification of 112 potentially relevant articles whose full text was retrieved: ultimately, 72 studies published between 2000 and 2014 met the inclusion criteria and thus were included for both qualitative and quantitative analyses (Fig. 1). Overall, 7451 patients were enrolled, with a mean of 103 patients per study (range: 14– 689). The main characteristics of these studies are presented in Tables 1 and 2. With regard to study quality, no high risk of bias was detected in any of the included study. SNB performance
SIR was investigated in 71 studies (n 5 7,368) and ranged between 58.8 and 100%: upon meta-analysis, the summary SIR resulted 89.6% (95% CI: 87.8–91.2), which poorly compared with the 95% reference value. Between-study heterogeneity was high (I2: 76.9%) and statistically significant (Q-test p values:
