Key Findings and Assessment of the Strength of Evidence

In this update of the 2009 Comparative Effectiveness Review on breast biopsy methods we synthesized evidence from a total of 316 studies (128 new studies and 188 from the original report). We found few studies providing information on the test performance of open surgical biopsy. In contrast, the evidence base on core needle biopsy methods now includes a large number of studies reporting on almost 70,000 breast lesions. This allowed us to assess the comparative performance of tests (when using the same type of imaging guidance), in addition to updating the 2009 report’s evaluation of the performance of individual biopsy methods. Tables E–G summarize our assessment of the strength of evidence. Following the original evidence report, and in view of the paucity of evidence on open surgical biopsy, we refrained from rating the strength of evidence for this technique for all Key Questions. For Key Questions 1 and 2, we assessed the strength of evidence by integrating our (subjective) judgments on the risk of bias of included studies, the consistency of their findings, the directness of the available data, and the precision of quantitative results. For Key Question 3 we only rated the strength of evidence for the outcome of additional surgical procedures required after biopsy. We did not rate the strength of evidence for other Key Question 3 outcomes because of the diversity of designs employed and outcomes addressed (see the Methods section for our approach to rating the strength of evidence). Interested readers should consult Appendix D for the detailed assessment of the strength of evidence.

Test Performance and Comparative Test Performance

Among women at average risk of cancer, core needle biopsy using ultrasound or stereotactic guidance had average sensitivities ranging from 0.97 to 0.99 and average specificities ranging from 0.92 to 0.98. Freehand biopsy methods appeared to have lower average sensitivity (0.91) compared to other methods, but similar specificity (0.98). Stereotactically guided automated techniques were associated with lower sensitivity and higher specificity compared to stereotactically guided vacuum-assisted methods. Although these results were fairly precise, they were derived from indirect comparisons across studies of moderate to high risk of bias. MRI-guided biopsies were evaluated in only six studies with small sample sizes, leading to substantial uncertainty around estimates of test performance. Table 13 summarizes our assessment of the strength of evidence for alternative biopsy methods in women at average risk of cancer and for comparisons among biopsy methods using the same imaging guidance modality. Of note, we rated the strength of evidence on both absolute and comparative test performance, whereas the original report considered absolute test performance only.

Table 13

Strength of evidence about comparative test performance in women at average risk of breast cancer.

We did not find a difference in test performance between women at low and high risk of breast cancer. Because the number of studies of women at high risk of cancer was small, comparisons of test performance between low and high risk women had substantial uncertainty and results were not sufficient to support definitive conclusions. Evidence on modifiers of test performance was also sparse for all biopsy methods, raising concerns about selective outcome and analysis reporting.

Underestimation Rates

Underestimation rates varied among alternative biopsy methods and were often imprecisely estimated because of the relatively small number of lesions contributing data for these analyses. In general, underestimation was less common with stereotactically guided vacuum-assisted biopsy methods, as compared to stereotactically or ultrasound-guided automated methods. Our assessment of the strength of evidence for this outcome is summarized in Table 14.

Table 14

Strength of evidence for underestimation rates in women at average risk of cancer.

Adverse Events and Additional Surgeries After Biopsy

In general, adverse events were reported inconsistently, raising concerns about selective outcome and analysis reporting. Few studies provided information on the harms of open surgical biopsy. Core needle biopsy was only infrequently associated with serious adverse events or adverse events requiring additional treatment. Comparisons between open and core needle biopsy are based on indirect comparisons and expert opinion, with limited empirical evidence. Open biopsy appeared to be associated with an increased incidence of adverse events (including serious adverse events) compared to core needle biopsy. Our assessment of the strength of evidence for adverse events is summarized in Table 15.

Table 15

Strength of evidence assessment for adverse events of biopsy.

Among core needle biopsy methods, vacuum-assisted methods appeared to be associated with increased bleeding and hematoma formation. Sitting upright during the biopsy procedure was associated with more vasovagal reactions. Information about the dissemination or displacement of cancer cells during the biopsy procedure was provided by a small number of studies with various designs. Cancer cell seeding along the needle tract was a rare outcome. Studies reported that women were generally satisfied with the cosmetic results of core needle procedures.

Women diagnosed with breast cancer by core needle biopsy were able to have their cancer treated with a single surgical procedure, more often than women diagnosed by open surgical biopsy. Although the magnitude of this association was large (the ratio of the odds was almost 15), women and their physicians are likely to choose biopsy methods on the basis of factors (e.g., lesion location, or characteristics of the lesion on imaging) that may also be associated with the need for additional surgeries. Thus, confounding by indication is likely, and we rated the strength of evidence for this association as moderate. A difference in the rate of additional surgeries among women diagnosed with alternative biopsy methods is likely, but we have less confidence that it is an effect of the biopsy methods per se or that the magnitude of the difference is known.

Limitations of the Evidence Base

We believe that the evidence regarding the performance of core needle biopsy for diagnosis of breast lesions is limited in the following ways:

• Published evidence on the test performance and adverse events of open surgical biopsy was sparse.
• Available studies, particularly for Key Questions 1 and 2, were at moderate to high risk of bias and the publications we reviewed did not follow the Standards for Reporting of Diagnostic Accuracy (STARD) guidelines.³⁵¹ Information on patient selection criteria, patient or lesion characteristics (e.g., granular reporting of pathology results), was often missing or inconsistently reported. Information on adverse events and patient-relevant outcomes was often incomplete, potentially selectively reported. Studies did not use standardized definitions and ascertainment methods for adverse events. Pathology results were not reported with adequate granularity in the majority of cases.
• Studies typically used lesions (or biopsy procedures) as the unit of analysis, instead of patients. This way, patients with multiple lesions contributed multiple observations to the analyses. Lesions belonging to the same patient are likely to have similar characteristics (i.e. they are correlated). Unfortunately, studies reported results in a way that did not allow for the correlation to be accounted for in our statistical models. As such, our analyses (and those of the original report) assume independence among lesions. If the correlation among lesions in the same patient is high (positive and close to one) individual study and meta-analytic results will underestimate uncertainty and may also be biased (the direction of bias is unpredictable). However, unless each patient contributes large numbers of lesions that are highly correlated, the underestimation of uncertainty will not be large. Further, bias is unlikely unless patients contributing large numbers of lesions also have lesions that are substantially harder (or easier) to diagnose compared to those of other patients. Without additional data on the test performance on individual lesions within patients it is not possible to ascertain the impact of this factors on our results.
• Studies provided limited information to assess the impact of various patient-, lesion-, procedure-, or system- related factors on the outcomes of breast biopsy. For example, the impact of patient age, breast density, lesion type, training and experience of the operators, and error rates of pathologists who read the samples, on test performance, adverse events, or clinical outcomes could not be assessed.
• We found very few studies on MRI-guided biopsy for women at average or high risk of cancer. Because MRI-guided biopsy is likely reserved for diagnostically challenging cases (e.g., when lesions cannot be visualized by other modalities) and may be available in specialized care settings indirect (i.e. across studies) comparisons between MRI-guided and other biopsy procedures may be confounded by factors unrelated to the diagnostic value of the tests compared.
• There is limited information on the comparative effectiveness of alternative biopsy methods on patient-relevant outcomes, resource use and logistics, and availability of technology and expertise for different core needle biopsy techniques.

Strengths and Limitations of This Review

We conducted an up-to-date review of the benefits and risks of breast biopsy methods for breast cancer diagnosis, with respect to test performance, underestimation rates, adverse events, and patient-relevant outcomes. Previous reviews on this topic have focused on special patient populations (e.g., patients with nonpalpable lesions), selected outcomes (e.g. DCIS underestimation³⁵² or seeding³⁵³), or biopsy methods (e.g., ultrasound-guided biopsy³⁵⁴). Nonetheless, our work has several limitations, which – to a large extent – reflect the limitations of the underlying evidence base. Studies were deemed to be of moderate to high risk of bias because of characteristics related to their design and conduct, limiting our ability to draw strong conclusions. Information for several outcomes of interest was not reported from all available studies (e.g., underestimation rates, adverse events) raising concerns about selective outcome and analysis reporting. Information on study- or population level characteristics that could be modifiers of test performance, adverse events, or clinical outcomes, was inadequate. Thus, our ability to explore between-study heterogeneity was limited. Further, because we relied on published information and did not have access to individual patient data, we were unable to evaluate the impact of patient- or lesion-level factors on outcomes of interest.

The reference standard in the reviewed studies was a combination of clinical followup and pathologic confirmation (following open biopsy or excisional surgery). We assumed that these diagnostic methods have negligible measurement error (i.e., that they represent a “gold” standard). It is unlikely that this assumption is exactly true (e.g., some degree of diagnostic error is possible for pathologic examination, and clinical followup may provide less than perfectly accurate information). However, we believe that the error rate of the reference standard is low enough that its influence on our estimates is unlikely to be substantial.

Applicability of Review Findings

The existing evidence base on core needle biopsy of breast lesions in women at average risk of cancer appears to be applicable to clinical practice in the United States. Studies enrolled patients with an average age similar to that of women undergoing breast biopsy in the United States, and for indications that represent the most prevalent indications in U.S. clinical practice (i.e. mammographic findings of suspicious lesions). While fewer than half of the studies in this review were conducted in the United States, almost all were carried out in either the United States or in industrialized European or Asian countries where core-biopsy methods are likely sufficiently similar to those used in the United States. However, the applicability of our findings to women at high risk of breast cancer may be limited because we found few studies explicitly reporting on groups of patients at high baseline risk of breast cancer on the basis of factors such as genetic testing, or family history of disease. Of note, this may be an instance of incomplete reporting rather than a true characterization of the baseline risk of included populations (i.e. some high risk populations may have been misclassified as “average risk”).

Evidence Gaps and Ongoing Research

Table 16 summarizes the evidence gaps with regards to the Key Questions of diagnostic test performance and adverse events. A search on ClinicalTrials.gov for randomized trials comparing alternative biopsy methods did not identify trials examining biopsy techniques for breast cancer diagnosis (last search: Dec 5, 2013; 141 records retrieved).

Table 16

Evidence gaps for biopsy methods for the diagnosis of breast cancer.

Future Research Needs

• Studies of test performance are needed to evaluate MRI-guided biopsy methods. Ideally, these studies will be large (powered to achieve adequate precision), prospectively designed, multicenter investigations enrolling patients representative of those seen in clinical practice. Patient selection criteria and the characteristics of included populations should be reported in detail. Studies should use standardized histological classification systems for pathological classification the specialty and experience of those performing the biopsy procedure should be reported. The reference standard for test negative cases should be regular monitoring for an adequate period of time (e.g., 2 years).
• Although a large number of studies were available for other core needle biopsy methods we believe that additional well-designed and fully reported prospective cohort studies are needed, primarily for addressing questions about the impact of patient-, lesion-, procedure-, or system-level factors on test performance, adverse events, and patient-relevant outcomes. Given that a large number of core needle biopsies are performed annually in diverse settings, such studies could be conducted at relatively low cost.
• Large-scale databases of prospectively-collected observational data on breast biopsy procedures and outcomes could be used to evaluate the effectiveness of alternative biopsy methods with respect to short and long term outcomes, and potential modifying factors. Such studies would need to collect detailed information on baseline factors that may be associated with both the choice of biopsy method and the outcomes of interest (e.g., lesion size, palpability, imaging characteristics, etc.), to adjust for potential confounding factors. Comparisons across methods should be performed only among patients that would be candidates for assessment with all methods being compared.
• In all future studies, baseline risk of cancer development should be characterized using consistent and widely accepted criteria to allow appropriate subgroup analyses.
• We believe that a randomized comparison of alternative biopsy methods is unlikely to be fruitful because existing studies indicate that biopsy procedures have sensitivities and specificities that are fairly similar and close to 1. Under these conditions randomized trials comparing alternative biopsy methods would need to enroll very large numbers of participants to allow reliable comparisons between tests.
• Additional information is also needed to define what patient and lesion factors may correspond with accuracy or adverse events of specific techniques. Future research needs to be better reported for progress to be made on these questions.

Conclusions

A large body of evidence indicates that ultrasound and stereotactically guided core needle biopsy procedures have sensitivity and specificity close to that of open biopsy procedures, and are associated with fewer adverse events. The strength of the evidence on the test performance of these methods is deemed moderate because studies are at medium to high risk of bias, but provide precise results and exhibit low heterogeneity. Freehand procedures have lower sensitivity than imaging-guided methods. The strength of conclusions about the comparative test performance of automated and vacuum-assisted devices (when using the same imaging guidance) is deemed low, because of concerns about the risk of bias of included studies and the reliance on indirect comparisons. There were insufficient data to draw conclusions for MRI-guided biopsy or women at high baseline risk of cancer. Harms were reported inconsistently, raising concerns about selective outcome and analysis reporting. There is low strength of evidence that vacuum-assisted procedures appear to have a higher risk of bleeding than automated methods. There is moderate strength of evidence that women diagnosed with breast cancer by core needle biopsy are more likely to have their cancer treated with a single surgical procedure, compared with women diagnosed by open surgical biopsy.