According to all controversies regarding the diagnostic/screening value of serum-HER2 in breast cancer, to overcome this debate, we conducted a meta-analysis to integrate the results of studies included with adequate data for calculation of test’s accuracy in this regard. Finally, we found that serum-HER2 indicated a high specificity for the detection of tissue HER2 status in breast cancer. In detail, specificity for a screening test can be defined as the test’s ability to identify the true negative cases or, in other words, identifying all cases who do not have the target disorder based on the performance relative to a gold standard (34). Accordingly, our meta-analysis indicated the specificity 79.27 (95%CI 73.02-85.51) for the serum-HER2 test, the ability of the test for detection of individuals with the negative result based on presence/absence of breast cancer is high and satisfactory.
In order to apply appropriate treatments for a disorder, we mainly need to more accurate diagnostic tests, especially with the development of modernity toward higher speed, cost-effectiveness, performance, and safety, which resulted in several available diagnostic tests for a particular condition (35). Focusing on our study, in several cancers, HER2 status is a necessary item for HER2 targeted therapy, which mostly will be determined using biopsy specimens. The problem with biopsy is that it is not always easily available, and the expression of HER2 is heterogeneous within tumor tissue, which might lead to a false negative outcome (36). Also, given our experiences, other limitations might affect the results as follows: 1. The lab is not equipped for running this test; 2. The cancer is HER2 negative; and 3. The cancer is small, and therefore, the amount of HER2 that is shed into the bloodstream is limited.
Among all available methods for HER-2 examination, IHC and FISH were the most preferred techniques among clinicians and researchers (37). However, the desire for the measurement of serum-HER2 levels as a non-invasive technique for breast cancer diagnosis and prognosis has attracted the investigators’ interests in two recent decades (38, 39).
In more detail, sensitivity for a screening test can be defined as the test’s ability to identify the true positive cases or, in other words, identifying all cases with the considered disorder based on performance relative to a gold standard (34). Accordingly, our findings showed a sensitivity of 53.05 (95%CI 40.82-65.28) for the serum-HER2 test; the ability of the test for detection of individuals with a positive result based on reference standard is low and unsatisfactory.
From another point of view, the indicator PPV demonstrates that how many people tested positive based on the screening test are actually have the considered condition (34). In this regard, our meta-analysis showed the PPV 56.18 (95%CI 44.16-68.20), which is low as like as sensitivity for serum-HER2 test and suggest unsatisfactory performance for this test in comparison to the gold standard.
Moreover, the indicator NPV indicates that how many people tested negative based on screening test truly do not have the target condition. In this regard, our findings showed the NPV 76.93 (95%CI 69.56-84.31), which is high as like as specificity for serum-HER2 test and suggest satisfactory performance for this test in comparison to the gold standard.
Altogether, owing mentioned metrics, another metric is proposed as a diagnostic test accuracy or effectiveness, which defined as the test’s ability to differentiate between subjects with or without the target disorder. In other words, the ability of the test to classify true positive and negative subjects among all subjects (40, 41). Our results revealed the accuracy of 72.06 (95%CI 67.04-77.08) for the serum-HER2 test, which is reasonably high and suggests satisfactory performance in this regard.
It is worth mentioning that the prevalence of target condition affects the diagnostic accuracy of the test. If the sensitivity and specificity do not change, the accuracy of the test will increase by decreasing condition prevalence (42).
As an alternative statistic, positive and negative likelihood ratios are powerful metrics for diagnostic accuracy summarising (43). The likelihood ratio defines as the probability of test results in subjects with the condition to the probability in the subjects without the condition (44). In this regard, our findings demonstrated that PLR 2.10 (95%CI 1.69-2.50) and NLR 0.58 (95%CI 0.44-0.71) for the serum-HER2 test, which shows an association with the presence and absence of the condition, respectively. However, it is reported that PLR greater than 10 and NLR less than 0.1 provide strong evidence for diagnosis (45).
It is significant that if the entered studies into meta-analysis use different cut-off values for presenting positive results for a test, the threshold effect will arise toward bias (46). To avoid the bias, we conducted the sensitivity analysis on studies with the same cut-off values, which revealed no significant differences (Supplementary File).
Considering all the facts mentioned above, although the serum-HER2 test has failed to be considered as a gold standard test, according to the shreds of evidence, this test can be truely useful for the detection of true negatives (HER2 negative status, or absence of breast cancer), especially in low-income regions due to its cost-effectiveness and ease of implementation.
Given the fact that limitations are unavoidable and all studies face that, as a potential limitation in the present study, although most studies used the same cut-off values, since we had no access to raw data, we could not define an ideal threshold for the serum-HER2 test in the diagnosis of breast cancer.