Prognostic Findings for HER2 in Breast Cancer not Reproducible

That HER2 is prognostic of outcome in breast cancer is unquestioned. As Jeffrey Ross at Albany Medical College put it: “Today, no one I know doubts in any way that, in the absence of anti-HER2 therapy, HER2+ breast cancer is an unfavorable subtype and HER2+ status by IHC or FISH is a significant and independent prognostic factor.”

Ross helped shape HER2’s reputation as a particularly aggressive form of breast cancer. In 1998, Ross and co-author Jonathan Fletcher published a review of 47 studies of HER2. Each study was checked for an “impact” on prognosis, either univariate or multivariate. (Appendix C lists the 47 studies.)

Univariate findings can be misleading, often losing significance when multiple factors are taken into account. Regarding the more robust multivariate analyses, Ross and Fletcher reported that 28 (60%) of 47 studies found multivariate impact. The remaining 40% of studies either found no multivariate impact or didn’t conduct a multivariate analysis.

Counted by cases, 10,142 (67%) patients out of 15,248 were in studies found by Ross and Fletcher to have a multivariate impact. Their review concludes: “The preponderance of evidence indicates that HER- 2/neu gene amplification and protein overexpression are associated with an adverse outcome in breast cancer.”

However, the review’s conclusion depends on miscategorizing 9 of the 47 papers examined. Correctly categorizing these 9 studies to reflect their actual findings overturns the conclusion that HER2 is prognostic. The preponderance of evidence is inverted and points to no adverse outcome from HER2 (Table 1). Similarly, the number of cases supporting a prognostic value for HER2 fall from two thirds to less than half (Table 2).

Table 1: Number of studies finding HER2 independently prognostic in multivariate analysis

Table 2: Number of cases in studies finding HER2 independently prognostic in multivariate analysis

Appendix A lists the 9 studies and justification for each recoding.

Ross did not dispute the recodings. Provided with the information in Appendix A and asked if he agreed with the recoding, Ross replied: “I am traveling in Europe and have limited time to review. It is certainly possible that the studies you have cited were not perfectly listed in my manuscript from so many years ago.”

Ross and Fletcher’s review suffers from multiple shortcomings. (Appendix B enumerates important but secondary flaws.) However, the miscoding of papers in Ross and Fletcher’s review is sufficient to overturn the paper’s conclusion.

Conflicts of interest

Investigations of HER2 as a prognostic factor produced contradictory findings and argument—resolved by Ross and Fletcher. Of note, commercial interests played a role in several of the studies they reviewed and the review itself.

Among the 47 papers examined, four [4, 10, 29, 48] list at least one author with a corporate rather than academic affiliation. One abstract [49] includes an author who was then a director of diagnostics at Oncor, maker of a HER2 test. All five studies reported HER2 as prognostic.

In their review, Ross and Fletcher report being consultants for Oncor. However, according to Bloomberg, Ross was Medical Director at Oncor beginning in late 1995 and later Chief Medical Officer when his review with Fletcher was published in 1998. Ross confirmed the accuracy of Bloomberg’s information. The FDA rejected Oncor’s test in 1995 but, as reported in the New York Times, Oncor won approval in 1998.


It is likely true, as Ross stated, that today no one questions that HER2 is prognostic in breast cancer. However, this supreme confidence needs to be recalibrated.

Appendix A: Recoded papers

Of the 47 studies, the nine below were recoded:

[11] (Battifora et al.): Yes to No

The paper reports: "Stepwise Cox Regression: This analysis identified independent prognostic factors of DFS and OS when all variables were considered together. Independent predictors of DFS included stage of disease, histology, and nuclear grade. Nuclear grade and stage were the only significant predictors of OS."

[14] (Lovekin et al.): Yes to No

The paper reports: “Multivariate analysis (Cox, 1972) was used to identify whether c-erbB-2 was of independent prognostic significance. In the context of the temporal variables, tumour size and lymph node stage, cell membrane staining was found to have independent significance as a prognostic factor but significance was lost when histological grade was included in the analysis."

[16] (Dykins et al.): Yes to NA

No multivariate analysis

[20] (Paterson et al.): Yes to No

The paper does not state HER2 is independently prognostic in a multivariate analysis or provide the statistics relevant to such a statement. The authors do suggest possible confounding of prognostic factors: “our study design precluded direct determination of the interrelationships of c-erbB-2 [HER2] amplification with conventional disease parameters.”

[22] (Molina et al.): Yes to NA

No multivariate analysis

[29] (Press et al.): Yes to NA

No multivariate analysis

[31] (Descotes et al.): Yes to NA

As its title states, the paper is a “correlation study between Her-2/neu amplification and prognostic factors.” No disease outcome data are included in the paper.

[34] (Têtu et al.): Yes to No

The paper reports that HER2 was predictive of treatment resistance, not prognostic: “The difference in survival rates between cases was only significant among patients submitted to adjuvant chemotherapy or hormone therapy."

[47] (Charpin et al.): Yes to NA

No multivariate analysis

Appendix B: Additional methodology issues

Inclusion criteria

How the 47 papers reviewed by Ross and Fletcher were selected is not described. In email, Ross wrote that “if you just limit the publications cited to those finding HER2 positive rates between 10 and 30% the prognostic impact of HER2+ status in the pre-anti-HER2 targeted therapy era was profound.”

However the review includes Dittadi et al. [44] which describes a “high risk” group comprising 44% of all cases, well above 30%. Ross and Fletcher count the study as supporting the independent, multivariate prognostic impact of HER2.

Berger et al. [5] and Descotes et al. [31] only examine correlations between biomarkers not with disease outcomes and should not have been included.

Ross and Fletcher included two studies [42, 49] for which there are only abstracts. More generally, the studies included were not graded for quality.

An unknown number of papers were omitted, potentially introducing a selection bias. An omitted paper from Zhou et al. (1989), for example, found no prognostic value for HER2. On the other hand, Wright et al.  (1989) also was not included but found HER2 independently prognostic. Other possible biases in the literature, against publishing, for example, are not examined.

Reviews frequently require a minimum number of cases for a study to be included. Indeed, a number of the papers reviewed by Ross and Fletcher attribute the conflicting results in HER2 studies in part to studies with small numbers of cases.

One study [43] had 37 cases. Ross and Fletcher record it as finding HER2 prognostic in univariate analysis but the paper contains no p values, perhaps because n is so small. O’Malley et al. [41] does not state the number of HER2 positive cases that provided the basis for the conclusion that HER2 was prognostic in multivariate regressions. (The corresponding author did not reply to an email inquiry.)

A 2002 review of prognostic factors in node-negative breast cancer specified inclusion criteria and set a minimum number of cases (200). The paper concluded HER2 is not prognostic.

No quantification of prognostic influence

Ross and Fletcher do not provide summary statistics based on a pooling of results. Heterogeneity of the study designs perhaps made this difficult or impossible. However, if heterogeneity prevented statistical summarization, that would be an important finding to report.

The review includes a table of 18 prognostic factors in breast cancer but makes no comparison of their relative strength and clinical value. The prognostic value of HER2 varied widely. In [18], the p value rested at precisely 0.05. Nodal status and tumor size were vastly more prognostic: p < 0.0001 and p = 0.003 respectively. Quénel et al. [39] found HER2 weakly prognostic: "in our hands, c-erbB2 [HER2] had a poor prognostic value in comparison with the classical prognostic variables…” However, whether such weak prognostic value is general among the papers finding HER2 prognostic is not examined by Ross and Fletcher.

Differences in treatment of cases occur within and between studies but the paper does not control for confounding of prognosis with predicting resistance to treatment.

HER2 positive undefined

Different studies used different definitions of HER2 positive. Even today, the definition of HER2 positive and the search for the best HER2 assay continue to be active areas of study. Ross and Fletcher identify the different assays used in HER2 determination (e.g. IHC, FISH) but cut points are not extracted.

Three studies [4, 20, 30] found amplification of HER2 prognostic. But each used a different cut off for gene copy number: six, three, and seven respectively. A single threshold would likely change the findings of these studies and affect the count of studies finding HER2 prognostic.

Some papers determined cutoffs and comparison groups based on achieving statistical significance. One study [44] found HER2 prognostic by creating a “high risk” group that combined cases with the lowest and the highest expression of p185. The low expression group had the worst outcome. Dittadi et al. go on to conclude p185 was independently prognostic in a multivariate analysis. Slamon et al. [4] simply dropped 23 cases with 2-5 copies of HER2. This remarkably unscientific omission enabled comparing a group with one copy of HER2 to those with six or more, providing the basis for the claim HER2 was independently prognostic in a multivariate analysis.

Negative findings not counted, contradictory findings are

Studies with even a single positive finding were counted by Ross and Fletcher as evidence supporting HER2 as a prognostic factor. The number of negative findings is not reported. For example, O’Reilly et al. [19] found HER2 prognostic for relapse-free survival but not overall survival in node-positive disease. Ross and Fletcher count [19] as one of 28 papers supporting the finding that HER2 is prognostic.

Quénel et al. [39], conducted multivariate analyses for three clinical outcomes for three groups. Among the nine tests in total, HER2 showed prognostic value in two and no prognostic value in seven. Ross and Fletcher count [39] among the papers showing that HER2 is prognostic.

Ross and Fletcher’s design also allows studies with opposing findings to be counted as finding HER2 prognostic.  For example Gusterson et al.  [27] found HER2 prognostic in node-positive but not node-negatives patients. However, Giai et al. [32] found the opposite. The papers contradict each other but both are counted as showing HER2 is prognostic by Ross and Fletcher.

Appendix C: Citations for the 47 studies reviewed by Ross and Fletcher

Footnote numbers are those used by Ross and Fletcher. The same numbers are used throughout this article.

4 Slamon DJ, Clark GM, Wong SG et al. Human breast cancer: correlation of relapse and survival with amplification of the Her-2/neu oncogene. Science 1987;235:177-182.

5 Berger MS, Locher GW, Saurer S et al. Correlation of c-erb B2 gene amplification and protein expression in human breast carcinoma with nodal status and nuclear grading. Cancer Res 1988;48:1238-1243.

6 van de Vivjer MJ, Peterse JL, Mooi WJ et al. Neu-protein overexpression in breast cancer. N Engl J Med 1988;319:1239-1245.

7 Heintz NH, Leslie KO, Rogers LA et al. Amplification of the c-erb B-2 oncogene in prognosis of breast adenocarcinoma. Arch Pathol Lab Med 1990;114:160-163.

8 Tsuda H, Hirohashi S, Shimosato Y et al. Correlation between histologic grade of malignancy and copy number of c-erbB-2 gene in breast carcinoma. A retrospective analysis of 176 cases. Cancer 1990; 65:1794-1800.

9 Borg A, Tandon AK, Sigurdsson H et al. HER-2/neu amplification predicts poor survival in node-positive breast cancer. Cancer Res 1990;50:4332-4337.

10 Paik S, Hazan R, Fisher ER et al. Pathologic findings from the nations’ surgical adjuvant breast and bowel project: prognostic significance of erb B2 protein overexpression in primary breast cancer. J Clin Oncol 1990;8:103-112.

11 Battifora H, Gaffey M, Esteban J et al. Immunohistochemical assay of neu/c-erb B-2 oncogene product in paraffin-embedded tissues in early breast cancer: Retrospective follow-up study of 245 stage I and II cases. Modern Pathol 1991;4:466-474.

12 Kallioniemi OP, Holli K, Visakorpi T et al. Association of Cerb B2 protein over-expression with high rate of cell proliferation, increased risk of visceral metastasis and poor long-term survival in breast cancer. Int J Cancer 1991;49:650-655.

13 Clark GM, McGuire WL. Follow-up study of HER-2/neu amplification in primary breast cancer. Cancer Res 1991;51:944-948.

14 Lovekin C, Ellis IO, Locker A et al. C-erb B2 oncoprotein expression in primary and advanced breast cancer. Br J Cancer 1991;63:439-443.

15 McCann AH, DeDervan TA, O’Regan M et al. Prognostic significance of C-erb B2 and estrogen receptor status in human breast cancer. Cancer Res 1991;51:3296-3303.

16 Dykins R, Corbett IP, Henry J et al. Long term survival in breast cancer related to overexpression of the C-erb B2 oncoprotein: an immunohistochemical study using monoclonal antibody NCL-CB11. J Pathol 1991;163:105-110.

17 Rilke F, Colnaghi MI, Cascinelli N et al. Prognostic significance of HER-2/neu expression in breast cancer and its relationship to other prognostic factors. Int J Cancer 1991;49:44-49.

18 Winstanley J, Cooke T, Murray GD et al. The long term prognostic significance of C-erb B2 in primary breast cancer. Br J Cancer 1991;63:447-450.

19 O’Reilly SM, Barnes DM, Camplejohn RS et al. The relationship between C-erb B2 expression, and s-phase fraction in prognosis in breast cancer. Br J Cancer 1991;63:444-446.

20 Paterson MC, Dietrich KD, Danyluk J et al. Correlation between C-erb B2 amplification and risk of recurrent disease in node-negative breast cancer. Cancer Res 1991;51:556-567.

21 Toikkanen S, Helin H, Isola J et al. Prognostic significance of Her-2 oncoprotein expression in breast cancer: a 30-year follow up. J Clin Oncol 1992;10:1044-1048.

22 Molina R, Ciocca DR, Candon AK et al. Expression of HER-2/neu oncoprotein in breast cancer: a comparison of immunohistochemical and western blot techniques. Anticancer Res 1992;12:1965-1991.

23 Noguchi M, Koyasaki M, Ohta N et al. c-erb B-2 oncoprotein expression versus internal mammary lymph node metastases as additional prognostic factors in patients with axillary lymph node-positive breast cancer. Cancer 1992;69:2953-2960.

24 Allred DC, Clark GM, Tandon AK et al. HER-2/neu nodenegative breast cancer: prognostic significance of overexpression influenced by the presence of in-situ carcinoma. J Clin Oncol 1992;10:599-605.

25 Babiak J, Hugh J, Poppeme S. Significance of c-erb B-2 amplification in DNA aneuploidy. Analysis in 78 patients with node-negative breast cancer. Cancer 1992;70:770-776.

26 Tiwari RK, Borgen PI, Wong GY et al. HER-2/neu amplification and overexpression in primary human breast cancer is associated with early metastasis. Anticancer Res 1992;12:419-426.

27 Gusterson BA, Gelber RD, Goldhirsch A et al. Prognostic importance of C-erb B2 expression in breast cancer. J Clin Oncol 1992;10:1049-1056.

28 Bianchi S, Paglierani M, Zampi G et al. Prognostic significance of C-erb B2 expression in node negative breast cancer. Br J Cancer 1993;67:625-629.

29 Press MF, Pike MC, Chazin VR et al. Her-2/neu expression in node-negative breast cancer: direct tissue quantification by computerized image analysis and association of overexpression with increased risk of recurrent disease. Cancer Res 1993;53:4960-4970.

30 Seshadri R, Firgaira FA, Horsfall DJ et al. Clinical significance of Her-2/neu oncogene amplification in primary breast cancer. J Clin Oncol 1993;11:1936-1942.

31 Descotes F, Pavy J-J, Adessi GL. Human breast cancer: correlation study between Her-2/neu amplification and prognostic factors in an unselected population. Anticancer Res 1993;13:119-124.

32 Giai M, Roagna R, Ponzone R et al. Prognostic and predictive relevance of C-erb B2 and ras expression in node-positive and negative breast cancer. Anticancer Res 1994;14:1441-1450.

33 Muss HB, Thor AD, Berry DA et al. Cerb-B2 expression and response to adjuvant therapy in women with node-positive early breast cancer. N Engl J Med 1994;330:1260-1266.

34 Têtu B, Brisson J. Prognostic significance of Her-2/neu oncogene expression in node-positive breast cancer. The influence of the pattern of immunostaining and adjuvant therapy. Cancer 1994;73:2359-2365.

35 Hartmann LC, Ingle JN, Wold LE et al. Prognostic value of CerbB2 overexpression in axillary lymph node-positive breast cancer. Results from a randomized adjuvant treatment protocol. Cancer 1994;74:2956-2963.

36 Jacquemier J, Penault-Llorca P, Viens P et al. Breast cancer response to adjuvant chemotherapy in correlation with erb B2 and p53 expression. Anticancer Res 1994;14:2773-2778.

37 Marks JR, Humphrey PA, Wu K at al. Overexpression of p53 and Her-2/neu proteins as prognostic markers in early stage breast cancer. Ann Surg 1994;219:332-341.

38 Rosen PP, Lesser ML, Arroyo CD et al. Immunohistochemical detection of Her-2/neu expression in patients with axillary lymph node-negative breast carcinoma. A study of epidemiologic risk factors, histologic features and prognosis. Cancer 1995;75:1320-1326.

39 Quénel N, Wafflart J, Bonichon F et al. The prognostic value of c-erbB2 in primary breast carcinomas: a study on 942 cases. Breast Cancer Res Treat 1995;35:283-291.

40 Sundblad AS, Pellicer EM, Ricci L. Carcinoembryonic expression in stages I and II breast cancer; its relationship with clinicopathologic factors. Hum Pathol 1996;27:297-300.

41 O’Malley FP, Saad Z, Kerkvliet N et al. The predictive power of semiquantitative immunohistochemical assessment of p53 and C-erbB2 in lymph node-negative breast cancer. Hum Pathol 1996;27:955-963.

42 Hieken TJ, Mehta RR, Shilkaitis A et al. Her-2/neu and p53 expression in breast cancer: valid prognostic markers when assessed by direct immunoassay, but not by immunochemistry. Proc Annu Meet Am Soc Clin Oncol 1996;15:113a.

43 Xing W-R, Gilchrist KW, Harris CP et al. FISH detection of HER-2/neu oncogene amplification in early onset breast cancer. Breast Cancer Res Treat 1996;39:203-212.

44 Dittadi R, Brazzale A, Pappagallo G et al. ErbB2 assay in breast cancer: possibly improved clinical information using a quantitative method. Anticancer Res 1997;17:1245-1247.

45 Fernandez-Acenero MJ, Farina Gonzalez J, Arangoncillo Ballesteros P. Immunohistochemical expression of p53 and c-erbB-2 in breast carcinoma: relation with epidemiologic factors, histologic features and prognosis. Gen Diagn Pathol 1997;142:289-296.

46 Eissa S, Khalifa A, el-Gharib A et al. Multivariate analysis of DNA ploidy, p53, c-erbB-2 proteins, EGFR, and steroid hormone receptors for short-term prognosis in breast cancer. Anticancer Res 1997;17:3091-3097.

47 Charpin C, Garcia S, Bouvier C et al. c-erbB-2 oncoprotein detected by automated quantitative immunocytochemistry in breast carcinomas correlates with patients’ overall and disease-free survival. Br J Cancer 1997;75:1667-1673.

48 Press MJ, Bernstein L, Thomas PA et al. Her-2/neu gene amplification characterized by fluorescence in situ hybridization: poor prognosis in node-negative breast carcinomas. J Clin Oncol 1997;15:2894-2904.

49 Ross JS, Muraca PJ, Jaffe D et al. Multivariate analysis of prognostic factors in lymph node negative breast cancer. Mod Pathol 1998;11:26a.

50 Depowski PL, Brien TP, Sheehan CE et al. Prognostic significance of p34cdc2 cyclin dependent kinase and MIB1 overexpression, and HER-2/neu gene amplification detected by fluorescence in-situ hybridization in breast cancer. Mod Pathol 1998;11:18A.