Delay to Formalin Fixation (Cold Ischemia Time) Effect on Breast Cancer Molecules

Thaer Khoury, MD


Am J Clin Pathol. 2018;149(4):275-292. 

In This Article


The terms cold ischemia time and delay to formalin fixation have been used in the literature to refer to the time lag between harvesting tumor tissue from the patient to immersing it in formalin.[16–28] The author of this review is the first to introduce the term DFF.[17] Warm ischemia time is another condition that has been recognized. It occurs during operations where the blood supply is cut while the temperature around the tumor remains the same like the body temperature. Under these conditions, the enzymes are still active and degrade the tumor molecules. The effect of this time is more difficult to study or control, and the time is difficult to measure. However, this time might have less effect on breast tumors than tumors from other organs. That is because the time of operation for breast tumors is shorter, particularly after introducing the use of seeds instead of needle localization.[42,43] In unpublished data, the author of this review has found that the operative room time is significantly shorter when a seed is placed in the tumor instead of a needle (paper in preparation). In most of the reviewed studies, DFF time period is defined as the time from the sample received to the time immersed in formalin. However, these studies did not measure the time from tissue excision to the time received at the laboratory. In Khoury et al,[17] this time ranged from 2 to 5 minutes.

In a setting of organ transplant, two distinct periods of ischemia have been recognized: ischemia during implantation, from removal of the organ from ice until reperfusion, and ischemia during organ retrieval, from the time of cross-clamping (or of asystole in non-heart-beating donors) until cold perfusion is commenced. This time is also referred to as "cold ischemia time."[44] When the term cold ischemia time is used, it might be confusing unless it is used in the context of handling tumor tissues. The terms delay to fixation, delay to formalin fixation, and cold ischemia time all refer to the same condition of exposing tissue to cold ischemia by delaying the fixation.

Many preanalytical variables in addition to DFF may play a role in accurate measurement of certain IHC markers, including type of fixative (the concentration, pH, and presence or absence of buffer in the formalin solution), time in fixative, variation between IHC runs, and antibody clones, as well as analytical variables, including the method of scoring and interobserver variability.[9–11,16,33–35,37,45–47] Therefore, a reliable study was defined as a study that controlled all or most of these variables and had DFF time the only changing variable. There are inherent variables that cannot be controlled such as tumor heterogeneity.[37] This variable is important when evaluating the DFF effect on IHC markers because most of the studies evaluated small sample sizes (TMAs or CNBs). On the other hand, variables already known prior to accruing the patient may have an effect on IHC markers such as neoadjuvant therapy.[35] This variable is easy to control by simply excluding such cases from the study.

The prospective studies were designed specifically to examine the hypothesis of the effect of DFF on breast biomarkers and other molecules.[16–22,27,28] Therefore, not surprisingly, the number of controlled confounding factors was higher than in the retrospective studies.[23–26] One issue common among all studies was comparing small samples such as TMAs or CNBs vs full sections. The author found tumor heterogeneity as a source of discordance in BC biomarker classifications, including ER, PR, and HER2,[36] and Ki-67 in another study.[40] However, in some of these reviewed studies, the investigators suggested ways of at least partially controlling this variable. The first way was described by Yildiz-Aktas et al.[22] They considered the reduction real when it was consistent and substantial, and it was considered equivalent to heterogeneity when it was focal. The other way was described by Khoury et al[17–19] and Qiu et al.[20] During the design of the study, they took into consideration tumor heterogeneity and planned ahead to acquire a relatively large number of cores (four 0.6-mm cores each) representing a single case in the TMAs. Later, the same authors suggested that the minimum acceptable size and number of TMA cores to test to avoid discordance in Ki-67 were three 0.6-mm cores or two 1.0-mm cores.[40] In the author's opinion, the results of the studies with a prospective design are more credible.

The studies with a prospective design agreed that ER and PR expressions were affected by DFF time period, either as a trend only[16,17,20–22] or with statistical significance.[20,22] There was no uniformity among these studies regarding the DFF time periods that the samples were subjected to. While Khoury et al[17] and Qiu et al[20] found the effect started to appear at the 1-hour mark, Yildiz-Aktas et al[22] found that these changes started to appear as early as 30 minutes. However, statistically significant change was observed only after 3 or 4 hours of DFF. The studies with a retrospective design also found a similar difference between the properly fixed samples and those that were exposed to DFF with no statistical significance.[23–25] Li et al[23] included cases with neoadjuvant therapy. However, conversion of biomarkers (pre- vs posttherapy) is not uncommon. Jin et al[35] found 13% changed from hormone receptor (HR) + to HR– and 5.4% changed from HR– to HR+. The mechanism of the conversion of HR status after neoadjuvant therapy is complex, including tumor heterogeneity,[48] downregulation of the HR caused by neoadjuvant therapy itself,[49] the inherent issue with variation of HR status between the CNBs and EBs,[13] and finally the effect of DFF. Therefore, the results of this study may not be credible.

Although complete conversion of ER/PR from positive to negative due to DFF occurs only rarely, decreased expression may by itself be clinically important. The degree of ER/PR expression predicts the response to hormonal therapy.[50] ER and PR scores are part of the criteria used to decide if a tumor should be sent for an Oncotype DX assay. The author and others have shown a strong correlation between ER and PR scores and Oncotype DX recurrence scores.[51–53] For example, if a tumor is small, with a low Nottingham grade, negative lymph node, and an ER Allred score of 8/PR Allred score of 8, then the tumor normally is not sent for a molecular assay (ie, Oncotype DX assay). However, if the same tumor has ER/PR Allred scores of 5 each (like when the sample is exposed to a prolonged period of DFF), then the sample is usually sent for testing. Therefore, a false decrease of ER/PR scores due to DFF may trigger an unnecessary test of Oncotype DX (Genomic Health, Redwood City, CA).

For HER2 IHC, there was a limitation in the number of examined cases.[28] However, Yildiz-Aktas et al[22] examined enough cases and overcame the issue of tumor heterogeneity as described above. Although it was not statistically significant, HER2 expression declined in samples exposed to DFF. Some of these changes were clinically significant, particularly in the samples that were not refrigerated. Six cases with 2+ staining converted to either 1+ or 0+. The clinical significance of this is that none of these cases would have been reflected to be tested in another method such as ISH as recommended by the guidelines.[29] Lee et al[27] also found clinically significant change in HER2 IHC staining where a single case converted from positive (3+) to negative (0+).

For HER2 tested by ISH (dual ISH or FISH), there was consistency among the studies that this assay was affected by DFF as well. In the properly designed studies, the investigators found that there was a significant change in the test with regard to troubleshooting. These changes appeared as early as 1 hour of DFF. There was no statistically significant effect on the HER2, CEP17, or HER2/CEP17 ratio.[19] However, there was a loss of signal, mainly HER2, when tested by FISH.[17] Therefore, there is potential for a case to be converted from amplified to nonamplified if the ratio is low positive.[17,19] In fact, Lee et al[27] found a single case with low positive HER2 that was converted from amplified to nonamplified by FISH when exposed to DFF. However, Moatamed et al[28] found no difference in the HER2/CEP17 ratio between the clinically submitted sample and the sample that was subjected to 4 days of DFF.

A major limitation in the Portier et al[26] study was with the experimental design. Rather than examining the same patient's tumor at sequential DFF time periods, the investigators evaluated a unique patient sample at each DFF time point. Therefore, this method did not allow direct detection of the time point when HER2 signal was lost in an individual patient. They also scored signal intensity using a 4-point system from 0 to 3 for both FISH and dual ISH assays. In addition to correlating this score to DFF time, they compared FISH to dual ISH at each DFF time. When they compared between FISH and dual ISH, there was a significant difference in signal intensity at every DFF point. There was an overall decline in signal intensity for both HER2 and CEP17 by FISH and dual ISH assays. However, the only significant decline in intra-assay signal with DFF time was in the FISH method for CEP17 in both tumor and stroma and for HER2 in stroma. It is worth noting that the studies that evaluated the signal intensity and the other troubleshooting variables are subject to tissue heterogeneity and technical variation.

While EB is a more representative sample of the patient's tumor, CNB can be used for HER2 analysis. The concordance rate for HER2 tested by either IHC or ISH between the CNB and the subsequent EB is substantial (98%-99%).[48,49] The American Society of Clinical Oncology–CAP guidelines recommend performing HER2 assays on CNBs with the following caveats. Repeat HER2 testing on EBs is recommended when (1) the result is negative and the sample is limited on the CNB, or (2) the results do not fall in the clearly positive or negative range (IHC or ISH) on the CNB.[29] When the assay is to be repeated on the EB, care should be taken in terms of tissue handling, including DFF time, as HER2 could convert from positive to negative if the sample is not properly fixed.[22,27]

It is more difficult to evaluate the effect of DFF on Ki-67. This antigen is more sensitive than the other antigens with regard to the preanalytical variables, such as how the specimen was stored long term, conditions and duration of long-term sectioned slide storage, freezing the specimen for frozen-section analysis before fixation, use of ethanol or Bouin solution rather than neutral buffered formalin fixation, use of EDTA or acid decalcification protocols, and the choice of Ki-67 antibody and staining protocol.[54,55] Moreover, using TMA or core cut biopsy specimens to assess the effect of DFF on Ki-67, as in the three reviewed studies herein,[18,21,24] is not ideal due to the high degree of tumor heterogeneity.[38–40] In a study conducted by us, we have found that the degree of Ki-67 heterogeneity in luminal-type/HER2– BC could reach up to 40%.[40] Therefore, the author thinks that the effect of DFF on Ki-67 is inconclusive and requires better designed studies that take into account the abovementioned factors.

The author briefly reviewed the effect of DFF on markers that are used in a research setting just to outline the importance of DFF and better tissue handling.[21,24] One of the interesting findings was the increased expression of markers that are normally expressed in hypoxia conditions (AKAP13 and HIF1A).[24] The tissue that was exposed to DFF was subjected to stress and hypoxia, which drove the expression high. Therefore, special care should be taken when handling tissue if the aim of the study is to investigate the expression of stress-related genes.

Finally, the author presents some suggestion on how to handle specimens to comply with CAP guidelines and prevent the sample from DFF Figure 3. When the sample is received from the operative room at the pathology department, it may undergo an intraoperative examination for margin check, gross visual examination, frozen section, or tissue procurement. These activities should not exceed 1 hour, even though some centers may claim otherwise.[23] The main hurdle that faces practicing pathologists in keeping DFF time under 1 hour is when the samples are shipped from one center to another. Some centers may immerse the tissue in formalin without sectioning it. This approach may not be sufficient as the formalin would penetrate the tissue at a very slow rate.[56] The periphery of the tumor would be properly fixed but the center would be exposed to DFF Image 1 and Image 2. Refrigerating the sample is not an option either. Two investigators who specifically addressed this issue found that even though the changes were less dramatic than for the nonrefrigerated samples, the effect was not negligible.[17,22] The author suggests that the sample should be bivalved at the center of the tumor and immersed in formalin to have direct contact of the formalin to the tumor. Another option is to inject formalin into the tumor, which was experimented by De Marzo et al,[57] who found modest improvement.

Figure 3.

Schematic figure showing suggestion on how to handle breast tissue specimens. The part below the bracket shows suggestion for when a test is compromised due to delay to formalin fixation. aNever immerse tissue in formalin without cutting, and never leave tissue overnight in the refrigerator. CNB, core needle biopsy; Dx, diagnosis; IHC, immunohistochemistry; ISH, in situ hybridization.

Image 1.

Tissue immersed in formalin without sectioning (H&E): (A) scanning magnification, (B) center of the section (solid outline) with poor fixation (×10), and (C) periphery of the section (dotted outline) showing proper fixation (×10).

Image 2.

Estrogen receptor (ER) staining of the section in Image 1: (A) scanning magnification (note the strong diffuse staining at the periphery compared with the weak sparse staining in the center), (B) center of the section (solid outline) with decreased ER staining (×10x), and (C) periphery of the section (dotted outline) with strong diffuse staining (×10).

In summary, DFF has the potential to be a confounding factor in breast biomarker assays. It may convert a biomarker status from positive to negative, decrease the overall score, or make the assay uninterpretable. The reviewed studies have shown that at least one of these situations could occur in a small number of cases. However, one must strive to accurately measure these biomarkers, as they are integral in therapeutic decision making. Although CAP guidelines require tissue to be fixed within 1 hour, clinically significant changes in the breast biomarkers may not occur until 3 or 4 hours of DFF. Pathologists should be aware of this fact, and measures should be taken to fix the tissue within a short period of time. When the assay shows unusual findings such as unevaluable tissue/autolysis, loss of HER2 signal, or edge effect for ER/PR, the pathologist should choose another block for testing. If the issue persists, it is recommended to test the CNB if it is available (Figure 3).