This retrospective study included all cases of CUP at Shahid Faghihi hospital from 2015 to 2020 which were available. Among total 307 cases, the final diagnosis for primary origin for 27(8.7%) cases remained unknown despite surveys and ancillary and diagnostic modalities. In a previous study by Brown et al., metastatic tumors were evaluated and the origin of 14% of cases remained indeterminate (12).
Gastrointestinal(GI) tract origin:
In this study, all patterns of CK7/CK20 expression were seen in GI origin cases with the CK7+/CK20- patients were the most prevalent pattern in Upper GI and CK7-/CK20 + in lower GI origins, respectively.
CDX2 was positive in 95.5% of cases of lower GI and 50% of upper GI origin carcinomas. Previous studies showed that CDX2 could be positive in more than 90% of colorectal cancer (1, 13, 14, 15, 16). However, Su et al. and Bayrak et al. showed that CDX2 is not a good predictor of the lower GI origin because it could be positive in all GI tract origin cases (17, 18). CDX2 positivity in in different studies ranged from 18% to more than 70% in gastric cancers (17, 15, 22, 24) and Vang et al. showed CDX2 expression in 74 % of metastatic gastric cancers to the ovary (25). Special AT-rich sequence-binding protein 2(SATB2) was positive in our lower GI cases. This marker also showed high sensitivity and specificity for colorectal origin (1, 14, 16) and Ma et al. demonstrated SATB2 as a novel marker with 85% expression in both primary and metastatic signet ring colorectal origins (19). But it is not a good marker for diagnosis of upper GI origin (13, 19, 26). Villin was positive in our upper and lower GI origin malignancies, Previous studies showed a high positive rate range from 75% to more than 90% in colorectal carcinomas (especially in well differentiated cases) but lower rates in upper GI tumors (1, 14, 20, 21, 22). Although it is noteworthy to mention that villin expression is not limited to GI malignancies. Werling et al. study stated that Villin was positive in 40% of pancreatic carcinomas, 78% of esophageal carcinomas, and 64% of mucinous ovarian adenocarcinomas(21).
It seems that using CDX2, Villin, and SATB2 is helpful for determining the lower GI origin. And combination of Villin and CDX2 in cases with suspicious upper GI origin will increase diagnostic yield.
Liver:
We had 4 cases with clinical diagnosis of liver as the primary source without accurate pathologic diagnosis. They were 66.7% CK7-/CK20- and 33.3% CK7+/CK20-.
Hepatocyte Paraffin 1(Hep Par1) and Arginase were checked and were negative.
In Kandalaft et al. study, Hep Par1 was positive in 70–100% of cases (1). Also, another experiment by Yan et al. showed 96 % sensitivity for arginase1 even for poorly differentiated hepatocellular carcinomas(27). Radwan et al. showed arginase1 and Hep Par1 expression of 84% and 70%, respectively, and they believed that in high-grade tumors, sensitivity and specificity of arginase1 were more than heppar1 (28). Also, a similar study by Koehne de Gonzales et al. estimated the sensitivity and specificity of Hep Par1 as 70% and 84%, and for arginase1 84% and 96%, respectively (29). No hepatic-specific markers were expressed in our cases and they may be better to be considered as undetermined origin.
Breast:
In our study. CK7+/CK20- pattern was seen in 86.2% and CK7-/CK20- pattern in 13.8% of breast origin cases.
Estrogen receptor(ER), was in 73.1%, GATA3 was in 88.9% and Gross cystic disease fluid protein 15(GCDFP15) was positive in 57.1% breast origin malignancies. Previous studies stated that GATA3 has more sensitivity for diagnosis of breast cancer compared to GCDFP15 (1, 30, 31, 32, 33, 34, 35, 36). The positive rate for GCDFP15 is still lower in metastatic cases (33)
In O’ Connell et al. research, ER and GCDFP15 expression in metastatic breast cancer was calculated to be 72% and 78%, respectively (30). Kandalaft et al. study recommended GCDFP-15, GATA3, and mammaglobin panel (1), but in our study it seem that combination of ER and GATA3 can be helpful in diagnosis of breast origin in a high proportion of our cases.
Kidney:
In this study, the most prevalent CK7/CK20 pattern in kidney origin cases were CK7-/CK20- followed by CK7+/CK20- pattern.
For determining kidney origin we used CD10, vimentin and PAX8 panel which showed positivity in more than 70% of renal malignancies.
In previous studies by PAX8 had more than 70% positivity in kidney carcinomas (37, 38). and the intensity of its expression is higher in the metastatic cases (39). Tong et al. (40) and Tacha et al. (37) showed that PAX8 is a specific marker for both primary and metastatic RCC. Liu et al. stated that CD10 is positive in 91% of clear cell RCC and vimentin is 100% sensitive for clear cell RCCs (41). Sharma et al. believed that PAX8 and RCC markers are 100% sensitive in metastasis with kidney origin (42). In our study combination of vimentin, PAX8, and CD10 was successful to identify kidney origin in acceptable number of patients.
Uterine cervix:
CK7+/CK20- was the most prevalent pattern seen in our uterine cervix origin cases. All cases were poorly differentiated; thus, types of cervical cancers (SCC and adenocarcinoma) could not be determined on histomorphology. Combination of P16, P63 and CK5/6 were used in our study. It has been stated that these markers have good sensitivity and specificity for diagnosis of SCC (43, 44, 45, 46, 47, 48).
Wang et al. showed strong P16 expression in 98% of metastatic cervix SCC to lung (49).
Ovary:
In our study, 83.3% of ovarian origin cancers were CK7+/CK20- and 16.7% were CK7+/CK20+.
In CK7+/CK 20 pattern carcinomas, Wilms tumor (WT)1 was positive in 72.7%, PAX8 was positive in 76.9% and ER was positive in 54.5% However one study by Tacha et al. revealed more than 90% positivity for WT1 and another study showed 70–99% for PAX8(37, 38). Also Ordo’nez et al showed ER expression in 85–90% of ovarian cancers (50).
WT1 and PAX8 combination were positive in about 66.7% of our patients. Since serous carcinoma of the ovary is the most common ovarian epithelial neoplasm with a high risk of dissemination we can use WT1 and PAX8 and ER as a good panel with high sensitivity for diagnosis of ovarian origin carcinoma especially in CK7+/CK20- pattern. In CK7+/CK20 + pattern mucinous carcinoma should be considered so other IHC markers such as CDX2 will be helpful.
Uterus:
All of uterine origin cases were CK7+/CK20-. PAX8 and ER marker was positive in 100% and 50% of cases, respectively.
PAX8 expression according to Tacha and Ozcan et al. was seen in 84–93% of endometrial carcinomas (37, 38)., and in Yemelyanova et al. study, PAX8 was reported as a useful marker to determine uterine origin of adenocarcinomas in extrauterine sites (51). Thus, all previous studies are concordant with our study for the PAX8 marker. In addition in Alkushi et al. study and the Kounelis survey, ER expression were seen in 73% and 54% of endometrial carcinomas, respectively(52, 53).
Lung:
In this study, 82.5% lung origin tumors were CK7+/CK20-, 12.5% CK7-/CK20-, and 5% were CK7+/CK20+.
Thyroid transcription factor (TTF)1 was positive in 73.2%. Which is in concordant to previous studies. The positivity of TTF has been reported from 70–100% in various lung malignancies (9, 45, 54–59). And lower frequency in poorly differentiated adenocarcinomas (9).
In our study, Napsin A was positive in 76.2% According to other studies, Napsin A is positive in more than 80% of patients (60, 59). It has been stated that this marker is more sensitive than TTF1 in the diagnosis of lung carcinoma (57). In our study, PAX8 was negative in lung origin carcinomas which is helpful to differentiate these tumors from thyroid origin tumors which both are TTF-1 positive. Ye et al., also stated that PAX8 was negative in their lung adenocarcinomas (59).
We recommend using both TTF1 and Napsin A as a panel with acceptable sensitivity to detect lung adenocarcinoma.
Thyroid:
Thyroid origin were CK7+/CK20-.
Follicular cell origin thyroid carcinoma:
Among 6 follicular cell origin cases, 83.3% were positive for TTF1. Thyroglobulin (TG) and PAX8 were positive in 100% of cases.
C cell type origin thyroid carcinoma:
Both medullary carcinomas cases found to co-express TTF1 and calcitonin and TG was negative.
Our study was concordant with previous studies, i.e. high sensitivity for TTF-1 and PAX8 was documented in thyroid origin cases. The positivity was more frequent in well differentiated thyroid follicular origin tumors (papillary carcinoma and follicular carcinoams) compared to medullary thyroid carcinomas and anaplastic carcinomas (11, 61–66).
However, Zhang et al. results is not consistent with other studies and showed PAX8 positivity in 31.2% of PTC and 40% of FTC cases (67).
TG expression showed variable results in thyroid follicular epithelial malignancies and in metastatic sites range from 60–100% in different studies (61, 62, 68).
So it seems that combination of TTF1, TG, and PAX8 as a panel of markers with 100% sensitivity for follicular cell origin cases. Positive TTF1 and calcitonin with negative TG are also a good predictor for C cell origin.
Prostate:
In our Prostate origin tumors 87.5% were CK7-/CK20- and 12.5% were CK7-/CK20+.
P504 and prostate specific antigen (PSA) were positive in 83.3% and 50% cases respectively.
Co expression of PSA and P504 were seen in 50% of patients. In previous studies PSA expression showed a high sensitivity and specificity for diagnosis of prostate carcinoma (69–70) but its utility will be less in metastatic sites (71).
According to Mai and Renshaw et al. studies, the possibility of PSA expression is higher in well-differentiated prostatic cancers in comparison with high-grade types (72, 73). Regarding p504 expression, 80–100% sensitivity has been reported for this marker in diagnosis of prostate cancers (74–79).
Bladder:
In our study, 57.2% showed CK7+/CK20+, 28.5% CK7+/CK20-, and 14.3% CK7-/CK20 + pattern, respectively.
GATA3 and high molecular weight keratin (HMWK) were positive in 50% and P63 expressed in 80% of cases.
Previous studies showed GATA3 expression is seen in 70–90% of urothelial carcinomas (80–83). Overall, HMWK is positive in 65–97% of urothelial tumor cases (84, 85, 86, 87), In Kunju et al. p63 was positive in 92% of urothelial carcinomas (85). Thus using a panel consisting of P63, HMWK, and GATA3 will have a high sensitivity for detection of bladder origin.
Pancreatobiliary system:
In pancreatobiliary origin tumors 65.4% were CK7+/CK20- and 34.6% were CK7+/CK20+. CK19 was positive in all patients and GATA3 was positive in 12.5%.
As in our study, Hornick et al. showed that CK19 is expressed in 100% of metastatic pancreatic adenocarcinomas to the liver (88) that is coordinate with a study by Zapata et al., which shows a sensitivity of 100% for CK19 marker in pancreatic adenocarcinoma (89).
In a study by Miettinen et al., pancreatic ductal carcinoma was one of the most GATA3 positive non-mammary cancers with 37% sensitivity (90) and Clark et al. showed GATA3 positivity in 10% of pancreatic cancers (36). Although no specific marker is available for detection of pancreaticobiliary cancer, using CK19 may be helpful for the diagnosis of pancreaticobiliary origin cases.
In our study, immunohistochemical diagnosis and final diagnosis agreement in CUP cases was 59%. In a study by Deyoung et al., IHC and final diagnosis concordance was seen in 67% of cases (91). Also, previous studies by Gambel et al., Brown et al., Dennis et al., and Park et al. showed that IHC predicted the primary origin of metastatic tumors in 64%, 66%, 67%, and 65% of cases, respectively, that are approximately in the same range (92, 12, 93, 94). Also, in another previous study by Weiss et al., in 69% of cases, the origin site was diagnosed by an IHC study (5). The lower successful rate of finding primary origin of carcinomas in our study compared to other studies may be caused by incomplete clinical studies and imaging before patients death. Also, new IHC markers are updating and emerging day by day and there is not any definite standard IHC panel for determination of each carcinoma origin. In our study a uniform and homogenous panels were not used for each individuals and this limitation affects the precise sensitivity of each marker.
Another issue is that the expression of cytoplasmic IHC markers are usually dependent on the grade of tumor differentiation. In metastatic tumors, in many cases, the loss of tumor cell differentiation will occur and many tissue-specific cytoplasmic markers will be lost, but nuclear markers have the least changes. In this study by using IHC panels upper and lower GI, breast, kidney, prostate, ovary, lung and thyroid origin carcinoma was more readily diagnosed compared to Cervix and pancreatobiliary and liver origin tumors