Population-based cancer registries are important sources of data for cancer control. However, most cancer registries lack data on the location of metastases and thus they have inherent weaknesses in following the process that eventually kills the patient. In this paper we present some alternative sources of data in order to address questions about survival depending on the location of metastasis. We investigated survival of patients with metastases from known primaries, compared with CUP patients, whose primaries were by definition unknown. Overall, survival in metastatic cancer was better if the primary location was known. However, metastatic pancreatic, liver, and stomach cancers –known for their poor prognosis- overall featured worse survival than CUP. The present results show 1) variations in risks of death in patients with defined metastases depending on the primary site of the malignancy, and that 2) survival in CUP is generally worse than with primary cancers metastatic to the same organ where CUP was detected. This may be due to the aggressive behavior of CUP. Although the primary tumor in CUP is thought to be dormant, CUP patients feature early distant metastases [15, 18, 19]. The metastatic tendency may explain the poor prognosis, and, as in primary cancers, metastases are thought to be the cause of death in most cancer patients [1, 2, 7]. Several genes have been implicated in metastasis [6, 20, 21]. Indeed, some important metastatic genes have been shown to be overexpressed in CUP: vessel endothelial growth factor, which induces angiogenesis , and matrix metalloproteinases, proteolytic enzymes mediating local invasion and metastasis .
The estimations of median survival times for included CUP patients, three months, is consistent with previous reports of approximately three to four months in population-based studies [8, 15, 22]. It has been earlier noted that some hospital-based studies have estimated substantially longer CUP survival, probably due to different inclusion criteria . Although the prognosis of CUP is overall poor, some 15–20% of the patients present with less aggressive and/or treatable tumors of favorable prognosis . Some of these include CUP diagnosed in lymph nodes only and others require clinical information not available in the present study, including some colorectal and breast cancers. The present metastatic sites would largely belong to the 80–85% of CUP on unfavorable prognosis. Although therapies have improved, particularly in the favorable subset, there has been no evidence that the overall survival would have changed, unfortunately alike many metastatic cancers [24, 25]. Fast diagnosis is important in CUP and new methods include immunohistochemical and gene expression based methods for tissue-of-origin identification [23, 26, 27]. If the primary cancer can be identified the diagnosis is changed to that cancer which would not be scored as CUP in the present analysis. CUP incidence has been declining during the past decade in many countries and improved detection of primary cancers may have contributed to this trend .
Recently, patients with CUP of brain have been shown to have better survival (HR = 0.85/0.79 men/women) compared to patients with known primaries. However, other investigators have not found any difference in brain metastasis survival between patients with a known or unknown primary location . In the present study, no differences could be found. Lung cancer was the most common source of brain metastases in the present study. Thus, we speculate that the primary source of CUP of the brain may in fact often be lung cancer, which is also the most common cause of death in CUP patients .
In our dataset of CUP patients, CUP of liver featured the shortest median survival, only two months. The poor prognosis of liver involvement in CUP is known: previous estimates have ranged between 1.7 and 10 months [19, 28]. Similarly, CUP patients with liver involvement have been shown to be at an increased risk of death (HR = 1.63) compared with CUP patients without liver involvement. Histological features consistent with neuroendocrine carcinoma have been associated with significantly better prognosis than adenocarcinoma. Liver metastases commonly arise from colorectal cancer, and the five year survival in patients not receiving surgery has been reported to be less than 5% . More recently, the five year survival in selected groups receiving surgery may approach 50% . The HR for colorectal cancer with liver metastasis was 0.42 but those from prostate (0.25) and breast (0.31) cancer were even more favorable. Pancreatic cancer, frequently featuring liver metastases, is associated with a dismal prognosis. Survival in patients with pancreatic liver metastases has been approximated to less than three months . However, median survival in patients receiving surgery has been estimated to 11.4 months . Novel developments regarding chemotherapeutic regimens in selected patients also show promising results .
The skeletal system has been described as the most common site for metastases . Present results show that patients with skeletal lung cancer metastases had similar risks of death than CUP of bone. The prognosis among lung cancer patients with bone metastases was unfavorable. Previous results are in line with our findings, the median survival being only three months even in patients receiving surgical treatment . Remarkably, the HR for prostate cancer was only 0.27 and those of kidney and breast cancer were 0.51 and 0.50, respectively.
Our study has several strengths. We used a national database, considered to be close to 100% complete in cancer registration . Population-based studies on metastatic cancers may encounter problems regarding exclusion of the metastatic sites in the TNM-coding system. The Database used in the present study incorporates data from both the Swedish Cancer Registry and the Swedish Cause of Death Registry. Therefore, we could use the death certificates of cancer patients to identify the locations of metastases. The validity of death certificates in Sweden has a considerable impact on the reliability of our results. In Sweden, the proportion of deaths occurring in hospitals is very high. In 2003, 62.5% of deaths occurred in hospitals, whereas the rest occurred in other health care facilities (nursing homes, hospices etc.) or at home . Furthermore, when considering only deaths with malignancies as the underlying cause, the proportion of hospital death has been shown to be as high as 85.1%. In hospital deaths, the issuing doctor of the death certificate is likely to have been involved in the treatment of cancer patients and therefore have insight in the patient’s history. Therefore, we believe that the high number of hospital deaths in Sweden strengthens the validity of death certificates. Also, the validity of death certificates with a malignancy as the underlying cause has been thought to be among the highest .
Can we be sure that the metastases mentioned in the death certificate were present at the time of diagnosis? This issue was addressed by only including cancer patients with positive distant metastatic status at diagnosis. Moreover, we excluded from our analyses decedents with more than one metastasis mentioned in their death certificate. Finally, in order to exclude the possibility that metastases might have been seeded from another primary site, we also excluded cancer cases with primary cancer diagnoses at multiple sites. Naturally, restriction of investigations to cancer patients with metastases at only one defined organ had a substantial impact on the number of cases available for analysis.
The poor survival in CUP may be due to its aggressive behavior. CUPs may undergo substantial phenotypic changes in order to avoid immunological surveillance, and the primary tumor may in fact reside in the same organ as the metastases themselves . Among different sites, CUPs of the liver have the worst prognosis. It is tantalizing to speculate that this is linked to the immune hypothesis relating to CUP . If many CUP cases are indeed due to prior immunological eradication of the primary, and CUP metastases thus represent immunological escape variants, it is perhaps logical that their growth is fastest in the liver, which has been proposed an “immune suppressive organ” . Overall, the survival of CUP patients was shorter than with patients with known primaries. This is compatible with CUP representing tumors subjected to significant prior immunoediting and/or featuring a high degree of immunosuppression. A tumor initially sensitive to immunological control might have ample time to become more malignant through accumulation of hundreds of mutations during a prolonged equilibrium phase . Upon escape from the equilibrium state an unstoppable killer is then unleashed. CUP accounts for 3–5% of cancer diagnoses, and although it is associated with a bad prognosis, some chemotherapeutic regimens have shown promising results [38–40]. Further research is motivated in order to increase understanding of this large group of cancer patients.