Excess premature (3-month) mortality in advanced pancreatic cancer could be related to fatal vascular thromboembolic events. A hypothesis based on a systematic review of phase III chemotherapy studies in advanced pancreatic cancer

Abstract

Introduction. An at least 3-month expected survival is a common inclusion criterion in cancer treatment trials, including advanced pancreatic cancer phase III studies. Published survival curves for advanced pancreatic cancer however seem to reflect a substantial survival shortfall. We wanted to assess the strength of this observation and search for an explanation by reviewing the literature. Methods. A Medline and EMBASE search was done for chemotherapy or chemotherapy based phase III studies in advanced pancreatic cancer published since 1997. Similar search was done at the American Society's of Clinical Oncology web site for abstracts presented since 2000. Three months mortality was based on the survival curves presented. Results. Fourteen papers and five abstracts met our criteria and are included in our review. Six thousand two hundred and twelve patients participated in these trials and 1 447 (23.3%) died in the first 3-month period. Figures were worse in patients with metastases and poorer performance status. Assuming that most deaths during treatment happened during the first 3-months, cause of death was reported in only 40 cases (2.8%). Progressive cancer was reported as cause of death in 21 of these cases. Less frequent causes of death were reported to be infections, ‘complications of cancer’, thromboembolic events and renal failure. Discussion. Overall treatment-related deaths represent a very small percentage of the deaths happening during the 3-month period, and are unlikely to be under-reported given the Good Clinical Practice oversight of these trials. Progressive cancer is likely to be an important cause of early mortality but given the very select nature of the trial-related population this cannot explain the phenomenon of 3-month early death burden of 23.3%. Our hypothesis, supported by multiple autopsy series, is that early death burden in advanced pancreatic cancer trial patients is likely to be due to under-reported vascular thromboembolic events. Thromboprophylaxis needs to be addressed in future trials.

Worldwide there were an estimated 232 306 new cases of pancreatic cancer diagnosed in 2002, accounting for 2% of all new cancer cases [1]. Pancreatic cancer has the poorest prognosis of all adenocarcinomas in the human body and its incidence rate almost matches its prevalence rate. It is one of the leading causes of cancer related deaths and it is the 4^th most common cause of cancer related death in the United States and the 6^th in the United Kingdom while only being 11^th in incidence [1], [2].

Because of the typically late onset of symptoms only around 15–20% of pancreatic cancer cases are amenable to potentially curative surgical resection. In about 40% of the patients, involvement of the major blood vessels around the pancreas (superior mesenteric artery, celiac axis, superior mesenteric vein-portal vein confluence, inferior vena cava and aorta) diagnosed either preoperatively or perioperatively precludes complete resection of the tumour. These patients are grouped as having locally advanced pancreatic cancer and surgery is not a therapeutic option. They are usually treated with radiotherapy, chemotherapy or both and a median survival between 7 and 11 months is the best achievable [3]. The remaining patients (45%) present with metastases. For these patients chemotherapy or best supportive care, if their general condition is poor, are the treatment options and a median survival of around 6–7 months and of just around 3–4 months respectively is to be expected [4].

These data derive from randomised phase III trials testing various treatments (mostly chemotherapy agents) in patients with advanced pancreatic cancer. One of the inclusion criteria, explicitly stated in many of these studies, is that the participants should have an at least 3-month expected survival. This means that other inclusion criteria usually contain ‘copy-paste’ parameters, commonly used for most advanced cancer-treatment trials, pertaining to performance status and end-organ status that generally are expected to distinguish patients likely to reach this 3-month target. The 3-month endpoint corresponds to a common assessment milestone for many trials that have response as an endpoint. Most imaging response data are conventionally obtained and confirmed at around week 12 it is therefore not a trivial point to identify these patients as survival shortfalls can compromise trial endpoints. While these common criteria tend to be by and large successful for many other cancers (e.g. breast, ovarian, colon), this does not seem to be the case in advanced pancreatic cancer even in the hands of trial-experienced oncologists. Looking at the survival curves of these studies, it seems to us that a significant number of patients die within the first 12 weeks from the day of their inclusion in the studies without ever reaching or concluding the first tumour-response assessment milestone. This was also our experience in phase II trial-work we were involved with [5], [6]. We have called this phenomenon ‘early death burden’, and wanting to get a better picture of it we undertook a systematic review of the recent studies and a discussion of the findings.

Methods

A search in the literature (Medline and EMBASE databases) was done for English-language articles presenting results from phase III studies dealing with chemotherapy or chemotherapy based first line treatment in advanced pancreatic cancer until the end of 2006. A study was eligible for our review if there was active treatment in all arms. At least the treatment in one trial's arm had to be chemotherapy or chemotherapy based. Articles dealing with regional treatment were excluded. We wanted our review to be relevant as much as possible to the current practice and as in 1997 the current standard of treatment was established (gemcitabine by Burris et al.) [7] we limited our search to articles published since that landmark article.

Medline search was done with the MeSH term ‘pancreatic neoplasms /drug therapy’ and it was limited to ‘randomized controlled trials’.

Search in the EMBASE database was done with the headings ‘advanced pancreatic cancer or pancreatic cancer or pancreatic adenocarcinoma or pancreatic carcinoma’ and ‘randomized or randomised or phase III’.

We have extended our search to include the web site of the American Society of Clinical Oncology in which abstracts from the annual meetings or from the Gastrointestinal Cancer Symposiums of the Society are posted. The period that our search covered on the above web site was from 2000 until 2006.

A trial was eligible for our review if its authors specifically stated inclusion of patients with an at least 3-month expected survival or if the authors stated that ‘common criteria used in oncology studies applied’ in their study as well (that means also an at least 3-month expected survival). Studies included had to have survival curves available for scrutiny.

Three month mortality was calculated from the survival curves presented in these studies.

Results

Our Medline search identified 79 articles of which 17 were related to our topic. Three of these 17 articles were excluded from our review as in one the authors explicitly stated that a 2 month expected survival was adequate for patient recruitment [8] and in the other two no survival curves were included [9], [10].

EMBASE search revealed 55 articles, 14 of which were related to our topic but all of these had already been identified through the Medline search.

Regarding our search on the American's Society of Clinical Oncology web site, we identified 14 abstracts. Two of those were excluded as Kaplan-Meier survival curves were not presented [11], [12] and seven further abstracts have not been included as since then the related studies have been published in peer-review journals and they had been identified through our literature search.

Overall 14 papers and five abstracts have been included in our analysis [7], [13–30].

Characteristics of the full papers

As it can be seen in Table I, 14 papers have been reviewed. Twelve had two treatment arms. The trial of 5-fluorouracil versus octreotide had three arms and the trial of gemcitabine versus marimastat four. But as in the first study one survival curve was given for both chemotherapy arms (single agent 5-fluorouracil and 5-fluorouracil plus leucovorin arms) [13] and as in the later study in three arms the same drug (marimastat) was used although at different doses [14], we have decided to present both studies as if they had two arms [14]. Gemcitabine was the treatment in the standard arm in 11 trials [14], [16–25], and 5-fluorouracil in the remaining three [7], [13], [15]. In nine of these the treatment in the experimental arm was a doublet that included the standard treatment [15–17], [19], [20], [22–25], in four it was a different single agent treatment [7], [13], [14], [18] and in one it was a regimen with four cytotoxic drugs [17]. In ten trials the drug or combination tested was chemotherapy [7], [15], [16], [19–25] while in the other four it was the somatostatin analogue octreotide, a matrix metalloproteinase inhibitor (marimastat or BAY 12-9566) or a combination of chemotherapy with a matrix metalloproteinase inhibitor [13], [14], [17], [18].

Table I.  Three-month mortality in patients with advanced pancreatic cancer received chemotherapy or chemotherapy based treatment in phase III studies which have been presented as full papers in the literature and are included in our review.

Study	Regimen	Number of patients	3-month mortality (%)
Burris [7]	5-Fu	63	32
	Gemc.	63	24
Burch [13]	5-Fu	53	21
	Octreotide	41	51
Bramhall [14]	Gemc.	103	26
	Marimastat	311	39
Maisey [15]	5-Fu	107	24
	5-Fu/mitomycin-c	102	22
Berlin [16]	Gemc.	162	19
	Gemc./5-Fu	160	15
Bramhall [17]	Gemc.	119	28
	Gemc/marimastat	120	28
Moore [18]	Gemc.	139	17
	Bay 12-9566	138	43
Rocha Lima [19]	Gemc.	180	24
	Gemc./irinotecan	180	27.5
Oettle [20]	Gemc.	282	22.5
	Gemc./pemetrexed	283	20
Reni [21]	Gemc.	47	12.5
	Gemc./PEF	52	11.5
Louvet [22]	Gemc.	156	13
	Gemc./oxaliplatin	157	10
Heinemann [23]	Gemc.	95	23
	Gemc./cisplatin	95	14
Stathopoulos [24]	Gemc.	70	25.5
	Gemc./irinotecan	60	22
Abou-Alfa [25]	Gemc.	174	22
	Gemc./exatecan	175	22

Abbreviations: Gemc.=Gemcitabine; 5-Fu = 5-fluorouracil; PEF = cisplatin, epirubicin, 5-fluorouracil.

In the 14 studies presented as full papers since 1997, 3 687 patients participated (mean number per study 263). The 3-months mortality varied from 10% (gemcitabine/oxaliplatin arm of the Louvet et al. study) [22] to 51% (octreotide arm of the Burch et al. study) [13] with a mean 3-month mortality of 23.6%.

Characteristics of the abstracts

In the five abstracts included in the review, all but one [30] had two treatment arms (Table II). Gemcitabine was the comparator in all of those. The experimental therapy in four of those was single agent chemotherapy or a combination of cytotoxic drugs [26], [28–30]. In the fifth study the combination of gemcitabine with the targeted drug erlotinib was tested [27].

Table II.  Three-month mortality in patients with advanced pancreatic cancer treated with chemotherapy or chemotherapy based regimens in phase III studies which have been presented only as abstracts so far and are included in our review.

Study	Regimen	Number of patients	3-month mortality (%)
Chevreton [26]	Gemc.	170	22
	Exatecan	169	28.5
Moore [27]	Gemc.	284	20
	Gemc./erlotinib	285	17
Riess [28]	Gemc.	236	26
	Gemc./5-Fu	230	30
Herrmann [29]	Gemc.	159	13
	Gemc./capecitabine	160	20
Poplin [30]	Gemc.	279	25
	Gemc. FDR	277	20
	Gemc./oxaliplatin	276	28

Abbreviations: Gemc.=gemcitabine; 5-Fu = 5-fluorouracil; Gemc. FDR = gemcitabine fixed dose rate infusion.

In the above five studies, 2 525 patients participated (mean number per study 505). The incidence of premature death varied from 13% (gemcitabine arm of Herrmann et al. study) [29] to 30% (gemcitabine/5-fluorouracil arm of the Riess et al. study) [28]. The mean 3-month mortality rate was 22.7%.

Three-month mortality in both full papers and abstracts

Combining results from full papers and abstracts, the mean 3-month mortality was 23.3%. The average 3-month mortality for patients receiving single agent gemcitabine, which is currently the treatment of choice and was the standard treatment for most studies reviewed, was 21.3%.

Three-month mortality in subgroups of patients

In two studies (Table III) survival curves have been presented for patients with stage III and stage IV disease. We calculated that ‘early death burden’ was worse in stage IV disease (mean values in stage III and stage IV patients were 10% and 19% respectively) [22], [30].

Table III.  Three-month mortality depending on stage or Karnofsky performance status (KPS) of patients with advanced pancreatic cancer as presented in four studies.

Study	Regimen	Subgroup of patients	No of patients	3-month mortality (%)
Louvet [22]	Gemc.	Stage III	47	8.5
		Stage IV	109	18
	Gemc./ oxalip.	Stage III	50	1.5
		Stage IV	107	13
Poplin [30]	Gemc. or gemc.FDR or gemc./oxalip.	Stage III	100	20
		Stage IV	733	27
Riess [28]	Gemc.	KPS 90-100	111	20
		KPS 60-80	125	37.5
	Gemc./5-Fu	KPS 90-100	101	15
		KPS 60-80	76	45
Herrmann [29]	Gemc.	KPS 90-100	84	13
		KPS 60-80	75	18.5
	Gemc./capec.	KPS 90-100	84	11.5
		KPS 60-80	76	33

Abbreviations: Gemc.=gemcitabine; oxalip.=oxaliplatin; Gemc. FDR = gemcitabine fixed dose rate infusion; capec.=capecitabine; 5-Fu = 5-fluorouracil.

In another two studies survival curves have been presented for good and worse performance status patients [28], [29]. From the data in the papers we calculated that the mean ‘early death burden’ was for patients with a Karnofsky performance status score of 90–100 and 60–80, 15% and 33.5% respectively (Table III)

Reported causes of patients mortality in the first 3 months

We scrutinised the papers for the mention of cause of death for patients who die early in these studies.

In only two studies the cause of 26 early deaths is reported (Table IV). Oettle et al. in their study of gemcitabine versus gemcitabine/pemetrexed mention that six patients died from progressive cancer prior to starting chemotherapy [20]. Also Herrmann et al. at the 2005 annual meeting of ASCO, reported 2-month mortality in patients with Karnofsky performance status score of 60–80%. Twenty participants died early and 15 of those died from progressive cancer, three from infections and two from vascular thromboembolic events (myocardial infarction and pulmonary embolism) [29].

Table IV.  Causes of death in 26 patients who died within 3 months from their inclusion in the studies.

Study	No of deaths cause was given	Progressive cancer	MI	PE	Infection in the absence of neutropenia
Oettle [20]	6	6	0	0	0
Herrmann [29]	20	15	1	1	3

Abbreviations: MI = myocardial infarction; PE = pulmonary embolism.

Cause of death in patients who died while on chemotherapy

In most of these studies reviewed the median duration of treatment was around 3 months, therefore we made the assumption that many of the deaths reported as having happened during treatment could be seen as proportionately reflecting events in the ‘early death burden’ period.

We found data regarding deaths while on chemotherapy in seven studies. In the study of Bramhall et al. testing marimastat and in the study of Bramhall et al. examining the efficacy of the combination gemcitabine/marimastat, all deaths that happened during treatment (although exact number not available) the authors attributed them to progressive cancer [14], [17].

Deaths during treatment from further five studies are shown in Table V. Burris et al., Burch et al., Berlin et al., Oettle et al. and Poplin et al. reported in total 14 deaths. It seems that most deaths were due to infections (neutropenic or not) (5 deaths) and ‘fatal complications of cancer’ (4 deaths) [7], [13], [16], [20], [30].

Table V.  Causes of death in 14 patients who died while on treatment (likely to be in the first 3 months).

Study	No of cases cause of death was given	ARDS	Complication of cancer	Diabetic ketoacidosis	Infection	Renal failure	Respiratory arrest
Burris [7]	4	0	4	0	0	0	0
Burch [13]	2	0	0	1	0	0	1
Berlin [16]	2	0	0	0	1	1	0
Oettle [20]	4	0	0	0	3	1	0
Poplin [30]	2	1	0	0	1	0	0

Abbreviations: ARDS = acute respiratory distress syndrome.

The authors of three further studies (Maisey et al., Louvet at al., Reni et al.), report that they had no toxic deaths during treatment but do not elaborate further on the cause of death of their patients during treatment.

Causes of death overall in patients who died the first 3 months and during chemotherapy (assuming they also died the first 3 months).

Overall, and given the above postulations, we have a cause of death reported for only 40 of the 1 447 premature deaths. From this reported group progressive disease seems to be the most frequently stated cause of death (52.5%). Other less frequent causes of death reported were infections (20%), complications directly related to cancer not further specified (10%), vascular thromboembolic events (myocardial infarction and pulmonary embolism) (5%), and renal failure that developed while patients receiving chemotherapy (5%).

Discussion

We have found that 23.3% (almost 1 in 4) of patients with advanced pancreatic cancer in recent phase III studies receiving chemotherapy or chemotherapy-based treatment were dead within 3 months from the day of inclusion in these studies. A specific cause of death was reported for only 2.8% of these deaths precluding any evidence-based conclusions for the cause of ‘early death burden’ in advanced pancreatic cancer. However some assumptions can be made.

First of all it seems to us that overall the conventionally viewed toxic deaths due to chemotherapy (neutropenic sepsis, haemolytic uremic syndrome, pulmonary fibrosis and thrombocytopenic bleeding) are likely to represent a very small percentage of the ‘early death burden’. Excluding fatal neutropenic sepsis and acute respiratory distress syndrome, which could also be related to an infection, we found only four toxic deaths from the chemotherapy (renal failure, diabetic ketoacidosis and respiratory arrest). Authors of other studies have reported that they had no toxic deaths and in the vast majority of studies there is no mention about deaths due to chemotherapy. Given the ever-increasing Good Clinical Practice-related regulatory oversight of these trials over the last few years it is unlikely that excess treatment toxicity, especially leading to death, would not have been stringently reported.

Certainly pancreatic cancer patients can be at risk of severe infections mostly biliary sepsis as a consequence of obstruction or due to the fact that many of those might have a biliary stent in situ. However, we believe that at least in the first 3 months when all patients are on active treatment, before first imaging assessment, if any fatal infections did occur they would have been reported within the context of a serious adverse event, as these infections are likely to still be viewed as potential complications of treatment. So that the eight deaths reported due to infection represent a very small percentage of the 1 447 deaths of the ‘early death burden’.

Should we therefore be comfortable with the deduction that almost a quarter of patients that are fit to enter trials of chemotherapy will have died of ‘progressive disease’ within 12 weeks of entry? Advanced pancreatic cancer patients certainly start dying from the minute the trial-treatment commences. Assuming a proportionate distribution it works out as 1.8% per week on trial. We tend to glibly assume that pancreatic cancer is an aggressive disease and biological evidence from in vitro work tends to be cited [31]. What do we mean by aggressive in the clinical sense? If we measure aggressiveness as an ability of this cancer to metastasise widely and kill the host from metastases (e.g. small cell lung cancer, melanoma, sarcoma) then we can demonstrate that in pancreatic cancer lung metastases are not common, bone metastases even less common [32], and brain metastases are the material of case reports [33]. If we measure it as clinical chemoresistance then quite remarkably the same agent (5-fluorouracil) that has provided us with a modest survival advantage in the adjuvant setting for colorectal cancer [34] has done so for pancreatic cancer as well [35].

There is no doubt that progressive disease will be in due course the main cause of death in advanced pancreatic cancer patients mostly through intraperitoneal and liver-related complications but to assume that almost a quarter of the trial-eligible patients with normal to minimally abnormal liver function tests, normal renal function, able to eat and drink and performance status as measured on Karnofsky scale of equal or better than 60% (usual trial-entry criteria apply) will suddenly obstruct or get liver failure or intractable peritoneal carcinomatosis just to fulfil the diagnosis of having died of ‘progressive disease’ is difficult to visualize. Moreover, this assumption is contradictory to the persistently replicated findings of a multitude of autopsy trials/series which actually draw attention to what we think is the most likely explanation of ‘early death burden’; namely the spectre of the vascular thrombotic events. The sheer volume of documented events in multiple studies likely, in our view, to explain a proportional ‘stealth’ mortality.

In the study of Moore et al., who reported cause of death for all participants, ‘progressive cancer’ represented 96% of deaths [18]. What does ‘progressive cancer’ mean in this context especially if in one arm of the study 43% of patients have died before 12 weeks? In a small but insightful study from 1984 in patients with bypassed locally advanced pancreatic cancer treated with helium ion irradiation with or without 5-fluorouracil chemotherapy, efforts were made to obtain a post-mortem at death [32]. This was achieved in 22 patients. All had ‘progressive disease’ at autopsy, 16 with liver metastases, seven with peritoneal metastases and most patients with multiple metastases [including lymph nodes, stomach, duodenum, colon, lung (5), bone (1), adrenal (1), skin(1)] but the cause of death was a pulmonary embolus in ten (45%) of these patients (complicated with evidence of infection in seven). Liver failure and gross ascites was documented only in three cases as the cause of death. No patients had gastrointestinal tract obstruction as cause of death and incidentally five of 22 patients (23%) died within an ‘early death burden’ context (i.e. 2 months after concluding treatment) three of which (60%) from pulmonary emboli.

Evidence of the burden of vascular thromboembolic events in advanced pancreatic cancer stretches back to 1938 [36] and comprehensively reviewed by Khorana et al. [37] so we will limit ourselves only to the citation of a most recent extensive autopsy series by Ögren et al. from Sweden who specifically examined the presence of venous thromboembolism in patients with various types of cancer, including 441 patients with pancreatic cancer. They found that 42% of the pancreatic cancer patients had pulmonary embolism and in at least 14% of patients it was the cause of death [38].

We therefore think that the concept of what constitutes ‘death from progressive disease’ in pancreatic cancer conceals within it a significant burden of mortality due to vascular thromboembolic events and this is deeply disturbing especially when related to fit trial patients that do not reach their first assessment milestone. This is specifically of concern since there is some evidence that chemotherapy per-se and possibly some of the newer anti-angiogenic agents that may become useful in advanced pancreatic cancer do carry a risk of exacerbating the thrombophilic state. How much chemotherapy contributes to thrombosis-related early death cannot be gauged through the methodological approach of our review but a recent epidemiological study by Blom et al. has attempted to quantify this risk and has determined that distant metastases raise the risk of vascular thromboembolic events by 1.9 fold (which may relate to the higher ‘early death burden’ we document in this setting in Table III), but chemotherapy had a notable 4.8 fold increase in the risk of venous thromboembolism [39]. This statistic gives serious impetus to the need to include vascular thromboembolic events as potential chemotherapy-related causes of ‘toxic death’, specifically in advanced pancreatic cancer, but also to the need to consider anti-thrombotic prophylaxis in patients undergoing chemotherapy.

Some evidence that vascular thromboembolic events represent the major cause of the excessive 3-month mortality can be obtained from some recently published studies of thromboprophylaxis in pancreatic cancer patients undergoing chemotherapy. Nakchabandi et al. presented the results of a retrospective study in 180 patients with locally advanced or metastatic pancreatic cancer. Approximately two thirds of the patients, in addition to the standard regional or systematic chemotherapy, received a fixed low dose of warfarin (1.25 mg daily) without monitoring the INR. With the warfarin the median survival was increased and the ‘early death burden’ dropped from 60 to 21% [40]. Icli et al. a few months ago presented the results of a non-randomised study where half of the 69 patients with advanced pancreatic cancer received prophylaxis with the low molecular weight heparin nadroparine calcium in addition to the chemotherapy doublet gemcitabine/cisplatin. The 3-month mortality as calculated from the survival curves provided was 10% in the arm receiving nadroparin and 36% in the arm receiving only chemotherapy [41]. We recently presented the interim analysis of a randomised IIb trial in patients with advanced pancreatic cancer randomizing to either gemcitabine single agent or gemcitabine and dalteraprin at 200 u/kg for 12 weeks. We found a reduction in vascular thromboembolic events from 30 to 0% and a reduction in ‘early death burden’ from 15 to 4% [42]. These early data from these three trials taken together seem to indicate that a staggering 67–75% of the ‘early death burden’ may be due to thromboembolism.

Finally there is also some interesting information that can be gleaned from a phase III trial. The MALT study that randomised ‘all-comer’ cancer patients to 6 weeks nadroparine and standard of care versus standard of care alone. Survival data given in the paper for the 6-week time-point show that mortality was more than halved in the nadroparine arm (from 10 to 4%) [43]. Although this is not strictly the 12 week endpoint we have attached to ‘early death burden’ it gives us some more information in a prospective sense and also suggests that this may not be solely an advanced pancreatic cancer related problem. There are at least two other malignancies with severe thrombosis related problems, non-small cell lung cancer and stomach cancer [44].

The final point we would like to raise is that of the unknown natural history of the ‘silent’ embolus both that of pulmonary embolism [45] but also of the splanchnic veins. These diagnoses are becoming more common with the more powerful multi-slice computer tomography (CT) scanners that we are using and in our experience seem to be a not uncommon finding during advanced pancreatic cancer staging.

Closing we present in Figure 1 (a–d) some of our own experience from staging CT scans of pancreatic cancer patients demonstrating ‘silent’ splanchnic thrombi. The first (a) image is from a 59-year-old patient who presented with jaundice and was found to have cancer of the head of the pancreas with liver metastases. The initial CT scan revealed a superior mesenteric vein thrombus. The second (b) image shows a celiac axis thrombosis in a 70-year-old patient complaining of weight loss and chronic epigastric pain and found to have locally advanced cancer of the body of the pancreas. Image (c) is from a 72-year-old patient presenting with anemia and early satiety. He was found to have a poorly differentiated adenocarcinoma of the body of the pancreas invading into stomach with a lymph node mass at the celiac axis and an the initial CT scan revealed inferior mesenteric vein thrombus. Image (d) is from a 65-year-old patient with locally advanced pancreatic cancer compromising the portovenous confluence and two thrombi in the left portal venous system (arrows) with no other evidence of distal (main) portal vein thrombosis.

Figure 1 a–d.  “Silent” thrombosis of the splanchnic vessels in advanced pancreatic cancer patients. a: Thrombus in the superior mesenteric vein (arrow), b: thrombus in the origin of the celiac axis (arrow), c: thrombus in the inferior mesenteric vein (arrow), d: thrombosis in the left portal vein system (arrows).

Conclusions

‘Early death burden’ in advanced pancreatic cancer seems to be a real phenomenon and only part of a continuum of a serious thromboembolic tendency that afflicts these patients. Oncologists need to be aware of the fact that a large proportion of their advanced pancreatic cancer patients are at risk of early death and that vascular thromboembolic events may be contributing significantly to this risk some of which may be treatment related. Modern trials in advanced pancreatic cancer need to start documenting carefully the elements that constitute ‘death from progressive disease’ in this cancer and possibly view venous thromboembolic events as a treatment related complication. The position of thromboprophylaxis needs to be investigated vis a vis best anticoagulant, optimal dose, and stage of disease setting to try and decrease stealth mortality that probably relates to the whole life of a pancreatic cancer patient but is nowhere felt more acutely than when being treated with chemotherapy in expectation of a response.

Acknowledgements

We would like to thank Dr Ged Avery, and Dr Alex Paddon for their assistance in the preparation of the figure with the CT scan images. Also Andrew Dove for his significant contribution in the literature search.