Time to progression of pancreatic ductal adenocarcinoma from low-to-high tumour stages (2024)

Abstract

Objective Although pancreatic ductal adenocarcinoma is considered a rapidly progressive disease, mathematical models estimate that it takes many years for an initiating pancreatic cancer cell to grow into an advanced stage cancer. In order to estimate the time it takes for a pancreatic cancer to progress through different tumor, node, metastasis (TNM) stages, we compared the mean age of patients with pancreatic cancers of different sizes and stages.

Design Patient age, tumour size, stage and demographic information were analysed for 13 131 patients with pancreatic ductal adenocarcinoma entered into the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) database. Multiple linear regression models for age were generated, adjusting for patient ethnicity, gender, tumour location and neoplastic grades.

Results African-American ethnicity and male gender were associated with an earlier age at diagnosis. Patients with stage I cancers (mean age 64.8 years) were on average 1.3 adjusted years younger at diagnosis than those with stage IV cancers (p=0.001). Among patients without distant metastases, those with T1 stage cancers were on average 1.06 and 1.19 adjusted years younger, respectively, than patients with T3 or T4 cancers (p=0.03 for both). Among patients with stage IIB cancers, those with T1/T2 cancers were 0.79 adjusted years younger than those with T3 cancers (p=0.06). There was no significant difference in the mean adjusted age of patients with stage IA versus stage IB cancers.

Conclusions These results are consistent with the hypothesis that once pancreatic ductal adenocarcinomas become detectable clinically progression from low-stage to advanced-stage disease is rapid.

Significance of this study

Introduction

Patients with pancreatic ductal adenocarcinoma usually present with advanced-stage cancers, have rapidly progressive disease and a poor prognosis. For this reason, pancreatic screening studies have evaluated the utility of using endoscopic ultrasound and MRI/MR cholangiopancreatography-based screening for individuals with a family history of pancreatic cancer in multiple first-degree relatives to try and detect asymptomatic early pancreatic cancers and precancerous lesions.1–9 There is a high prevalence of precancerous lesions among patients who meet criteria for undergoing pancreatic screening.1 These lesions include intraductal papillary mucinous neoplasms (IPMNs) typically identified as pancreatic cysts, and pancreatic intraepithelial neoplasia (PanIN) generally identified only after pathological examination of the resected pancreas.10 Pancreatic screening protocols base their recommendations for surveillance11 largely on the characteristics of the pancreatic cysts identified by screening and on the overall experience of following patients with incidentally-detected pancreatic cysts,10 ,12 despite evidence that most pancreatic cancers are thought to arise from undetected PanIN rather than from cysts, even in participants who are candidates for pancreatic screening.11 ,13 An important goal of pancreatic screening is to identify and treat any pancreatic cancers that develop while they are still small and confined to the pancreas and ideally to identify and treat carcinoma in situ (ie, lining pancreatic cystic neoplasms or as PanIN-3 lesions) before invasive pancreatic cancer develops.11 ,14 Pancreatic cysts with suspicious features can be imaged and sampled directly, but tests to detect PanIN-3 lesions such as by sampling secretin-stimulated pancreatic fluid are still investigational.15

Since pancreatic imaging tests cannot reliably detect PanIN lesions and can only detect a pancreatic cancer mass once it reaches several millimetres in diameter,16 it is important to understand the natural history of the early clinical stages of invasive pancreatic cancer. Reliable estimates of the time it takes pancreatic cancers to progress from its earliest detectable stage to more advanced stages are needed to determine surveillance intervals for patients undergoing pancreatic screening. A study using pancreatic cancer passenger mutations to model tumour progression concluded that pancreatic cancer remains confined to the pancreas for many years.17 This prediction contrasts with the experience of many patients who present with advanced-stage pancreatic cancer and have rapidly progressive disease. One reason for this apparent discrepancy is that this model estimates time starting from when the first cancer cell emerges until it progresses to an advanced cancer and includes the period when the cancer is too small (<4–5 mm in diameter) to be detectable by clinical tests. This undetectable period of growth may represent the majority of the overall time to progression, since a founding cancer cell has to undergo many tumour doublings to become a detectable tumour (a 1 cm diameter tumour contains ∼1 billion cells), but only a few more doublings to grow into an advanced-stage tumour. The most relevant timeframe for early detection purposes is the time from when a cancer first becomes detectable by diagnostic tests up until it begins to progress beyond stage I disease. Although pancreatic cancer has a poor prognosis overall, patients with small stage I cancers have a better outcome than most patients with resectable cancers (mostly stage IIB).18–20

If it takes several years for most small early-stage clinically detectable pancreatic cancers to progress to advanced-stage disease, screening intervals can account for this and focus on trying to detect small stage I pancreatic cancers. On the other hand, if pancreatic cancers usually progress rapidly through its clinical stages (such as 1 year), then patients destined to develop pancreatic cancer are much more likely to present with advanced-stage cancers even if they undergo regular and frequent surveillance, unless screening protocols can reliably detect and treat preinvasive lesions.

Several studies have estimated the growth rate of various cancers using serial tumour marker measurements or serial CT scanning,21 but only limited data of this kind is available for pancreatic cancer. One study that measured serial serum CA19-9 and carcinoma embryonic antigen (CEA) measurements in patients with advanced stage disease estimated the average tumour doubling time of pancreatic cancers to be ∼40–60 days.22 Studies have attempted to estimate tumour growth using experimental models,23 but these models may not sufficiently represent the growth of primary human cancers.

If the average time required for pancreatic cancers to progress through different stages is long enough, it should be reflected in the average age of patients diagnosed at each stage of disease. In this study we analyse patient and tumour data in the National Cancer Institute's (NCI) Surveillance, Epidemiology and End Results (SEER) database to evaluate factors associated with age and tumour size at diagnosis and then used this information to compare the average age of patients with small and low-stage pancreatic cancers versus those with larger and higher stage cancers in order to estimate the average time it takes pancreatic cancer to progress through its clinical stages.

Patients and methods

We analysed the data of patients with pancreatic ductal adenocarcinoma entered into the NCI's SEER database, considered to be representative of the US population.24 In the SEER database, there were 13 131 patients aged between 30 and 95 years and diagnosed with stage I to IV pancreatic ductal adenocarcinoma (American Joint Committee on Cancer (AJCC), Seventh edition, described in online supplementary table S1) between January 2004 and December 2011 who had all the required data for this study (demographics, primary site, tumour size, tumour stage, neoplastic grade, diagnostic confirmation, surgery of the primary site for patients with stage I and stage II disease). The SEER data includes patients from 18 registries: Alaska, Atlanta, Connecticut, California (excluding San Francisco, Los Angeles and San Jose), Detroit, Georgia (excluding Atlanta and rural Georgia), rural Georgia, Hawaii, Iowa, Kentucky, Los Angeles, Louisiana, New Jersey, New Mexico, San Francisco-Oakland SMSA, San Jose-Monterey, Seattle and Utah. We included cases with pancreatic ductal adenocarcinoma who had data available for all the following inclusion criteria: malignant behaviour, known age, age at diagnosis (codes: 30-115), race (codes: white, black, other), sex (codes: male and female), primary site (codes: C25.0-head of pancreas, C25.1-body of pancreas, C25.2-tail of pancreas), neoplastic grade (codes: well/moderately/poorly differentiated and undifferentiated), diagnostic confirmation, AJCC stage (codes: IA, IB, IIA, IIB, III and IV), tumour size (codes: 5-988), therapy; surgery of the primary site for patients with stage I and stage II disease (codes: 0-90). We excluded cases with variant histology such as those with cystic adenocarcinomas and neuroendocrine cancers.

Results

The patient characteristics associated with age at diagnosis in the SEER database are listed in tables 1 and 2. Males with pancreatic ductal adenocarcinoma were significantly younger than females (mean age (95% CI), 65.2 (64.9 to 65.5) vs 66.8 (66.5 to 67.1) years; p<0.0001). Males also had slightly larger primary tumours on average than females (mean tumour diameter; 3.9 (3.9 to 4.0) vs 3.7 cm (3.7 to 3.8) (p<0.0001). African-Americans were also significantly younger at the time of diagnosis than Caucasians (62.7 (62.2 to 63.3) vs 66.4 (66.2 to 66.6) years) and other ethnic groups (65.6 (64.9 to 66.4) years; p<0.0001 for both comparisons). Caucasians were older on average than all other ethnic groups (p=0.034). African Americans also had larger tumours on average than other ethnic groups (mean tumour diameter, 4.0 (3.9 to 4.1) vs 3.8 (3.8 to 3.8) cm; p=0.0001). There was no significant difference in the age of patients whose tumours were located in the head, body or tail of the pancreas (data not shown). The average size of tumours located in the head of pancreas was significantly smaller than those in the body and those in the body were significantly smaller than those in the tail; (mean tumour diameter; 3.5 (3.5 to 3.6) vs 4.3 (4.2 to 4.4) vs 4.8 (4.7 to 4.9) cm; p<0.0001).

Patients with well-differentiated cancers were significantly younger at diagnosis (mean age, 63.9 (63.3 to 64.5) years) than patients with cancers of higher neoplastic grade (moderately, poorly or undifferentiated; 66.3 (66.0 to 66.5), 66.4 (66.1 to 66.7) and 65.3 (64.0 to 66.6) years; p<0.0001, p<0.0001 and p=0.043; respectively). Patients with well-differentiated (mean tumour diameter, 3.7 (3.6 to 3.8) cm) and moderate-differentiated cancers (3.6 (3.6 to 3.7) cm) had smaller tumours on average than patients with poorly differentiated cancers (4.0 (4.0 to 4.1) cm; p=0.0001 for both). Patients with undifferentiated cancers had the largest tumours (4.7 (4.4to 5.0) cm; p<0.0001).

Discussion

In this study we have calculated differences in the average age at diagnosis of patients with pancreatic cancers of different sizes and stages after adjusting for cofactors that are associated with age at diagnosis. These age differences provide an estimate of the average time it takes pancreatic cancers to grow within and through different stages of the disease. As a representative example, we found that among patients diagnosed with localised or locally advanced disease those diagnosed with T3 or T4 pancreatic cancers were 1.06 and 1.19 adjusted years older, respectively (13–14 months), than patients with T1 tumours (p=0.03; table 3). Such small age differences between patients with high versus low T stage cancers are in contrast with estimates derived from molecular clock data17 and are consistent with the hypothesis that once a pancreatic cancer is detectable by diagnostic tests, its growth and progression to more advanced stage disease is rapid. It should be noted while that our analysis calculated an average time to progression through different TNM stages, there is likely to be significant heterogeneity in the rate of progression of pancreatic cancers in individual patients.

Evidence that pancreatic cancers rapidly grow through its clinical stages has implications for efforts at detecting pancreatic cancers while it is still at its lowest stage. Ideally, screen-detected cancers would be smaller than the average-sized stage I pancreatic cancer in the SEER database, and so theoretically would have a longer average time-to-progression. There have been only a few screen-detected pancreatic cancers reported in the literature, including cases detected at baseline screening,1–9 ,11 and while most of these screen-detected cancers have been resectable cancers,9 few have been stage I cancers, and some were not detected until they were at an advanced stage. Until pancreatic screening programmes have identified more screen-detected pancreatic cancers and understand the challenges of detecting these earliest stage cancers, our results point to the possibility that many pancreatic cancers diagnosed by pancreatic screening will have progressed to advanced stage even when affected patients have undergone regular surveillance unless screening can identify the precancerous lesions likely to progress to invasive cancer in these patients and treat these lesions before they progress to invasive cancer. Better screening tests that can reliably detect very small T1 cancers (<1 cm, or better yet <5 mm), and if possible PanIN-3 (carcinoma in situ) would minimise the chance of having patients progressing to advanced stage disease while under surveillance. Current screening programmes emphasise the detection and management of precancerous cystic lesions, because these are the most common lesions detected by current screening tests and removing them provides an opportunity to prevent the development of pancreatic cancer. Resecting the pancreas in these individuals will also remove associated PanIN. Another reason why the focus on identifying and treating pancreatic precursor lesions is important is that efforts focused solely on detecting small T1 primary pancreatic cancers (those <1 cm) may be inadequate because many T1-stage cancers have spread beyond the pancreas. In our SEER sample, only 179 of 13 131 patients (1.4%) had pancreatic cancers with a diameter of ≤1 cm, 55 of these 179 (30.1%) had node-positive disease and 18 (10.1%) had distant metastases. These data predict that screening protocols that continue surveillance until a pancreatic mass emerges and surgical intervention is undertaken may often fail to prevent death from pancreatic cancer. To increase the numbers of patients diagnosed with stage I pancreatic cancers more high-risk individuals need to undergo pancreatic screening and pancreatic screening programmes need to have well-validated diagnostic tests that can reliably detect carcinoma in situ and very small (<5 mm) pancreatic cancers.

Expanding pancreatic screening to other at-risk populations might help increase the proportion of patients who had early-stage subcentimetre pancreatic cancers, particularly if better screening tests were developed,14 ,15 ,25 but long-term studies are still needed to evaluate the utility of screening the ‘high-risk’ populations with the appropriate age, family history and/or gene mutation status where screening is currently recommended.11 The benefits of pancreatic screening would likely improve if we were able to better estimate pancreatic cancer risk in the general population26 and in families affected by the disease.27

Our analysis also confirmed prior associations between pancreatic cancer risk factors and age at diagnosis. Thus, it has been reported that African Americans are diagnosed with pancreatic cancer at an earlier age than other ethnic groups,28–30 as are individuals who smoke.31 ,32 The prevalence of known pancreatic cancer risk factors in African Americans versus Caucasians does not explain the earlier average age of pancreatic cancer diagnosis in African Americans.30 It has also been shown that cancers in the tail are generally larger than those in the head of the pancreas.33 We also found that larger primary pancreatic ductal adenocarcinomas were more likely to be poorly differentiated. There is evidence that tumour hypoxia which is likely to be greater in larger tumours contributes to the differentiation state of the tumour.34

We were not able to identify any significant differences in the age of patients with different sized stage I cancers, perhaps because of the small number of patients available for analysis with stage I pancreatic cancers. Another limitation of our study is that although we adjusted for factors in our data set associated with patient age and tumour size at diagnosis, other risk factors for pancreatic cancer development not in the SEER data set may influence the age of pancreatic cancer diagnosis that if accounted for could provide more reliable estimates of the average age of patients at diagnosis and the size and stage of their tumours. One limitation to using tumour size as a measurement of pancreatic cancer progression is that such measurements do not take into account factors such as the contribution of tumour stroma to tumour size and the limitation of tumour size measurements of resection margin-positive cancers. Although the SEER database is a valuable resource, it does not collect data on resection margin status. It is possible that primary tumour size estimates that accounted for margin status would be more accurate than those without. Incomplete ascertainment can be a concern for SEER-based studies and for cross-sectional studies in general,35 ,36 particularly studies of outcomes of patients assigned to different treatments.37 Our inclusion criteria only required cases have a complete set of pathological staging and demographic so we would not expect there to be significant bias related to the exclusion of cases without a complete set of data.

In conclusion, we find using the adjusted average ages of patients with localised or locally advanced pancreatic cancer at diagnosis that disease progression is rapid, with an average estimated time of 14 months for a T1 pancreatic cancer to progress to the T4 stage.

References

1. ↵
   1. Canto MI,
   2. Hruban RH,
   3. Fishman EK, et al

   . Frequent detection of pancreatic lesions in asymptomatic high-risk individuals. Gastroenterology 2012;142:796–804. doi:10.1053/j.gastro.2012.01.005

   OpenUrlCrossRefPubMedWeb of Science

2. ↵
   1. Ludwig E,
   2. Olson SH,
   3. Bayuga S, et al

   . Feasibility and yield of screening in relatives from familial pancreatic cancer families. Am J Gastroenterol 2011;106:946–54. doi:10.1038/ajg.2011.65

   OpenUrlCrossRefPubMed

3. ↵
   1. Verna EC,
   2. Hwang C,
   3. Stevens PD, et al

   . Pancreatic cancer screening in a prospective cohort of high-risk patients: a comprehensive strategy of imaging and genetics. Clin Cancer Res 2010;16:5028–37. doi:10.1158/1078-0432.CCR-09-3209

   OpenUrlAbstract/FREE Full Text

4. ↵
   1. Langer P,
   2. Kann PH,
   3. Fendrich V, et al

   . Five years of prospective screening of high-risk individuals from families with familial pancreatic cancer. Gut 2009;58:1410–18. doi:10.1136/gut.2008.171611

   OpenUrlAbstract/FREE Full Text

5. ↵
   1. Canto MI,
   2. Goggins M,
   3. Hruban RH, et al

   . Screening for early pancreatic neoplasia in high-risk individuals: a prospective controlled study. Clin Gastroenterol Hepatol 2006;4:766–81. doi:10.1016/j.cgh.2006.02.005

   OpenUrlCrossRefPubMedWeb of Science

6. ↵
   1. Canto MI,
   2. Goggins M,
   3. Yeo CJ, et al

   . Screening for pancreatic neoplasia in high-risk individuals: an EUS-based approach. Clin Gastroenterol Hepatol 2004;2:606–21. doi:10.1016/S1542-3565(04)00244-7

   OpenUrlCrossRefPubMed

7. ↵
   1. Brentnall T,
   2. Bronner M,
   3. Byrd D, et al

   . Early diagnosis and treatment of pancreatic dysplasia in patients with a family history of pancreatic cancer. Ann Intern Med 1999;131:247–55. doi:10.7326/0003-4819-131-4-199908170-00003

   OpenUrlCrossRefPubMedWeb of Science

8. ↵
   1. Poley JW,
   2. Kluijt I,
   3. Gouma DJ, et al

   . The yield of first-time endoscopic ultrasonography in screening individuals at a high risk of developing pancreatic cancer. Am J Gastroenterol 2009;104:2175–81. doi:10.1038/ajg.2009.276

   OpenUrlCrossRefPubMed

9. ↵
   1. Vasen HF,
   2. Wasser M,
   3. van Mil A, et al

   . Magnetic resonance imaging surveillance detects early-stage pancreatic cancer in carriers of a p16-leiden mutation. Gastroenterology 2011;140:850–6. doi:10.1053/j.gastro.2010.11.048

   OpenUrlCrossRefPubMed

10. ↵
    1. Brune K,
    2. Abe T,
    3. Canto M, et al

    . Multifocal neoplastic precursor lesions associated with lobular atrophy of the pancreas in patients having a strong family history of pancreatic cancer. Am J Surg Pathol 2006;30:1067–76.

    OpenUrlPubMedWeb of Science

11. ↵
    1. Canto MI,
    2. Harinck F,
    3. Hruban RH, et al

    . International consensus recommendations on the management of patients with increased risk for familial pancreatic cancer (The Cancer of the Pancreas Screening (CAPS) consortium summit). Gut 2013;62:339–47. doi:10.1136/gutjnl-2012-303108

    OpenUrlAbstract/FREE Full Text

12. ↵
    1. Tanaka M,
    2. Fernandez-Del Castillo C,
    3. Adsay V, et al

    . International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology 2012;12:183–97. doi:10.1016/j.pan.2012.04.004

    OpenUrlCrossRefPubMedWeb of Science

13. ↵
    1. Norris AL,
    2. Roberts NJ,
    3. Jones S, et al

    . Familial and sporadic pancreatic cancer share the same molecular pathogenesis. Fam Cancer 2014. In press.

14. ↵
    1. Lennon AM,
    2. Wolfgang CL,
    3. Canto MI, et al

    . The early detection of pancreatic cancer: what will it take to diagnose and treat curable pancreatic neoplasia? Cancer Res 2014;74:3381–9. doi:10.1158/0008-5472.CAN-14-0734

    OpenUrlAbstract/FREE Full Text

15. ↵
    1. Kanda M,
    2. Sadakari Y,
    3. Borges M, et al

    . Mutant TP53 in duodenal samples of pancreatic juice from patients with pancreatic cancer or high-grade dysplasia. Clin Gastroenterol Hepatol 2013;11:719–30. doi:10.1016/j.cgh.2012.11.016

    OpenUrlCrossRefPubMed

16. ↵
    1. Khashab MA,
    2. Yong E,
    3. Lennon AM, et al

    . EUS is still superior to multidetector computerized tomography for detection of pancreatic neuroendocrine tumors. Gastrointest Endosc 2011;73:691–6. doi:10.1016/j.gie.2010.08.030

    OpenUrlCrossRefPubMed

17. ↵
    1. Yachida S,
    2. Jones S,
    3. Bozic I, et al

    . Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature 2010;467:1114–17. doi:10.1038/nature09515

    OpenUrlCrossRefPubMedWeb of Science

18. ↵
    1. Ferrone CR,
    2. Pieretti-Vanmarcke R,
    3. Bloom JP, et al

    . Pancreatic ductal adenocarcinoma: long-term survival does not equal cure. Surgery 2012;152:S43–9. doi:10.1016/j.surg.2012.05.020

    OpenUrlCrossRefPubMed

19. ↵
    1. Winter JM,
    2. Cameron JL,
    3. Campbell KA, et al

    . 1423 pancreaticoduodenectomies for pancreatic cancer: a single-institution experience. J Gastrointest Surg 2006;10:1199–210; discussion 1210–1191. doi:10.1016/j.gassur.2006.08.018

    OpenUrlCrossRefPubMedWeb of Science

20. ↵
    1. de Jong MC,
    2. Li F,
    3. Cameron JL, et al

    . Re-evaluating the impact of tumor size on survival following pancreaticoduodenectomy for pancreatic adenocarcinoma. J Surg Oncol 2011;103:656–62. doi:10.1002/jso.21883

    OpenUrlCrossRefPubMed

21. ↵
    1. Honda O,
    2. Johkoh T,
    3. Sekiguchi J, et al

    . Doubling time of lung cancer determined using three-dimensional volumetric software: comparison of squamous cell carcinoma and adenocarcinoma. Lung Cancer 2009;66:211–17. doi:10.1016/j.lungcan.2009.01.018

    OpenUrlCrossRefPubMedWeb of Science

22. ↵
    1. Nishida K,
    2. Kaneko T,
    3. Yoneda M, et al

    . Doubling time of serum CA 19-9 in the clinical course of patients with pancreatic cancer and its significant association with prognosis. J Surg Oncol 1999;71:140–6. doi:10.1002/(SICI)1096-9098(199907)71:3<140::AID-JSO2>3.0.CO;2-T

    OpenUrlCrossRefPubMed

23. ↵
    1. Gerlee P

    . The model muddle: in search of tumor growth laws. Cancer Res 2013;73:2407–11. doi:10.1158/0008-5472.CAN-12-4355

    OpenUrlAbstract/FREE Full Text

24. ↵
    1. Edwards BK,
    2. Brown ML,
    3. Wingo PA, et al

    . Annual report to the nation on the status of cancer, 1975–2002, featuring population-based trends in cancer treatment. J Natl Cancer Inst 2005;97:1407–27. doi:10.1093/jnci/dji289

    OpenUrlAbstract/FREE Full Text

25. ↵
    1. Kanda M,
    2. Knight S,
    3. Topazian M, et al

    . Mutant GNAS detected in duodenal collections of secretin-stimulated pancreatic juice indicates the presence or emergence of pancreatic cysts. Gut 2013;62:1024–33. doi:10.1136/gutjnl-2012-302823

    OpenUrlAbstract/FREE Full Text

26. ↵
    1. Klein AP,
    2. Lindstrom S,
    3. Mendelsohn JB, et al

    . An absolute risk model to identify individuals at elevated risk for pancreatic cancer in the general population. PLoS ONE 2013;8:e72311. doi:10.1371/journal.pone.0072311

    OpenUrlCrossRefPubMed

27. ↵
    1. Klein AP

    . Identifying people at a high risk of developing pancreatic cancer. Nat Rev Cancer 2013;13:66–74. doi:10.1038/nrc3420

    OpenUrlPubMed

28. ↵
    1. Chang KJ,
    2. Parasher G,
    3. Christie C, et al

    . Risk of pancreatic adenocarcinoma: disparity between African Americans and other race/ethnic groups. Cancer 2005;103:349–57. doi:10.1002/cncr.20771

    OpenUrlCrossRefPubMedWeb of Science

29. ↵
    1. Wray CJ,
    2. Castro-Echeverry E,
    3. Silberfein EJ, et al

    . A multi-institutional study of pancreatic cancer in Harris County, Texas: race predicts treatment and survival. Ann Surg Oncol 2012;19:2776–81. doi:10.1245/s10434-012-2361-z

    OpenUrlCrossRefPubMed

30. ↵
    1. Arnold LD,
    2. Patel AV,
    3. Yan Y, et al

    . Are racial disparities in pancreatic cancer explained by smoking and overweight/obesity? Cancer Epidemiol Biomarkers Prev 2009;18:2397–405. doi:10.1158/1055-9965.EPI-09-0080

    OpenUrlAbstract/FREE Full Text

31. ↵
    1. James TA,
    2. Sheldon DG,
    3. Rajput A, et al

    . Risk factors associated with earlier age of onset in familial pancreatic carcinoma. Cancer 2004;101:2722–6. doi:10.1002/cncr.20700

    OpenUrlCrossRefPubMedWeb of Science

32. ↵
    1. Anderson MA,
    2. Zolotarevsky E,
    3. Cooper KL, et al

    . Alcohol and tobacco lower the age of presentation in sporadic pancreatic cancer in a dose-dependent manner: a multicenter study. Am J Gastroenterol 2012;107:1730–9. doi:10.1038/ajg.2012.288

    OpenUrlCrossRefPubMed

33. ↵
    1. Sohn TA,
    2. Yeo CJ,
    3. Cameron JL, et al

    . Resected adenocarcinoma of the pancreas-616 patients: results, outcomes, and prognostic indicators. J Gastrointest Surg 2000;4:567–79. doi:10.1016/S1091-255X(00)80105-5

    OpenUrlCrossRefPubMedWeb of Science

34. ↵
    1. Axelson H,
    2. Fredlund E,
    3. Ovenberger M, et al

    . Hypoxia-induced dedifferentiation of tumor cells—a mechanism behind heterogeneity and aggressiveness of solid tumors. Semin Cell Dev Biol 2005;16:554–63. doi:10.1016/j.semcdb.2005.03.007

    OpenUrlCrossRefPubMedWeb of Science

35. ↵
    1. Bull K,
    2. Spiegelhalter DJ

    . Survival analysis in observational studies. Stat Med 1997;16:1041–74. doi:10.1002/(SICI)1097-0258(19970515)16:9<1041::AID-SIM506>3.0.CO;2-F

    OpenUrlCrossRefPubMedWeb of Science

36. ↵
    1. Cain KC,
    2. Harlow SD,
    3. Little RJ, et al

    . Bias due to left truncation and left censoring in longitudinal studies of developmental and disease processes. Am J Epidemiol 2011;173:1078–84. doi:10.1093/aje/kwq481

    OpenUrlAbstract/FREE Full Text

37. ↵
    1. Park HS,
    2. Lloyd S,
    3. Decker RH, et al

    . Limitations and biases of the Surveillance, Epidemiology, and End Results database. Curr Probl Cancer 2012;36:216–24. doi:10.1016/j.currproblcancer.2012.03.011

    OpenUrlCrossRefPubMed