br Beyond surgery the appropriate use
Beyond surgery, the appropriate use of adjuvant chemoradiation in older patients is controversial ; a prospective randomized study showed a significant longer median survival in patients who received DETANONOate and chemotherapy . However, two randomized controlled trials (RCTs) showed no survival benefit [17,18]. On the other hand, studies have clearly shown the benefit of adjuvant chemotherapy [17, 18[ and adjuvant Gemcitabine was the standard of care in many centers before 2012 [19–21]; standard of care nowadays is Gemcitabine/Cape-citabine or FOLFIRINOX [22,23]. Nonetheless, population-based studies show that a lower proportion of older patients receive chemotherapy and due to the inclusion criteria in the RCTs, leaving older patients out, its efficacy in (frail) older patients is still unclear. A large proportion of the pancreatic cancer patients will present with metastatic (unresectable) disease; Gemcitabine is widely used in the treatment of unresectable pancreatic cancer [24,25]. More recently, trials have shown an improvement in efficacy with combination chemotherapies [26–29] over gemcitabine alone. However, there are only a few studies evaluating the efficacy and safety of these treatments for older patients .
Beyond RCTs, observational studies offer an important alternative source of evidence; in particular, comparisons of treatment and survival using the instrumental variable assumptions could provide clues to op-timize the treatment strategy for older patients. Country may be a suit-able instrumental variable as place of residence determines a patient allocation to one of the cohorts, assuming that there are differences in treatment between the countries to study and that patient and tumor characteristics are equally distributed and that, in general, health systems are similar in both countries. The aim of the present study was to compare treatment and survival for pancreatic cancer patients aged 70 years and older in the Netherlands or treated at Moffitt Cancer Center, Tampa (Florida, US) which offers specialized care tailored to ge-riatric cancer patients.
All age-eligible patients with histologically confirmed pancreatic ad-enocarcinoma (ICD-O morphology codes 8140 and 8500, where known) diagnosed between 2008 and 2012 were identified. At Moffitt Cancer Center, patients were identified through the Moffitt Cancer Reg-istry and the Total Cancer Care™ program and details were retrieved from medical records. Only patients who had their first treatment at Moffitt (and not in another hospital) were included in the cohort. For the Netherlands, treatment and survival data was retrieved from the Netherlands Cancer Registry for all hospitals in the Netherlands; this population-based registry contains data from all cancer patients in the Netherlands. An institutional review board of both institutions reviewed and approved the study protocol and results (MCC 17877 Moffitt and K14.079 for the Netherlands).
For the present comparison, the instrumental variable methodology was followed, and the ‘three assumptions’ to use country as a valid in-strument were assessed. In short, the three assumptions are that  “country” (cohort from a specific country) should be related to the chance of receiving a specific treatment strategy,  that the instrument should not be related to the prognosis of the patients and  that coun-try should not have an effect on the outcome other than through the chance of receiving a certain treatment strategy. Baseline patient and tumor characteristics were compared between the two cohorts; as stage was differentially distributed between the cohorts (p b .001) and associated with survival, analyses were stratified for stage (early stage T1–2, N0, M0 (UICC stage IA and IB); T3 or node positive T3, N0, M0 or T1–3, N1, M0 (UICC stage IIA and IIB) or advanced disease (T4, N0–
1, M0 or metastatic disease M1). Baseline patient and tumor character-istics were compared between the two cohorts. Neo-adjuvant treat-ment (none, chemo-radiation, chemotherapy), surgery (yes or no), adjuvant treatment (none, chemo-radiation, chemotherapy) and non-surgical treatment (no treatment, chemo-radiation, chemotherapy) were compared. For advanced stage, palliative treatment by type (no treatment, chemo-radiation, chemotherapy or radiotherapy) was compared. Treatment guidelines for all ages showed some differences between the countries [30,31], however specific guidelines for older patients with pancreatic cancer are lacking.
Overall Survival (OS) and 95% Confidence Intervals (95%CI) at one and three years after diagnosis were calculated with death due to any cause as event with time from diagnosis (first pathological confirmation of malignancy) to death, stratified by stage. Cause of death was not re-corded for the Netherlands cohort, we used Overall Survival; this seems justified as cause of death was known for the Moffitt cohort and 92.8% died as a result of pancreatic cancer. Besides, a Dutch study showed that 94.7% of the deaths was attributable to pancreatic cancer . Cox proportional hazards models were used to compare the short-term survival with the Netherlands cohort as reference. Two ad-justed models were constructed; one with adjustment for age, sex, grade and year of incidence and one model with an additional adjust-ment for treatment (treatment modalities entered as separate vari-ables). Kaplan-Meier survival curves for both cohorts were generated according to stage at diagnosis. Finally, stratified survival analyses (ad-justed for sex, grade, year and age (where appropriate), with the Netherlands as reference cohort) were performed according to age (70–74, 75–79 and 80 years and older) and type of hospital in the Netherlands (academic, teaching or non-teaching) as pancreatic (surgi-cal) cancer care is mostly centralized in specialized hospitals. STATA/SE version 10.0 (Texas, USA) was used for the analyses.