Some studies attempted to address unmeasured confounding Tab
Some studies attempted to address unmeasured confounding (Table 2). Wright used hospital referral region as an instrumental variable . Two studies performed sensitivity analyses to assess the impact of unmeasured confounders, including preoperative disease burden [19,36], performance status , and breast cancer susceptibility gene (BRCA) status . One study restricted to patients <70 years old with Klabunde-Charlson comorbidity index of 0 to mitigate confounding by health status .
Four studies [19,36,37,40] attempted to assess heterogeneous treatment effects (Table 1, Table 2). Six studies restricted their study populations based on age [20,36,, , ], body mass index , or comorbidity index . One study  stratified survival by a risk score based on age, comorbidity index, and stage. Three studies stratified survival estimates by stage [19,36,40].
Five studies had evidence of immortal time bias (Table 2). Of those, three [20,35,36] introduced immortal time by including person-time from diagnosis to first treatment in the survival calculation, and two [37,39] introduced immortal time by conditioning survival on events occurring after treatment began (i.e., interval debulking surgery). One study did not provide enough details on survival measurement to make a determination . Another study measured survival from interval debulking surgery . No studies reported the length of time between diagnosis and treatment, making the direction and magnitude of immortal time bias difficult to quantify.
Author contribution statement
Sources of funding
Conflict of interest
Introduction Worldwide, lung cancer is the most commonly diagnosed cancer. In 2012, 1.8 million new lung cancer cases were estimated to have occurred globally, accounting for almost 13% of all new cancer cases (excluding non-melanoma DETA NONOate cancer) . In South America, over 64,000 new cases are diagnosed each year, the second most common cancer among men (after prostate) and the fourth most common among women . While lung cancer mortality rates have begun to stabilize or decrease among males in the region, they continue to increase in most countries among females, reflecting their more recent uptake of tobacco smoking . Uruguay, a country with a high Human Development Index  has an estimated 3.4 million inhabitants and an area of approximately 176,000 square kilometers, making it one of the smaller nations in South America. It has however one of the highest lung cancer rates among men in the Americas [5,6], and the Uruguayan government has implemented an unprecedented series of stringent tobacco control measures upon ratifying the WHO Framework Convention on Tobacco Control (FCTC) in 2004, including Latin America’s first ban on smoking in enclosed public places [, , , ], the world’s largest pictorial warnings (80% of the front and back of cigarette packs), the first ban on differentiated branding , alongside a series of tobacco excise tax hikes . While lung cancer mortality trends have been reported via collaborative studies in the region , the existence of a high-quality and national population-based cancer registry (PBCR) in Uruguay provides a unique opportunity to assess incidence trends by sex over the last quarter century, including analyses by main histological subtype, as well as a quantification of the future lung cancer burden in the country. Using age-period-cohort approaches and national data on smoking prevalence, we aim to provide baseline evidence from which to assess the future impact of the tobacco control measures now in place in Uruguay.
Materials and methods
Discussion The overall declining incidence rates observed among men in Uruguay are not dissimilar to those observed in several western countries, including the UK, Nordic and North American countries and Australia, where the decreasing lung cancer rates in men follow the trend of an earlier tobacco epidemic that peaked during middle of the last century . Considering the tobacco epidemic model proposed by Lopez and Collishaw  and further developed by Thun , one could argue that Uruguay is now in a late stage phase in men and an advanced stage in women . The estimated smoking prevalence of 24.8% for both sexes combined (30.7% and 19.8% in men and women, respectively), in Uruguay represented the highest prevalence seen among several Latin American countries that performed the GATS between 2009 and 2012 . Historical information on tobacco prevalence in Uruguay can only be derived from varying sources using different smoking definitions ; nevertheless, it has permitted estimations that indicate a statistically significant annual decrease of −3.3% (95% CI 2.4–4.1) in the prevalence of current use of tobacco occurred between 2005 and 2011 . Other estimations  have indicated smaller declines, with a −1.5% annualized rate of change (for both sexes combined). Additionally, the last results for the GATS indicate a decline in tobacco smoking of 3.2% in both sexes (5.1% in men and 1.8% in women) over the period 2009–2017.