In describing cancer patterns and trends as a means to plan and evaluate cancer control policies, incidence, mortality and survival are considered the standard set of indicators. This study provides a methodological framework for estimating an indicator that integrates the above measures with metrics related to non-fatal outcomes. Disability-adjusted life years (DALYs) measure loss of health as a result of illness in the population relative to the ideal scenario where everyone in the population lives into old age in full health [3]. It is a time-based measure that combines the time lost due to premature mortality (years of life lost, or YLLs) and the duration of disability (years of life lived with a disability, or YLDs) in survivors. One lost DALY equates to one lost year of “healthy” life, either as a result of premature death from the disease, or because of disease-related illnesses or disability [3]. Survival for a number of common cancers has been increasing in the last decades, resulting in an ever-increasing number of survivors [4], some living with cancer sequelae. By estimating disability-adjusted life years (DALYs), two key components of the burden of cancer are captured: one related to premature mortality, the other to the loss of ‘healthy’ life years related to the morbidity that follows a diagnosis of cancer [3].

YLLs were calculated by multiplying the number of cancer-specific deaths at a given age by the standard life expectancy for that age (or age group, Formula2). As has been done in earlier global burden of disease project, we used the model life-table West with a life expectancy at birth of 82.5years for women and 80years for men [7]. We used the life expectancy at the mid-point for each age group (0–14, 15–39, 40–44, 45–49,… , 70–74, 75 and over), corresponding to the age-specific mortality data available on a global basis (see below).

where d=death, x=age, e* standard expectation of life at age x.

YLDs were derived as the product of the number of new cases, the average duration of the disability and disability weightings for the condition (or disease state, Formula3). Disability weights represent a value preference that scales a condition or state from 0 (full health) to 1 (death) [3]. To allow comparison between countries, YLLs and YLDs were age standardized by the direct method, using the world standard population [8].

where i=incidence, x=age, y=disease phase, dw=Disability weight, d=duration of disability.

The general form of the proposed natural history model for cancer is illustrated in Figure ​Figure1.1. We assumed three possible pathways for newly-diagnosed cancer cases:

I. Those who were treated (p) and then cured (s) from cancer, underwent a period of disability during the primary diagnosis and therapy phase (duration=L_D), and a period of disability during remission (L_R1)_, during which patients underwent intensive follow-up to detect recurrence or dissemination. A proportion of treated patients who were considered cured continued to live permanently with one (or more) cancer sequelae.

II. Those who died from cancer after treatment (ps), underwent a period of disability in the following phases: (i) primary diagnosis and therapy (L_D), (ii) the remission period L_R2 (for which the duration was estimated as the time between treatment and death minus 4months) [5], (iii) the pre-terminal phase L_M and (iv) the terminal phase L_T. L_M constitutes the phase where cancer has disseminated and was set to last for 3months [5]. The terminal phase L_T, where patients are at the final stages of living, was uniformly set for 1month [5].

III. Those who did not receive treatment (1-p) underwent a period of disability during the pre-terminal (L_M) and terminal phases (L_T). This group was considered to comprise advanced cancer cases, who would have received palliative treatment only.

On the basis of this natural history model and the parameters contained within, we estimated DALYs for 27 cancers. In this paper we present and validate DALYs for four countries, one within each category of the HDI: Norway, Bulgaria, India and Uganda, in 2008. The main sources of data and the estimation methods are set out below.

Cancer survival has been increasing over the last four decades in many more developed countries such as Norway and Bulgaria. Inequalities in survival are reflected in the larger proportion of the DALYs that are attributable to YLD in these countries, compared to the less developed countries such as India or Uganda. Advances in the treatment of several cancers - such oral cavity, leukemia, testicular cancer has resulted in increasing survival [34]. Earlier detection of breast, cervical and colorectal cancer has increased the rate of treatable cases and thus survival in developed countries [33]. Yet facilities for cancer prevention, diagnosis and treatment in developing countries remain inadequate, calling for action to scale up these activities. In addition we also observed large rates of DALYs attributed to highly preventable cancers such as lung, oral cavity or stomach in all countries. This points to the importance of intervention to reduce cigarette smoking. In Uganda, cancers related to infection (Kaposi sarcoma, cervix cancer, non Hodgkin lymphoma, liver cancer) make a large contribution in DALYs highlighting the value of vaccination or treatment of the various infectious diseases.

Previous studies of the global burden of cancer were performed by WHO as part of a wider effort to map the global burden of disease in general [1,3]. This project, commissioned in 1990 [3], introduced DALYs as a means of facilitating comparison across diseases, countries and over time. In 2004, WHO published disease- and country-specific DALYs [1], and their estimates are generally comparable to our current estimates (as illustrated in Figure ​Figure4).4). Some differences can be attributed to changing rates of incidence or mortality for some cancer sites, as observed for lung cancer in Norway (decreasing) and in Bulgaria (increasing) [35]. The DALY rates for breast, cervical and prostate cancer were relatively high in the present study, most likely due to the improvement in the YLD calculation, by incorporating a more detailed disease quantification of phases and sequelae.

This project has been followed by more recent assessments of the global disease burden, as well as various national initiatives [36-40]. These studies are mostly confined to more developed countries, where similar observations to our findings have been noted, with lung, colorectal, and breast cancers sharing the largest proportion of DALYs. Our study was able to draw on more current epidemiological data to derive incidence and mortality estimates [2,41], population-based data on cancer specific-treatment and outcomes including survival, on the basis of more recent reviews [15-17,19]. A major improvement in this study relative to previous studies is the use of observed survival data in estimating the proportion cured and the median survival of uncured patients [13] as a means to calibrate country-specific estimates of the proportion cured and proportion treated. In addition, larger lists of incorporated sequelae and more information on the relative proportions of treated and non-treated patients, using population based registry data, will have resulted in a more valid set of estimates than has been possible in previous exercises.

We modeled proportion cured based on the relation between the mortality to incidence ratio and the human development index. In an earlier paper we have shown a strong correlation between gross domestic product (GDP) and cancer specific survival [2]. In this paper, HDI was chosen over GDP because HDI that also covers wealth, health and education [12] gave better fit than GDP to survival in our internal analysis. This was confirmed on establishing reasonably well fitting models and predicted estimates for the proportion cured on the basis of HDI. The modeled proportion treated also corresponded well to that reported by several cancer registries. Finally the estimated DALYs in Norway based on the modeling approach were very similar to those calculated using observed data from the Registry, serving as further indication of the validity of proportion cured and proportion treated based on the former procedures.

The simpler two-stage [36] natural history model was compared with the three-stage model to calculate DALYs, in view of the principle that it may reduce the complexity of the data required, and calculations. When the two-stage model was used, it generally increased the DALYs, although the differences with the three-stage disease model were rather modest. In the two-stage disease model all patients received curative-intended treatment, and thus all patients who would eventually die from cancers went through remission followed by pre-terminal and terminal phases. In this model, the increase in time to death increased YLD. Therefore for cancers with long remission times (for example, prostate cancer), the differences between the two- and three-stage models becomes larger.

As the intention of treatment (curative or palliative) tends to be recorded less well than treatment modality, we assumed that the proportion of patients receiving any cancer treatment was a reasonable proxy of proportion curatively treated. This is probably a slight overestimation of the true proportion of patients who received curative treatment, as was indicated by the sensitivity analysis. We also considered the use of advanced cases as a proxy of patients who did not receive curative treatment; this had an effect of overestimating the YLD by up to 4%. Such an observation likely resulted from a larger proportion of YLD contributed by patients who were treated but eventually died from cancer.

The limitations of this study pertain to the inputs and necessary assumptions we have made, given the lack of information available at the appropriate level of detail as input for the calculation of DALYs. Firstly incidence and mortality data were derived from the GLOBOCAN2008 which are estimates with varying accuracy depending on the availability of country-specific data [2]. In most low-income countries, there are no comprehensive national-level data on cancer incidence and mortality, and the estimates are based on model-based rates of mortality (from WHO) and/or limited data on cancer incidence from regional cancer registries.

Secondly, the YLD calculations are based on rather limited country-level data, and most parameters are based on data from high-income countries. As an example country-specific proportion cured was modeled using survival:proportion cured ratio as observed in Norway. Applying this ratio to other higher income country may not cause substantial bias as treatment and follow-up practices may be comparable. Third line treatments might delay deaths from relatively advanced disease and hence after “statistical” cure, patients may still have a higher mortality relative to that observed in the general population. On the other hand, it may be hypothesised in less developed countries, treated patients who are considered cured effectively have a similar mortality experience to the general population. In such circumstances, we may underestimate the proportion cured in the lower and middle income countries.

Furthermore, we assumed a uniform period of two months between onset of symptoms and treatment (allowing for delays due to the patient and the medical care system). While the duration of delay might be reasonable for high income countries, it is probably too optimistic in low and middle income settings. While there are many population-specific studies related to specific cancers [42-44], we are not aware of any overall comparative assessment. In any case, a longer period of disability in this phase is likely to be offset by shorter pre-terminal and terminal phases: and the contributions of YLDs from these disease phases comprise a very small proportion of the cancer-specific DALYs in lower income countries.

Fourthly, in calculating DALYs, we clearly missed many long-term consequences from cancers. For example infertility after chemotherapy for haematological cancer patients is well reported [45], but we assumed that these patients were completely cured from cancer. In addition to the observable clinical sequelae many survivors continue to live with psychological stress related to their cancer diagnosis [46]. The proportion living with sequelae is calculated based on studies and data from developed countries that may approximate the proportion in countries with similar level of development. Yet, in addition to low availability and access to modern cancer care, in developing countries cancer cases show a less favorable stage distribution that requires more aggressive treatment, hence a higher proportion of survivors with cancer sequelae. Suboptimal treatment and follow-up for cancer patients in less developed countries may also cause higher levels of disability, and yet we assumed similar disability weights for each sequelae in all countries. Taking into account of these factors are important in future studies where the burden and hence the priorities in cancer control can be better determined.