The diet of participants who reported food insecurity was found to be less healthy than that of other participants. They consumed more energy and their food was more energy dense. The expenditure on food was similar in the two groups but the expenditure of the food insecure was significantly higher on non-nutritious food, alcohol and tobacco, their expenditure on nutritious food lower, and the food price per kJ lower.
The data was geographically representative of all of Greenland and the study base made up more than 6% of the adult Inuit population. Although the Inuit in Greenland is an indigenous people with a cultural past as hunters, today they mostly live a life as wage earners in modern towns by and large comparable to the life in Scandinavia and other Western countries. In our sample, less than 4% of participants were hunters/fishermen. It is therefore reasonable to assume that the sample is representative of Western populations, and in particular of indigenous populations in the affluent Western countries.
The key variables in the study will be discussed first, followed by a more general discussion of the findings. The definition of food and nutrition security of the Food and Agriculture Organization of the United Nations (FAO) emphasises availability, accessibility and utilisation of food . Numerous methods have been used to measure food security , mostly targeting the access aspect. In the present study, the household hunger scale  was used, which has been specifically developed for cross-cultural use. The household hunger scale has three main questions about food insecurity during the past 4 weeks plus three questions about the number of times each of the three problems happened. In the course of adaptation for use in a large population health survey in Greenland, the questions about number of times were left out and the time frame was changed from the past 4 weeks to the past 12 months. Questions were translated from English to Greenlandic and independently back-translated by skilled interpreters. In Arctic Canada, the use of a single item food security questionnaire has proven feasible  whereas in Greenland, the use of three instead of one question about food security slightly increased the proportion of food insecure participants (from 7.9 to 8.8%) .
The household asset score has certain advantages over other measures of social position, in particular that it is defined for all participants irrespective of age and is normally distributed. Education was skewed with almost half of the participants having no education beyond school and only 1.3% having a higher education. Job category excludes students and old-age pensioners and is only defined for the age group 25-64. Household asset score as an indicator of social position has been used in several studies from Greenland [24,25,36]. It differs from information about personal income in that it is a measure of the economic conditions of the entire household and that it is a measure of what the income is spent on by ways of durable goods . Similar measures have been used elsewhere in the Arctic . A population health survey from 2005-2010 showed a significant association between household asset score and income reported by Statistics Greenland based on information from the tax authorities: the disposable income was DKK 103,500 per person in the lowest household asset category compared with DKK 195,900 in the highest category (unpublished results).
Information on diet was collected from a food frequency questionnaire which has been validated by biomarkers for the traditional marine diet . Information was obtained by face-to-face interviews, which are usually considered more reliable than self-administered questionnaires. Bias in reporting of diet in food frequency questionnaires may be introduced by the inability to recall and synthesise a general diet over a long period. The questions did not specify a short recent time period and although the interviewers were instructed to ask about the usual diet over the last 12 months, it is reasonable to assume a certain seasonal bias in particular concerning kalaalimernit. However, interviews were evenly spread over the year. The question of social desirability may also introduce bias, which may vary among social groups . The definition of dietary pattern that was used in the present paper has proven its usefulness as independent variable in the analysis of diabetes  and food basket expenditure .
Energy intake was calculated from published values of macronutrients in the food. The accuracy of the estimate is dependent on the calculated food intake and the correspondence of actual food eaten with food items in the published tables. The food tables used were Greenlandic and Canadian values for kalaalimernit and recently updated Danish values for imported food. These values are believed to correspond well with actual food in Greenland. After exclusion of participants with unrealistically low or high reported energy intake, the reported average energy intake was 79% of the calculated energy expenditure. Underreporting of total energy intake estimated by food frequency questionnaires is common, ranging from 4.6 to 42% , so the 20% underreporting of the present study was not unduly high. It must be noted that relative to our previous article based on the same data , the reference values for energy and macronutrients as well as the method for estimation of criteria for exclusion have been updated with newly published information. The number of participants as well as energy and expenses are therefore not identical, although very similar.
Food prices were collected from 14 shops, including the two major food chains in Greenland (Pisiffik/Pilersuisoq and Brugseni) and some private kiosks in 7 of the 12 towns where interviews were conducted. Both normal prices and special offers were recorded. The interviewers who collected the prices were asked to choose from the many varieties of similar food at their own discretion. Although the collection of prices was not further standardised, it is believed that the variety of towns and shop types give a reasonably estimate of average prices.
It was confirmed that food insecurity is inversely associated with socioeconomic position measured by household asset score, education and job. The social indicators are closely associated  and it is uncertain whether one of the indicators is more germane to the study of food security than the others.
As hypothesised, the food insecure participants consumed less kalaalimernit, fruit and vegetables. Their dietary pattern was more often classified as unhealthy, they consumed more non-nutritious food, and the cost per 1000 kJ of their food was lower than that of other participants, which all points to a diet of lower quality. Our hypothesis that the energy intake was lower among the food insecure was not supported. The total energy intake of the food insecure was actually higher than that of other participants due to a higher consumption of energy from non-nutritious food. Also, the energy density of their diet was higher. Contrary to our hypothesis, the weekly expenditure on food was similar among the two groups but our hypothesis that the food insecure spent more money on alcohol and tobacco than other survey participants was supported.
Food insecure persons in Greenland ascribed their unhealthy diet to the high cost of nutritious food , but in our study, the price per kJ was only slightly lower for non-nutritious food and the difference was outweighed by the purchase of relatively more of the non-nutritious food.
Whereas the calculated energy requirement was lower among the food insecure than among other participants, the reported energy intake was higher. The calculated energy requirement among the food insecure participants was lower predominantly because they weighed less than other participants (62.3 vs. 74.2 kg adjusted for age and sex; p<0.0001). Their body mass index was also lower and fewer were overweight or obese. This is at variance with most other studies where food insecurity is directly associated with overweight and obesity . At present, we do not have an explanation for this. There could be several reasons why the reported energy was higher among the food insecure. First, their energy consumption might actually be higher than that of other participants. Second, the food insecure participants reported an energy intake equivalent to 92% of their calculated requirement whereas other participants reported only 77% (p<0.0001; adjusted for age and sex). The food insecure may have a less varied diet, which may be easier to report. There were indeed significant differences in ratios between reported and calculated energy among dietary patterns. Thus, participants eating an unhealthy diet reported 85% of their calculated requirement and those eating a kalaalimernit diet 81% compared with 73-76% among participants with other dietary patterns (p<0.0001).
The food secure and food insecure participants spent equal amounts of money on food and the results thus suggest that there must be other explanations for food insecurity than poverty. Several studies agree that income is an important factor for food choice especially among persons living in low-income, food insecure households [9-13], but attitudes toward food prices also influence food purchase decisions  and the potential influence of cost may be overestimated . Other motives are important and individual priorities in food choice motives play a role in producing social disparities in diet . It is therefore probable that the differences in diet and expenses between food secure and food insecure participants are explained by other mechanisms than purchase power. Food security can be construed as a proxy for social position and the literature about social position and health behaviour is therefore pertinent for the discussion of food security and the unhealthy dietary habits of food insecure participants. A review of socioeconomic disparities in health behaviour outlined nine possible explanations for, among other things, poor nutrition and smoking . Several of these possible explanations are relevant in the Greenland context, but in particular the following seem plausible. First, disadvantaged social position and inequality has been shown to induce stress and reduce the capacity to cope. Inequality does exist in Greenland which has a Gini coefficient of 35 , notably higher than the EU average of 31 , and persons who report food insecurity did perceive inequality . Second, education increases the efficacy, problem-solving skills, ability to process information and locus of control needed to govern daily life and overcome obstacles to good health . Although participants with food insecurity cloaked their explanations for their situation in high prices and lack of money, there was a distinct undercurrent telling of reduced ability to plan ahead . Third, less educated persons may have limited knowledge of the harm of unhealthy behaviour and therefore have less motivation to adopt healthy behaviours . This may particularly be the case for dietary advice, which in the public debate is often perceived as being complicated, ever changing, and often conflicting.
Table 5 is a different approach to explaining the possible roles of poverty and competing expenses on alcohol and tobacco on food insecurity. In total, 91% of the cases of food insecurity could be explained by family unemployment and/or a high expense on alcohol and tobacco. This is in line with a qualitative study from Greenland where the overall narrative was one of lack of money to buy food while several participants admitted that another reason for their food insecurity was that they prioritised alcohol and tobacco over food . Studies from other parts of the world showed similar associations of food insecurity with alcohol and tobacco [48,49].
Strengths and weaknesses
Strengths and weaknesses pertaining to individual variables have been discussed above. The study is part of a long series of population studies about health and social issues that date back to 1993 and thus build on a sizable amount of knowledge. It is a general strength of the study that it is countrywide and includes a large percentage of Greenlandic Inuit and that, geographically, the whole country is represented. The data on diet and cost are unique to this study. It is the first countrywide study of food security in Greenland and the first one to include expenditure on food.
The lack of data on food prices in the villages where 14% of the Inuit population lives is a weakness. The recruitment of participants was random but with a participation rate of 52% the possibility of recruitment bias is another weakness. The participation rate was lowest among young adults and among men. Few did not participate because of illness (2.8% of the sample). More did not participate because the interviewers were unable to contact them (12%), and most did not participate because they chose not to (31%) . In general, persons without employment and other socially vulnerable persons, including those with a risky pattern of alcohol and drug use, are underrepresented in population health studies. This was the case for Greenland but only moderately so . If the sample is socially skewed, we would expect the proportion of food insecure participants to be underestimated, the true population average of food expenses to be lower and the overall dietary pattern to be unhealthier. However, the differences between food secure and food insecure participants would not necessarily be biased.
An average price per food item was used to calculate the cost of food for all participants, but this reduces the interpersonal variation in actual expenses. Individual priorities in food choice play an important role . Searching for bargain prices may be an economic necessity for some, and for a given food item, there may be cheap variants of low quality and more expensive variants that only the wealthy can afford. This is not reflected in our data and may result in bias when comparisons are made between food insecure participants and other participants. A lack of kalaalimernit may contribute to the perception of food insecurity as suggested in a previous study  but the survey questions did not cover this aspect, having a narrow economic definition of food security. However, food insecure participants did eat less meat of marine mammals than other participants but the same amount of fish.