Malnutrition and cachexia among cancer out-patients in Nairobi, Kenya

Cancer is the third leading cause of death in Kenya. However, there is scarce information on the nutritional status of cancer patients to guide in decision making. The present study sought to assess the risk of malnutrition, and factors associated with malnutrition and cachexia, among cancer out-patients, with the aim of informing nutrition programmes for cancer management in Kenya and beyond. This was a facility-based cross-sectional study performed at Kenyatta National Hospital and Texas Cancer Centre in Nairobi, Kenya. The risk of malnutrition was assessed using the Malnutrition Universal Screening Tool (MUST). Diagnoses of malnutrition and cachexia were done using the European Society of Clinical Nutrition and Metabolism (ESPEN) and Fearon criteria, respectively. A total of 512 participants were assessed. Those at risk of malnutrition were 33·1 % (12·5 % at medium risk, 20·6 % at high risk). Prevalence of malnutrition was 13·4 %. The overall weight loss >5 % over 3 months was 18·2 % and low fat-free mass index was 43·1 %. Prevalence of cachexia was 14·1 % compared with 8·5 % obtained using the local criteria. Only 18·6 % participants had received any form of nutrition services. Age was a predictor of malnutrition and cachexia in addition to site of cancer for malnutrition and cigarette smoking for cachexia. The use of the MUST as a screening tool at the first point of care should be explored. The predictive value of current nutrition assessment tools, and the local diagnostic criteria for malnutrition and cachexia should be reassessed to inform the development of appropriate clinical guidelines and future capacity-building initiatives that will ensure the correct identification of patients at risk for timely care.

The 2013 National Guidelines for Cancer Management in Kenya identifies nutrition care for persons with cancer as an integral part of cancer treatment (7) . It singles out the need for early initiation of management plans that incorporate nutritional assessment, diagnosis and intervention. BMI and weight loss are currently the main indicators of malnutrition and cachexia, respectively, in Kenyan public hospitals. However, previous studies have shown the limitations of using BMI as the sole indicator of malnutrition in cancer because it can yield high or normal results in patients with large tumours, ascites or oedema, thereby masking weight loss. Unintentional weight loss is instead recommended (8,9) .
The Malnutrition Universal Screening Tool (MUST) is designed to identify adults who are underweight or obese and at risk of malnutrition, and who may benefit from appropriate intervention (2) . It uses BMI, weight loss and acute disease effect scores in estimating the overall risk of malnutrition. The European Society of Clinical Nutrition and Metabolism (ESPEN) consensus statement provides a minimum set of criteria for the diagnosis of malnutrition that can be applied independently of clinical setting and aetiology (10) . It allows for inclusion of parameters such as fat-free mass where applicable in diagnosing malnutrition. The international consensus on definition and classification of cancer cachexia also provides diagnostic criteria for cachexia that include weight loss or loss of skeletal muscle mass (3) . These assessments enrich understanding of the variability in nutritional status, and have the potential to alter cancer treatment paradigms for improved health outcomes (11,12) .
There is scarce information on the burden of malnutrition and cachexia among cancer out-patients in Kenya to inform decision making and policy (1) . In view of the lack of information and the recent advances in screening and diagnosis of malnutrition and cachexia, this study sought to establish the risk of malnutrition, and factors associated with the occurrence of malnutrition and cachexia, with the ultimate aim of informing nutrition programmes for cancer management in Kenya and beyond.

Study design and population
This was a facility-based cross-sectional study among cancer out-patients seeking treatment at Kenyatta National Hospital (KNH) and Texas Cancer Centre (TCC) in Nairobi, Kenya. KNH is the largest public referral hospital in Kenya. The KNH cancer treatment centre receives more than 22 000 cancer patients annually (13) . TCC, on the other hand, is a private facility that receives approximately 1092 cancer patients annually (14) .

Inclusion criteria
The study included all out-patients with confirmed stage I-IV cancer disease, aged above 18 years and receiving treatment (radiotherapy, chemotherapy, surgery or hormonal) in the two selected cancer treatment sites.

Sample size estimation and allocation
Prevalence of malnutrition among cancer patients in Kenya remains unknown. Using the Fischer formula (15) , a sample size of 512 participants was calculated factoring in a 10 % nonresponse rate. The calculated sample size was distributed between the two facilities using the square root allocation method to ensure appropriate representation of the two facilities based on the number of patients received per facility per year. A systematic random sampling method was used in recruiting participants until the desired sample size was achieved. This entailed obtaining the total number of patients expected in each facility from the hospital records and selecting a random starting point. The sampling interval was determined by dividing the total number expected by the desired sample size.

Recruitment and training of research assistants
Research assistants with a medical background were recruited and trained on the protocol, study procedures including the consenting process, study tools and on proper data collection using pretested data collection tools. Training and certification were done to harmonise data collection methods between researchers. This was followed by piloting of the study tools and definitive data collection.

Assessments
Sociodemographic assessments. Data collection was done between June and July 2016. Interviews were carried out and information on socio-economic and demographic characteristics collected using a pretested structured questionnaire. Education level was categorised into four categories: none, primary (1-8 years), secondary (9-14 years), and tertiary (>14 years). Occupation was classified as formal, non-formal, unemployed and retirees.
Anthropometric assessments. Weight and height were measured using calibrated Seca 762 classic mechanical medical weighing scales and UNICEF standard height boards, and findings recorded to the nearest 0·1 kg and 0·1 cm, respectively. In measuring height, participants were asked to stand upright, without head gear and shoes, arms hanging loosely at the sides, feet slightly apart, and with the back of the heels and head against the stadiometer. In measuring weight, the scale was placed on a flat surface and zero checked before every measurement. Participants were weighed without shoes while wearing minimal clothing. Weight and height measurements were used to compute BMI using the formula of BMI = weight (kg)/height (m 2 ). Information on weight history was retrieved from patients' medical records and used to calculate changes in weight. The current weight was deducted from the previous weight to calculate the amount of weight lost.
Body composition assessments. Total body fat, body water and lean mass percentages were estimated using a commercially available single-frequency four-electrode bioimpedance analyser system (Bodystat 1500). Tetrapolar hand to foot measurements were performed in a supine position lasting for 15 min. The gel-filled electrodes were placed on the dorsal surfaces of the right hand and foot, distal ones being, respectively, proximal to the metacarpal and metatarsal phalangeal joints in accordance with standard tetrapolar electrode placement to minimise gap impedence (16) . The instrument recorded whole body impedance from the hands to the feet by applying an electric alternating current flux of 0·8 A at an operating frequency of 50 kHz. Fat-free mass index (FFMI) was calculated using the formula FFMI = lean body mass (kg)/height (m 2 ).
Clinical and dietary assessments. Details on participants' diagnosis (clinical history, types of cancer and clinical stages) and treatment were retrieved from patient files. Information on symptoms such as lack of appetite, feeling nauseated or vomiting was self-reported. Dietary diversity was assessed using individual Food Variety Scores (FVS) that collect data on specific food items consumed (17) . Participants provided details on the variety of foods consumed 7 d and 24 h prior to the interview. Each food item was given a score of 1 if consumed at least once over 24 h and the 7 d period regardless of the frequency. The FVS were classified as either adequate by creating a cut-off point where the mean of the upper tercile was used to classify the respondent as either having adequate or inadequate dietary diversity.
Risk of malnutrition. Participants were screened using the MUST to identify those at risk of malnutrition. The overall risk was an aggregate of BMI, weight loss and acute disease scores whose cut-off points and interpretation of the data were done in accordance with MUST guidelines (2) . BMI cut-offs of >20, 18·5-20 and <18·5 kg/m 2 were scored as 0, 1 and 2, respectively. Unplanned weight loss of <5, 5-10 and >10 % in the past 3-6 months was scored as 0, 1 and 2, respectively. The acute disease effect did not apply to the out-patients and was scored a zero. A score aggregate of 0, 1 and 2 represented low, medium and high risk of malnutrition, respectively.
Diagnosis of malnutrition. Diagnosis of malnutrition was done in accordance with the ESPEN criteria for the diagnosis of malnutrition (10) . The criteria provide for the use of low BMI (<18·5 kg/m 2 ), or combined finding of unintentional weight loss of >5 % over 3 months and at least one of either a reduced BMI of <20 kg/m 2 or <22 kg/ m 2 in subjects younger and older than 70 years, respectively, or a low FFMI of <15 kg/m 2 and <17 kg/m 2 in females and males, respectively, to diagnose malnutrition. Similar to ESPEN, the local criterion uses low BMI in diagnosing malnutrition; hence further bivariate and multivariate analyses were done using low BMI.
Diagnosis of cachexia. Diagnosis of cachexia was in accordance with the Fearon criteria for the assessment of cachexia (3) . Diagnosis of cancer cachexia was based on BMI < 20 kg/m 2 and any degree of weight loss >2 %. The local definition of cachexia is >5 % weight loss/month or >10 % weight loss over 6 months (7) .
Information from the MUST will be used to identify patients at risk of malnutrition for targeted intervention. In addition to information on malnutrition based on low BMI, differences in malnutrition based on weight loss and either a reduced age-specific BMI or low sex-specific FFMI diagnostic criteria will help inform clinical practice, the development of guidelines and appropriate care plans for cancer out-patients in Kenya.
Ethical considerations. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Kenya Medical Research Institute (KEMRI) Scientific and Ethics Review Unit (reference: KEMRI/SERU/CPHR/ 001/3026) and the KNH Ethics and Research Review Committee (registration certificate: P462/07/2015). Written informed consent was obtained from all study subjects.
Data management and statistical analysis. Quantitative data were double-entered using the Microsoft Access application. Clean and validated data were exported to SPSS version 20.0 statistical software (IBM Corp.) for data analysis. Exploratory data analysis techniques were used at the initial stage of analysis to uncover the data distribution characteristics of continuous variables and identify outliers. Descriptive statistics such as proportions and frequency distributions were used to summarise categorical variables and measures of central tendency and dispersion for continuous variables. Pearson's χ 2 test or Fisher's exact test, where applicable, was used to assess the relationship between dependent and independent categorical variables. To test for association between independent continuous variables and dependent categorical variables, the unpaired Student's t test for normally distributed continuous variables and the Mann-Whitney U test for non-normally distributed continuous variables were used for dichotomous dependent categorical variable. For dependent categorical variables with more than two categories, one-way ANOVA for normally distributed continuous variables and Kruskal-Wallis one-way ANOVA for non-normally distributed continuous variables were carried out. The threshold for statistical significance was set at α = 0·05 (two-sided). Post hoc grouping of study participants was done during analysis since this was a comparative study.
The overall weight loss >5 % over 3 months was 18·2 % and low FFMI 43·1 %. Malnutrition based on the combined findings of unintentional weight loss of >5 % over 3 months together with age-specific reduced BMI was 8·6 %, and with sex-specific low FFMI was 11·5 %.

Discussion
The present study sought to establish the prevalence and factors associated with malnutrition and cachexia among cancer UD, undetermined; MUST, Malnutrition Universal Screening Tool. * Risk of malnutrition assessed using the MUST ((BMI cut-offs of >20, 18·5-20 and <18·5 kg/m 2 representing low, medium and high risk, respectively) + (unplanned weight loss of <5 %, 5-10 % and >10 % in the past 3-6 months representing low, medium and high risk, respectively)).
out-patients in Nairobi, Kenya. A third of the patients were found to be at risk of malnutrition, while 13·4 and 14·1 % had malnutrition and cachexia, respectively. Our findings confirm previous results that show malnutrition and cachexia as common nutritional challenges among cancer patients, and that negatively make an impact on the response to cancer treatment, quality of life and overall prognosis (18)(19)(20)(21) . Malnutrition and cachexia are poor prognostic factors and should be prevented or detected and managed early enough (18,22) . Those found at risk of malnutrition should therefore be closely monitored to avert progression to frank malnutrition. Mechanisms should also be put in place to identify patients at risk of cachexia for the initiation of preventive interventions (3) . The leading cancers were breast and cervical cancer in females and prostate and oesophageal cancer in males. Our findings agree with the Global Burden of Disease 2016 report that showed prostate and breast cancer as the leading types of cancer in men and women, respectively (23) . Up to 24 % of males and close to 10 % of females presented with oesophageal cancer that has been found to contribute to some of the highest incidences of malnutrition (21) . This might partly explain the significant association observed between malnutrition and cancer of digestive organs in the present study.
Up to 28·2 % presented with stage III cancer disease. However, the odds of being at risk and having malnutrition were higher among patients with stage II cancer disease. Previous studies have shown that many cancer patients in Africa are diagnosed with advanced cancer and easily become malnourished (24,25) . In light of the present findings on increased odds of malnutrition in early cancer and the fact that 81 % of the patients do not receive nutrition services, there is need for enhanced early screening that can be achieved by integrating nutrition services across the four levels of health care in Kenya.
Many factors can modify nutritional status in cancer patients including nausea, vomiting, decreased energy intake or oncologic treatments resulting in malnutrition (26,27) . More than 50 % Previous studies have pointed out alterations that occur in components of energy expenditure that can contribute to malnutrition if not compensated for by increased energy intake (27) . Thus, a holistic approach is required to address the physical, social, psychological and nutritional needs of cancer outpatients to reduce the consequences of cancer-associated nutritional decline (27,28) . BMI is the most common and easily available anthropometric assessment method despite its limitation. The local diagnostic criterion is BMI < 18 kg/m 2 , which is similar to the ESPEN criteria. The low BMI criterion was able to identify a higher proportion of patients with malnutrition compared with the combined finding of unintentional weight loss and either a reduced age-specific BMI or a low sex-specific FFMI criteria also recommended by ESPEN. However, the observed low FFMI of 43·1 % highlights the need to incorporate body composition assessments in nutritional assessments. Body composition is not routinely measured in Kenyan public hospitals despite the morbidity risk associated with increased loss of muscle mass. Given its relevance, assessment of body composition should be supported by building the necessary capacity and infrastructure (29)(30)(31)(32)(33)(34)(35)(36) . On the other hand, the Fearon criteria for diagnosis of cachexia identified 14·1 % cachexia compared with 8·5 % using the local criteria. There is need therefore to evaluate the current local diagnostic criteria for cachexia while being cognizant of the local clinical practice and health system structures.
In this study, malnutrition and cachexia were significantly associated with treatment type and site of cancer. A previous study in Kenya among cervical cancer patients undergoing radiotherapy projected a 2-year survival of <20 % (37) . Cancer management therefore requires a multimodal approach, and the Kenyan Government recommends establishment of 'multidisciplinary tumour boards' to consider all aspects of the patients' conditions (7) . A multidisciplinary team is required to consider the implications beyond patients' dietary needs to patients' nutritional and functional state throughout the prolonged course of treatment, and is best started earlier rather than later (20,35,38) . The role of nutrition as a therapy that complements basic treatment and improves treatment outcomes needs to be appreciated (22,38,39) . Inclusion of nutritionists in mainstream oncology practice has the potential to improve early detection and screening of malnutrition, access to nutrition services and communication for improved patient management (22,40) . Awareness and consideration of nutritional issues among oncologists and other related health disciplines are also vital to the success of nutrition support in cancer care (24,36) . In settings with limited nutritionists/dietitians, provision of protocols for screening can enable task shifting that has been shown to be effective and affordable in improving access to healthcare (41) . Findings on sociodemographic and behavioural factors add insights to the likely influence of social factors in the management of cancer. Disease-aggravating factors such as cigarette smoking and alcohol consumption were prevalent and significantly associated with the presence of malnutrition. Cigarette smoking was a predictor of cachexia. Public sensitisation on the ill-effects of such risk factors and raising the suspicion index among healthcare workers is important. The Global Burden of Disease 2016 report (23) showed that mortality due to all cancers is largely located on the African continent and expected to increase due to the epidemiological transition. Continued research and surveillance therefore remain critical. Further research is needed to understand the nutrition requirements in various care settings, and explore novel cancer preventive and control strategies (19,38) . The emergence of population-based cancer registries in East Africa should be supported, and so should prioritisation of cancer control programmes within the health-care systems (19,(42)(43)(44) .

Conclusions
Malnutrition and cachexia remain a challenge among cancer out-patients in Nairobi, Kenya, yet the provision of nutritional services remains low. The use of the MUST as a screening tool at the first point of care has the potential to identify patients at risk of malnutrition and who are likely to benefit from appropriate interventions for better health outcomes. The local diagnostic criteria for cachexia should be reassessed to ensure correct identification of patients at risk of cachexia for initiation of appropriate care. There is need for inclusion of body composition assessments in Kenyan public hospitals for refined nutritional assessments that will in turn inform patient management plans. Age is a common predictor of malnutrition and cachexia in addition to site of cancer for malnutrition and cigarette smoking for cachexia. Tailored education and sensitisation campaigns should therefore be encouraged and multidisciplinary partnerships formed to provide holistic cancer care.

Limitations
The study did not measure inflammatory markers or ascertain the aetiology of malnutrition observed. The contribution of inflammation and other catabolic drivers to the development and progression of cachexia was also not assessed. This was partly due to the design of the study and the choice of outpatients in whom tests for inflammatory markers are not routinely done in the selected study facilities.

Acknowledgements
The study team acknowledges and appreciates the input of various partners who contributed to the development, planning and execution of the study. We would like to express our heartfelt appreciation to experts from the Kenya Medical Research Institute (KEMRI), Kenya Cancer Association and KNH who participated in the planning and implementation processes. Our earnest gratitude goes to the management of collaborating institutions for providing the necessary administrative, technical and logistical support. We would like to thank the following facility staff for their dedication and support during data collection: Lydia Wataka and Onesmus Kariuki of TCC and Irene Karanja and Carol Wambua of KNH. We also thank the following research assistants for their hard work and commitment to this study: Rodgers Ochieng, Melvin Obuya, Schiller Mbuka, Peter Shigholi, Dorine Njeri, Mercy Chepkirui, Paul Odhacha, Isaac Kisiangani and Fidel Muendo. We wish to thank all the study participants and caregivers for their cooperation. This work was supported by KEMRI (grant no. IRG 176/6) and the Coca-Cola Company (grant no. E-1176). KEMRI and the Coca-Cola Company had no role in the design, analysis or writing of this article. L. U. K., Z. N. B., Y. O., R. M. and C. F. L. M. contributed to study conceptualisation, design, data collection, analysis and manuscript writing. E. M. and M. M. contributed to the study design, data processing, data analysis and manuscript writing. A. K. participated in study conceptualisation, data analysis and manuscript writing. V. T. and C. N. participated in data collection and manuscript writing.
There were no conflicts of interest.