Effective nutritional screening, nutritional care planning and nutritional support are essential in all settings, and there is no doubt that a health service seeking to increase safety and clinical effectiveness must take nutritional care seriously. Screening and early detection of malnutrition is crucial in identifying patients at nutritional risk. There is a high prevalence of malnutrition in hospitalized patients undergoing treatment for colorectal cancer.
To synthesize the best available evidence regarding the diagnostic test accuracy of nutritional tools (sensitivity and specificity) used to identify malnutrition (specifically undernutrition) in patients with colorectal cancer (such as the Malnutrition Screening Tool and Nutritional Risk Index) compared to reference tests (such as the Subjective Global Assessment [SGA] or Patient Generated Subjective Global Assessment [PG-SGA]).
Types of participants
Patients with colorectal cancer requiring either (or all) surgery, chemotherapy and/or radiotherapy in secondary care.
Focus of the review
The diagnostic test accuracy of validated assessment tools/instruments (such as the Malnutrition Screening Tool and Nutritional Risk Index) in the diagnosis of malnutrition (specifically under-nutrition) in patients with colorectal cancer, relative to reference tests (Subjective Global Assessment or Patient Generated Subjective Global Assessment).
Types of studies
Diagnostic test accuracy studies regardless of study design.
Studies published in English, German, Danish, Swedish and Norwegian were considered for inclusion in this review. Databases were searched from their inception to April 2014.
Methodological quality was determined using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) checklist.
Data was collected using the data extraction form: the Standards for Reporting Studies of Diagnostic Accuracy (STARD) checklist for the reporting of studies of diagnostic accuracy.
The accuracy of diagnostic tests is presented in terms of sensitivity, specificity, positive and negative predictive values. In addition, the positive likelihood ratio (LR+) (sensitivity/ [1 - specificity]) and negative likelihood ratio (LR-) (1 - sensitivity)/ specificity), were also calculated and presented in this review to provide information about the likelihood that a given test result would be expected when the target condition is present compared with the likelihood that the same result would be expected when the condition is absent. Not all trials reported true positive (TP), true negative (TN), false positive (FP) and false negative (FN) rates, therefore these rates were calculated based on the data in the published papers. A two-by-two truth table was reconstructed for each study, and sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), LR+ and LR- were calculated for each study. A summary receiver operator characteristics (SROC) curve was constructed to determine the relationship between sensitivity and specificity, and the area under the SROC curve which measured the usefulness of a test was calculated.
Meta-analysis was not considered appropriate, therefore data was synthesized in a narrative summary.
One study evaluated the Malnutrition Screening Tool (MST) against the reference standard Patient-Generated Subjective Global Assessment (PG-SGA). The sensitivity was 56% and the specificity 84%. The positive likelihood ratio (LR+) was 3.100, negative likelihood ratio (LR-) was 0.59, the diagnostic odds ratio (CI 95%) was 5.20 (1.09–24.90) and the Area Under the Curve (AUC) represents only a poor to fair diagnostic test accuracy. A total of two studies evaluated the diagnostic accuracy of Malnutrition Universal Screening Tool (MUST) (index test) compared to both Subjective Global Assessment (SGA) (reference standard) and PG-SGA (reference standard) in patients with colorectal cancer. In MUST vs SGA the sensitivity of the tool was 96%, specificity was 75%, LR+ 3.826, LR- 0.058, diagnostic OR (CI 95%) 66.00 (6.61–659.24) and AUC represented excellent diagnostic accuracy. In MUST vs PG-SGA the sensitivity of the tool was 72%, specificity 48.9%, LR+ 1.382, LR- 0.579, diagnostic OR (CI 95%) 2.39 (0.87–6.58) and AUC indicated that the tool failed as a diagnostic test to identify patients with colorectal cancer at nutritional risk. The Nutrition Risk Index (NRI) was compared to SGA representing a sensitivity of 95.2%, specificity of 62.5%, LR+ 2.521, LR- 0.087, diagnostic OR (CI 95%) 28.89 (6.93–120.40) and AUC represented good diagnostic accuracy. In regard to NRI vs PG-SGA the sensitivity of the tool was 68%, specificity 64%, LR+ 1.947, LR- 0.487, diagnostic OR (CI 95%) 4.00 (1.23–13.01) and AUC indicated poor diagnostic test accuracy.
There are no single, specific tools used to screen or assess the nutritional status of colorectal cancer patients. All tools showed varied diagnostic accuracies when compared to the reference standards SGA and PG-SGA. Hence clinical judgment combined with perhaps the SGA or PG-SGA should play a major role.
Implications for practice
The PG-SGA offers several advantages over the SGA tool: 1) the patient completes the medical history component, thereby decreasing the amount of time involved; 2) it contains more nutrition impact symptoms, which are important to the patient with cancer; and 3) it has a scoring system that allows patients to be triaged for nutritional intervention. Therefore, the PG-SGA could be used as a nutrition assessment tool as it allows quick identification and prioritization of colorectal cancer patients with malnutrition32 in combination with other parameters.
Implications for research
This systematic review highlights the need for the following:
- Further studies needs to investigate the diagnostic accuracy of already existing nutritional screening tools in the context of colorectal cancer patients.
- If new screenings tools are developed, they should be developed and validated in the specific clinical context within the same patient population (colorectal cancer patients).