The purpose of this study was to determine the intra- and inter-observer agreement between paediatric head and neck malignancy findings of single CT procedures and double CT procedures for children aged 15 years and below at Uganda Cancer Institute. Majority of the images selected in the study belonged to children in the age range of 3 to 13 years with a median age of 9 years. We determined both inter-observer agreement as well as intra-observer agreement in the diagnosis of the tumours using the various CT protocols.

When compared with previous literature, this study resonates with what has been previously reported by Just da Costa e Silva et al. [14]. The observed substantial agreement obtained on contrast enhanced CT only and unenhanced CT only procedures demonstrated that using either procedure provided the same value in regard to determining the primary tumour location. These findings resonate with reports in documented literature by Brennan et al. (2006) and Stambuk et al. (2005) [18, 19].

This demonstrated that unenhanced CT only procedures did not offer a significant edge over contrast enhanced CT only procedures for detection of calcifications. This discordance was neither reflected in the diagnosis where a substantial agreement was obtained for contrast enhanced CT only procedures, unenhanced CT only procedures and double CT procedures. The findings from this study do resonate with the findings reported in previous literature [14] where they reported that contrast enhanced CT procedures had a sensitivity greater than 70% and specificity of 100% at detecting calcifications in paediatric abdominal malignancies.

Findings from this study demonstrated that there was better agreement on presence of lymphadenopathy with unenhanced CT only procedures as opposed to contrast enhanced CT only procedures which was different from reports in literature [20]. The explanation for these findings is unclear however may be explained by the intra-observer agreement findings for the different observers when contrast enhanced CT only procedures are compared to the findings on the double CT procedures.

This demonstrated that using both unenhanced and contrast CT procedures did not provide an added advantage over using either contrast enhanced or unenhanced CT only procedures only. This is similar to the results presented by Just da Costa e Silva et al. [14] where diagnoses made on contrast enhanced CT only procedures were compared to histopathology results. For practice in our setting, this implies that contrast enhanced CT only procedures are sufficient in diagnosing PHNM.

This demonstrated that either unenhanced or contrast enhanced CT only procedures are sufficient in determining the primary location of a tumour and using both does not offer an advantage over use of only one. The implication of this finding for clinical practice in our study is that patients with contraindications to contrast enhanced CT procedures and lack of access to MRI will still benefit from unenhanced CT procedures.

Contrast enhanced CT only and unenhanced CT only procedures were comparable in terms of tumour calcification identification which is similar to what was reported by Just da Costa e Silva et al. [14]. The possible explanation for the differences among the observers could be due to lack of documentation of presence of calcifications while reporting the images which pertains to the presence of intrinsic differences among the observers [21, 22]. This may however also allude to the lack of importance of presence or absence of tumour calcifications when diagnosing PHNM. Indeed, this compares with what has been previously reported in literature by Lloyd et at. (2010) where presence of calcifications was only important in diagnosing retinoblastoma [23].

Observers A and C had an almost perfect intra-observer agreement when contrast enhanced CT only procedures were compared to double CT procedures as opposed to a substantial agreement when unenhanced CT only procedures were compared to double CT procedures, inferring that contrast enhanced CT only procedures are the preferred for demonstrating lymphadenopathy. Reports by Chong et al. [24], Restrepo et al. [25] and Som et al. [26] had similar conclusions. The fair intra-observer agreement by observer B may explain the moderate inter-observer agreement for lymphadenopathy demonstrated on contrast enhanced CT only procedures. Although observer A and C demonstrated a substantial agreement when unenhanced CT only procedures and double CT procedures were compared, metastatic spread to lymph nodes may occur in normal sized lymph nodes and contrast enhanced CT procedures provide an advantage as they delineate increased and heterogeneous enhancement of the lymph nodes unlike unenhanced CT [27].

The apparent differences could be explained by intrinsic differences between the different observers and their years of experience [28, 29]. For observers B and C, there was no advantage offered by use of either contrast enhanced CT only procedures over use of unenhanced CT only procedures in making a diagnosis. Observer A, however, had a better agreement with use of contrast enhanced CT only procedures as opposed to unenhanced CT only procedures. This finding, which is similar to reports in literature, demonstrates that contrast enhanced CT only procedures have an advantage over unenhanced CT only studies in PHNM [6, 30, 31].

Findings from this study suggest key implications for practice particularly that in evaluation for paediatric head and neck malignancies, in a bid to reduce radiation exposure to patients, contrast enhanced only CT procedures are sufficient. The paediatric patients do not require unenhanced CT procedures for the sole purpose of identifying the tumour calcifications as radiologists are able to detect them on contrast enhanced CT only procedures. In consideration of these findings, radiology residents should be trained to report on paediatric head and neck malignancies with contrast enhanced CT only procedures.

Generally, the findings in this study do agree with previous literature, but the strength of the study lies in its being conducted in a resource-limited setting whose findings can be adaptable to many other resource-limited contexts. Another strength lies in using three observers of varying experiences, which offered a rich context to get opinion from different radiologists hence increasing the validity of the findings.

Despite the observed strengths of the study, there were some limitations. For example, all observers knew that the final diagnosis was a malignancy, which may have accounted for the substantial agreement on all protocols. Agreement for the diagnoses that the radiologists mentioned was determined only from the first diagnosis listed; however, in clinical practice; radiologists make a list of differential diagnoses, which are all considered. Intra- and inter-observer agreement on tumour size was not evaluated due to missing data from the observers. The study used a structured reporting template, as opposed to a checklist, which may have increased the chances of missing data. However, the study still provides useful findings that can inform clinical practice in similar settings. We do suggest more research particularly a level II phase III (multi-centre multi-observer) study, as suggested by Fryback and Thronbury [32], with the use of contrast enhanced paediatric head and neck CT only procedures and correlation with histopathology results is recommended to determine the generalizability of these results.