We thank Dr. Friborg and colleagues for their thoughtful comments on our recent article1 in which we reported that “the SIRs for other malignancies examined in the Cantonese population were less than 1 relative to the Hong Kong and Shanghai reference population.”2 Below, we discuss the two weaknesses related to this statement in terms of the statistical methods used in our analysis, as raised by Dr. Friborg.
First, Dr. Friborg and colleagues correctly point out that the information on malignant diseases in relatives was collected by interview with the probands and therefore was based on recall. They questioned whether the comparison between information based on recall and the data from the Hong Kong Cancer Registry was appropriate and whether the two data sources were comparable in terms of their completeness. To address this concern, we would like to reveal more details on how the data were collected.
For the purposes of quality control, we performed a secondary investigation of 100 pedigrees selected randomly from a cohort that included 2252 pedigrees. It is noteworthy that the interviewers for the second study were different from the ones who conducted the first interview with these 100 probands. The second interview team was composed of 14 medical students trained at the Cancer Center of Sun Yat-sen University (Guangzhou, China). These students visited all 100 probands at their residences. Although it was difficult to obtain complete data on information such as the precise date of diagnosis, we administered a detailed questionnaire that included items on clinicopathologic diagnosis, surgical history as a result of the malignancy, and whether the relatives had received either radiotherapy or chemotherapy. If the proband responded to one of the above four items in the affirmative, we coded his/her relative as a case. When we compared the results from both interviews using a chi-square test, no significant difference was found.
We interviewed only first-degree relatives (n = 718). In the study population examined, first-degree relatives and probands tend to live in similar areas. Therefore, the information we collected is likely to be more reliable than that provided by more distantly related relatives. We believe that it is reasonable to use the Hong Kong Cancer Registry as a reference population, although we indicated in our previous publication that the comparison was rather crude. Our goal is to eventually establish a cancer registry that encompasses all areas of Guangzhou.
We agree with Dr. Friborg that, due to the extended period covered by our comparisons, we are likely to have overestimated the risks of other malignancies among relatives of probands. Increases in lung and breast carcinoma risk over the last 20–30 years could have led to this overestimation and, therefore, to an underestimation of the standardized incidence ratio. Therefore, our results regarding other malignancies must be interpreted with caution. We believe that our estimation of nasopharyngeal carcinoma (NPC) risk among relatives is accurate, because the temporal trend in incidence runs counter to the observed risk. Over the last 30 years, the NPC rate appears to have remained stable according to the Sihui Cancer registry. (Sihui is known as a high-risk area for NPC in Guangzhou.)
Finally, we have completed a matched case–control study involving the same study cohort as the one discussed here. We hope to provide a more accurate estimation of the risk of other malignancies among relatives of probands in the near future.