A detailed description of the design and methods of the case-control study is available elsewhere.6
Briefly, the study was carried out in the eastern part of Golestan province, northern Iran. Case participants were recruited between December 2003 and March 2007 at Atrak Clinic, the only gastrointestinal specialty clinic in the study area. Doctors in the study catchment area were asked to refer any patients suspected of having cancers of the upper gastrointestinal tract to Atrak Clinic, where all the patients underwent video oesophagogastroduodenoscopy. Patients with suspicious lesions in the oesophagus during endoscopy were invited to participate in the study; nearly all of those who were invited participated. Fixed biopsy samples were sent to the Tehran University Digestive Disease Research Center, where they were examined by experienced pathologists. Only newly diagnosed patients with histologically confirmed oesophageal cancer were finally included as case participants in the study. Cancer registry data show that nearly 70% of all incident cases of oesophageal cancer in this area were enrolled in the case-control study (unpublished data). For each case participant we tried to select two population based controls matched to the cases on place of residence, age (SD 2 years), and sex, using data from a family health census that is carried out by the Iranian primary healthcare system and is updated annually. A roster of eligible controls was prepared for each case, from which the controls were selected at random. If the first randomly selected control did not agree to participate, the second person was approached, and so on. Most of the selected controls agreed to participate; of the enrolled controls 77% were the first randomly selected patients and 11% the second. In nearly all instances in which selected patients were not enrolled, the reason for not participating was the absence of an eligible control at the time of invitation. Written consent was obtained from all participants.
Using a structured questionnaire trained interviewers collected information on personal characteristics, several other potential confounders of interest, and tea drinking temperature in face to face interviews. The participants were asked whether they drank tea and, if so, whether they usually drank it warm, lukewarm, hot, or very hot. They were also questioned about the interval (in minutes) between tea being poured and drunk. We asked separately about consumption of black and green tea, using a food frequency questionnaire specifically designed for this population18
; this questionnaire was administered by a trained nutritionist. For this we asked participants about the usual frequency of drinking tea and the volume of the cups usually used. We provided photographs of five types of cups and mugs that are commonly used in the study area. A limited number of staff carried out the interviews and no proxies were used. For cases, the questions referred to the period before symptoms began. Case participants were interviewed on the same day that they underwent diagnostic endoscopy of the upper gastrointestinal tract at Atrak Clinic.
The Golestan Cohort Study is a prospective study that recruited 50 045 adults, aged 40-75, from Golestan province between January 2004 and June 2008. A total of 16 599 inhabitants in the specified age range were selected randomly from Gonbad City, the main urban area in eastern Golestan, by systematic clustering based on the household number. In rural areas, all residents of villages in the study catchment area in the specified age range were invited to participate. Eligible participants were enrolled in the study unless they were unwilling to participate at any stage for any reason, were temporary residents, or had a current or previous diagnosis of an upper gastrointestinal cancer. In urban areas, 10 032 participants were enrolled, with participation rates of about 70% for women and 50% for men. In rural areas, 40 013 participants were enrolled from 326 villages, with participation rates of 84% for women and 70% for men.
The same questions for tea temperature were asked as in the case-control study. In addition we measured the temperature of tea drunk by the participants. To achieve this we prepared a fresh cup of tea for each participant and measured the temperature of the tea using a digital thermometer. When the temperature was 75°C we asked the participants to sip the tea and say whether that was the temperature at which they usually drank tea. If not, the tea was allowed to cool to 70°C and the question was asked again. This procedure was repeated, at 5°C intervals, until the temperature at which tea was usually drunk was reached. This method for measuring the temperature of tea showed good reliability when tested in the pilot phase of the cohort study.19
The usual interval between tea being poured and drunk was categorised as four or more minutes, 2-3 minutes, and less than two minutes. Within the case-control study the amount of black tea consumed each day (millilitres) was categorised into fifths. For green tea, however, because only a small number of participants drank this kind of tea, we present only the frequency of consumption. Data on tea temperature were available for over 99% of cases and of controls. The amount of tea, however, was available for 89% of the cases and 67% of the controls. We excluded participants with missing data on a variable from the corresponding analysis. Conditional logistic regression was used to calculate odds ratios and corresponding 95% confidence intervals. By design, case and control participants were matched for age, sex, and place of residence. We used logistic regression models to adjust for potential confounders, including ethnicity, daily vegetable intake, alcohol consumption, tobacco or opium use, duration of residence in rural areas, education level, and car ownership, as indicators of socioeconomic status, and other tea drinking variables. P values for trend were obtained from adjusted conditional logistic regression models by assigning consecutive numbers to categories within each categorical variable. We tested the agreement between tea temperature categories and the interval between tea being poured and drunk by weighted κ statistics and Spearman’s rank correlation coefficients.
Less than 3% of the cohort participants had one or more missing values in tea drinking variables. These participants were excluded from the current analyses. We calculated means and standard deviations for daily intake of black and green tea among the cohort participants and the percentage of participants who drank these two kinds of tea daily, weekly, or less. In addition, we examined the validity of the questionnaire data on tea temperature within the cohort study. For this we categorised tea temperature measurements as less than 65°C, 65-69°C, and 70°C or more, because their distribution was close to that of the categorical variable used in the questionnaire. We then compared them with the questionnaire data, using weighted κ statistics and Spearman’s rank correlation coefficients.
The methods of sample size calculation for both studies are presented in the web extra. Throughout the analyses we considered two sided P values <0.05 as significant. All statistical analyses were done using Stata version 10.0 software.