Oral cancer in southern India: The influence of smoking, drinking, paan-chewing and oral hygiene

Authors


Abstract

Between 1996 and 1999 we carried out a case-control study in 3 areas in Southern India (Bangalore, Madras and Trivandrum) including 591 incident cases of cancer of the oral cavity (282 women) and 582 hospital controls (290 women), frequency-matched with cases by age and gender. Odds ratios (ORs) and 95% confidence intervals (CIs) were obtained from unconditional multiple logistic regressions and adjusted for age, gender, center, education, chewing habit and (men only) smoking and drinking habits. Low educational attainment, occupation as a farmer or manual worker and various indicators of poor oral hygiene were associated with significantly increased risk. An OR of 2.5 (95% CI 1.4–4.4) was found in men for smoking ≥ 20 bidi or equivalents versus 0/day. The OR for alcohol drinking was 2.2 (95% CI 1.4–3.3). The OR for paan chewing was more elevated among women (OR 42; 95% CI 24–76) than among men (OR 5.1; 95% CI 3.4–7.8). A similar OR was found among chewers of paan with (OR 6.1 in men and 46 in women) and without tobacco (OR 4.2 in men and 16.4 in women). Among men, 35% of oral cancer is attributable to the combination of smoking and alcohol drinking and 49% to pan-tobacco chewing. Among women, chewing and poor oral hygiene explained 95% of oral cancer. © 2002 Wiley-Liss, Inc.

Cancer of the oral cavity and pharynx is the first and third commonest cancer in Indian men and women, respectively.1 Whereas in most areas at high risk for cancer of the oral cavity other than India (e.g., central and Eastern Europe, South America), the ratios between male and female incidence rates range between 3 and 10, in India the male-to-female ratio is approximately 1 (e.g., Madras) or lower than 0.5 (Bangalore).2 Such very high incidence rates in Indian women reflect the persistent importance in India of paan chewing, a habit that is equally common in the 2 genders.3 Paan generally includes calcium hydroxide, areca nut (from the Areca catechu tree) and betel leaf (from the Piper betle vine). Tobacco and/or various spices are commonly added.4 Paan represents a cheap pharmacologically addicting stimulant, principally used by members of low social classes in South Asia. Fewer efforts have been made in Asia to discourage paan chewing than tobacco smoking,5 and only recently have links been established between paan and oral cancer that cannot be explained by the presence of tobacco.3, 6

Annual per capita consumption of cigarettes in India was maximal in the 1970s and 1980s and declined by approximately 40% in the early 1990s.7 Two nation-wide surveys8, 9 showed a somewhat lower prevalence of tobacco use in any form in 1993–1994 (23% in urban and 34% in rural areas in men and 4% and 9%, respectively, in women) than in 1987–1988 (26% and 35% in men and 6% and 11% in women, respectively). It is estimated that 150 million males and 34 million females used tobacco in India in 1996.8, 9

Relatively few case-control studies have recently addressed the impact of paan chewing and smoking on oral cancer in India,10–12 and information on women and on risk factors other than smoking or chewing is scanty.12

The present case-control study was conducted in 3 areas of Southern India in order to evaluate the relative importance of smoking, alcohol drinking and paan chewing, with or without tobacco, on cancer of the oral cavity in men and women and the modifying effect, if any, of various indicators of oral hygiene. Our study is part of an international study on oral cancer coordinated by the International Agency for Research on Cancer and carried out also in Italy,13 Cuba,14 Spain, Northern Ireland, Poland, Canada, Sudan and Australia, whose major aim is to evaluate the role of human papillomavirus (HPV).15 In fact, many case-series and a few case-control studies have raised the possibility that HPV may be causally associated with a subset of head and neck cancer, most notably tonsillar carcinoma.15

MATERIAL AND METHODS

Between July 1996 and May 1999 the incident cases of cancer of the oral cavity were identified in 3 Indian centers: Bangalore, Madras and Trivandrum, Southern India. Among identified cases, 20 were too sick to be interviewed. A total of 309 male cases (median age 56; range 22–85 years) and 282 female cases (median age 58; range 18–87 years) were thus enrolled (Table I). Twenty-nine cases (24 males) of oropharynx cancer were also interviewed but were not included in the analysis. The distribution by cancer stage among men was as follows: stage 1, 16%; stage 2, 18%; stage 3, 28%; and stage 4, 38%. Among women, it was as follows: stage 1, 8%; stage 2, 14%; stage 3, 38%; and stage 4, 40%. All cases had their interview and oral examination before any cancer treatment.

Table I. Distribution and Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) for Cancer of the Oral Cavity by Selected Characteristics and Gender (591 Cases and 582 Controls, India, 1996–1999)1
 MenWomen
CasesControlsOR2 (95% CI)CasesControlsOR2 (95% CI)
  • 1

    Distribution: some strata do not add up to the total because of missing values.

  • 2

    Estimates from unconditional regression equations, including terms for age, center, education and chewing habits and (men only) smoking and drinking habits.

  • 3

    Reference category.

Age (yr)
 <55126136109159
 55–64103929877
 ≥6580647554
Education (years)
 ≥731011631131121
 1–6138881.89 (1.24–2.88)45721.09 (0.44–2.69)
 070412.06 (1.21–3.49)2221045.52 (2.36–12.90)
 χmath image for trend9.83; p = 0.00220.12; p < 0.001
Number of siblings
 <335054187731
 3–458561.32 (0.72–2.43)62540.98 (0.45–2.11)
 ≥555750.84 (0.46–1.55)58940.97 (0.46–2.01)
 χmath image for trend0.47; p = 0.490.04; p = 0.85
Religion
 Hindu or buddhist319919311201201
 Christian27211.88 (0.90–3.91)10140.86 (0.22–3.37)
 Muslim25161.30 (0.60–2.83)20181.14 (0.38–3.39)
Occupation
 Clerical34489113231
 Industrial workers89822.19 (1.26–3.78)1081032.29 (0.80–6.58)
 Farmers153902.76 (1.62–4.70)57212.18 (0.62–7.66)
 Others23291.69 (0.77–3.70)1021401.50 (0.51–4.42)
Smoking habit
 Never38612712742851
 Ever2231651.77 (1.17–2.69)853.18 (0.58–17.46)
Drinking habit
 Never313723212732851
 Ever172902.18 (1.43–3.33)650.31 (0.07–1.40)

Control subjects were frequency-matched with cases by center, quinquennium of age and gender. They were all identified and interviewed in the same hospital where cases were found. In Madras and Bangalore, control subjects were identified among relatives and friends who were attending patients admitted for cancer other than oral cancer to, respectively, the Madras Cancer Institute or the Kidwai Memorial Institute of Oncology. In Trivandrum, control subjects were chosen among outpatients who attended the clinics of the Medical College Hospital or of the Regional Cancer Center but were found to be free from malignant diseases. In all 3 centers, over 90% of eligible controls accepted participation in the study. Overall, the control group included 292 men (median age 55; range 20–76 years) and 290 women (median age 52; range 18–80 years) (Table I).

Cases and controls were interviewed by social workers. The section of smoking habits included questions of smoking status (never, ex-smoker or current smokers), daily number of cigarettes, cigars or bidi smoked, age at starting and duration of the habit. Bidi is a local cigarette made by wrapping less than 0.5 g of coarse tobacco dust in a dry temburni (Diospyros melanoxylon) leaf. When estimating risk associated with tobacco smoking, 1 bidi was considered equivalent to 1 cigarette or ¼ of a cigar. The consumption of the commonest alcoholic beverages was also investigated. The alcoholic beverages used are mainly a locally fermented and distilled sap from palm trees called “toddy” (approximately 4% ethanol) and another locally brewed liquor called “arrack” (approximately 40% ethanol). Taking into account the different ethanol concentration, 1 drink corresponded to approximately 40 ml of hard liquor (arrack included), 450 ml of beer and toddy, and 150 ml of wine, equivalent to 15 g of ethanol. In Bangalore, a simplified questionnaire was used for drinking habits, and study subjects could be classified as ever/never drinkers only.

The habit of paan chewing was investigated by considering the chewing status (never, ex-chewer or current chewer) before cancer onset, different kinds of products (i.e., paan with or without tobacco), number of paan consumed per day, age at starting and duration of the habit. Paan chewing involved the addition of locally cured dried tobacco leaves and/or stem in most study subjects. Never-smokers, never-drinkers and never-chewers were individuals who had abstained respectively from smoking, alcoholic beverages and chewing, lifelong. Former smokers, former drinkers and former chewers had abstained respectively from any type of smoking, chewing or drinking for at least 12 months before cancer diagnosis or interview (for controls).

Indicators of oral hygiene were self-reported by means of 9 specific questions. The number of missing teeth that had not been replaced and the general oral condition, on the basis of presence of tartar, decayed teeth and mucosal irritation, were evaluated by the interviewer through inspection of the mouth. The questionnaire also included information on sociodemographic characteristics, prior occurrence of sexually transmitted diseases and other infections, family history of cancer and a dietary questionnaire.

The present project was reviewed and approved by the Ethical Committee of IARC and the local ethical and research committees.

Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were computed for the 3 centers together using unconditional multiple logistic regression models. Men and women were assessed separately. All models included terms for center, age quinquennium, educational years and chewing habit in addition to other variables as specified. Detailed evaluation of and adjustment for smoking and drinking habits was restricted to men, since very few women reported any consumption of cigarettes or alcoholic beverages (Table I). Attributable risk fractions were computed, separately for men and women, according to a method that implies knowledge of the risk estimates and of the joint distribution of risk factors among cases only, and is therefore applicable to hospital-based case-control studies.16

RESULTS

Oral cancer cases reported significantly fewer years of education than control subjects. The difference was more marked in women (OR for 0 versus ≥7 years of education 5.5) than men (OR 2.1). Industrial manual workers and farmers were at an approximately 2-fold increased risk compared with clerical workers in either gender. Housewives represented the majority of occupations in the “other” category. A direct association also emerged between cancer risk and spouse's education (OR for 0 versus ≥ 7 years of education 1.9; 95% CI 1.1–3.4 in men and 1.6; 95% CI 0.6–4.4 in women, not shown). Number of siblings was unrelated to oral cancer risk in either gender, whereas Christian men, but not Christian women, were at a 1.9-fold greater risk than Hindus or Buddhists. Tobacco smoking was associated with oral cancer risk among men (OR 1.8) and women (OR 3.2), but less than 3% of female cases had ever smoked. Consumption of alcoholic beverages was associated with an OR of 2.2 among men, but no risk increase was detected among the few drinking women (2% of female cases).

Smoking and drinking habits in men only are considered in detail in Table II. Fifty-three percent of cases and 39% of controls were current smokers. The majority of them smoked bidi, alone or in combination with cigarettes or cigars (OR for ≥ 20 bidi or equivalent/day versus never smokers 2.5; 95% CI 1.4–4.4). Only 28 cases and 40 controls smoked cigarettes only (OR 1.1). Age at starting among current smokers was relatively late (median age starting at 20 years among both cases and controls), and it was not related to oral cancer risk. Quitting smoking was associated with a nonsignificant decline in risk compared with current smokers (OR for ≥ 10 years after quitting 0.7), but former smokers were few. Tobacco snuffing was rare (7% of male cases and 5% of controls) and not significantly associated with oral cancer risk (not shown).

Table II. Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) for Cancer of the Oral Cavity in Men by Smoking and Drinking Habits (309 Cases and 292 Controls, India, 1996–99)1
 CasesControlsOR2(95% CI)
  • 1

    Some strata do not add up to the total because of missing values.

  • 2

    Estimates from unconditional regression equations, including terms for age, center, education, smoking, drinking and chewing habits.

  • 3

    Reference category.

  • 4

    Current smokers or drinkers only.

  • 5

    Information not available for Bangalore.

  • 6

    One drink corresponds to approximately 150 ml of wine, 450 ml of beer and 40 ml of liquor (i.e., 15 g of ethanol).

Smoking habit
 Never smokers8612713
 Former smokers59501.38(0.78–2.47)
 Current smokers
  Cigarettes only28401.08(0.56–2.09)
  Cigars only8110.17(1.12–92.18)
 Bidi or equiv. (no./day)
   <2055332.04(1.10–3.79)
   ≥2073412.50(1.41–4.42)
Age started smoking (yr)4
 ≥23543413
 20–2263460.82(0.42–1.61)
 <2047350.84(0.41–1.73)
 χmath image for trend0.23; p = 0.63
Years since quit smoking4
 Current smoker16411513
 <1039330.71(0.37–1.34)
 ≥1020170.73(0.32–1.68)
 χmath image for trend1.07; p = 0.30
Drinking habit5
 Abstainers10215213
 Former drinkers65341.78(0.97–3.28)
 Current drinkers (drinks/wk)6
  <329182.17(1.00–4.69)
  3–1322132.14(0.89–5.19)
  ≥1429121.97(0.85–4.57)
 χmath image for trend6.02; p = 0.01
Age at start drinking4,5 (yr)
 ≥31261313
 23–3029122.11(0.69–6.48)
 <2325180.67(0.20–2.26)
 χmath image for trend0.08; p = 0.78
Years since quit drinking4,5
 Current drinkers844413
 <1049270.94(0.43–2.09)
 ≥101670.62(0.19–2.05)
 χmath image for trend0.36; p = 0.55

Current drinkers of alcoholic beverages were 32% among male cases, and 19% among male controls (Table II). A significant trend of increase in oral cancer risk with increasing number of drinks per week was found (χ2 = 6.0; p = 0.01). Toddy accounted for 38% of the alcohol consumption, whereas arrack and liquors such as whisky or gin represented 33 and 28%, respectively, of the total amount. Only 1% of alcohol intake came from wine and beer. Neither age at start drinking nor cessation of the habit were related to oral cancer risk.

Table III shows paan chewing habits in men and women separately. Among cases, 59% of men and 90% of women were ever-chewers (OR 5.1; 95% CI 3.4–7.8 and 42.4; 95% CI 23.8–75.6, respectively). Ninety-one percent of chewers, in both genders, reported the use of paan with tobacco (OR 6.1 in men and 45.9 in women). However, a significantly elevated risk was also found in the few subjects who reported chewing paan without tobacco (OR 4.2 in men and 16.4 in women). Among chewers of paan without tobacco, 9 male cases and 4 male controls, but no women, reported tobacco smoking. Median number of paan consumed per day was 5 in either female or male cases. A significant trend of increase in oral cancer risk by number of paan per day was seen in both genders. The OR for ≥10 paans per day was substantially greater in women (OR 112) than in men (OR 7.9). Women reported starting at an earlier age (median 20 years) than men (median 22 years), and early starting of chewing (<20 versus ≥ 25 years of age) was associated with a 5-fold elevated OR in women, but not in men. There were few former chewers. No clear decline of oral cancer risk was seen after chewing cessation in either gender.

Table III. Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) for Cancer of the Oral Cavity by Paan Chewing Habits and Gender (591 Cases and 582 Controls, India, 1996–99)1
 MenWomen
CasesControlsOR2 (95% CI)CasesControlsOR2 (95% CI)
  • 1

    Some strata do not add up to the total because of missing values.

  • 2

    Estimates from unconditional regression equations, including terms for age, center, education, chewing and (men only) smoking and drinking habits.

  • 3

    Reference category.

  • 4

    Current chewers only.

Chewing habit
 Never chewers31272321292511
 Ever chewers182605.12 (3.38–7.76)2533942.40 (23.78–75.59)
Type of paan
 With tobacco139376.10 (3.84–9.71)2223145.89 (25.02–84.14)
 Without tobacco1564.16 (1.46–11.83)14516.42 (4.77–56.48)
No. of paan/day
 Former chewers
  <528114.24 (1.87–9.63)17620.24 (6.40–63.94)
  ≥53195.77 (2.53–13.16)31360.42 (15.83–230.67)
 Current chewers
  <540183.06 (1.58–5.91)511322.10 (10.06–48.52)
  5–946128.15 (3.93–16.90)1011358.58 (26.61–128.99)
  ≥103477.91 (3.23–19.41)513112.41 (30.85–409.55)
 χmath image for trend18.37; p < 0.00171.21; p < 0.001
Age started chewing (yr)4
 ≥2535121156131
 20–2442101.53 (0.56–4.18)74121.92 (0.69–5.34)
 <202761.54 (0.47–5.02)7345.43 (1.50–19.65)
0.73; p = 0.396.86; p = 0.01
Years since quit chewing
 Current chewer3120371203291
  <1045141.02 (0.45–2.29)3160.72 (0.23–2.21)
  ≥101460.75 (0.23–2.52)1730.97 (0.23–4.11)
 χmath image for trend0.50; p = 0.480.17; p = 0.68

To elucidate the difference between genders, the influence of paan chewing was examined separately in men who, like the vast majority of women in our study, never smoked or drank alcoholic beverages (63 cases and 110 controls, not shown). ORs were more elevated (OR for ≥ 5 versus 0 paan/day 18; 95% CI 6.2–53.8) than in the total male population but were still lower than among women. When the gender-specific ORs for paan chewing were examined in 3 separate strata of education, no difference was found between male (OR 5.2) and female (OR 3.7) chewers who reported 7 years of education or more.

Various indicators of oral hygiene and dentition are shown in Table IV according to gender. Female, but not male, cases reported that they cleaned their teeth less often than controls. For this purpose, the majority of study participants, most notably women, reported using fingers (OR 1.8 in men and 3.4 in women) or other instruments (OR 3.7 in men and 2.9 in women), instead of a toothbrush. A few subjects reported using a soft wooden stick. Regular toothpaste was used by 25% of oral cancer cases and 60% of control subjects. Few cases and controls reported wearing dentures and having dental check-ups. Dental check-up seemed to be significantly protective in women (OR 0.4), but not in men. Conversely, gum bleeding (OR 2.8 and 3.4 in men and women, respectively), having 6 or more missing teeth (OR 3.9 in men and 7.6 in women) and interviewer-reported poor general oral condition (OR 4.9 in men and 6.0 in women) were associated with a significantly increased risk in both genders.

Table IV. Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) for Cancer of the Oral Cavity by Indicators of Oral Hygiene and Dentition and Gender (591 Cases and 582 Controls, India, 1996–99)1
 MenWomen
CasesControlsOR2(95% CI)CasesControlsOR2(95% CI)
  • 1

    Some strata do not add up to the total because of missing values.

  • 2

    Estimates from unconditional regression equations, including terms for age, center, education and (men only) smoking and drinking habits.

  • 3

    Reference category.

Self-reported:
 Tooth cleaning (times/day)
  ≥235360133731
  ≤12542320.96(0.59–1.59)2442173.39(1.65–6.98)
 Instrument used
  Tooth brush3961771351771
  Finger1831031.75(1.11–2.76)2361093.40(1.80–6.45)
  Other30123.65(1.50–8.84)1142.87(0.54–15.40)
 Wearing dentures
  No329627612742631
  Yes11160.86(0.35–2.06)4260.26(0.05–1.25)
 Dental check-ups
  Never325221712461981
  Yes52720.89(0.56–1.42)31880.41(0.19–0.87)
 Gum bleeding
  No319923811241981
  Yes108532.83(1.71–4.68)154923.35(1.82–6.15)
Interviewer-reported
 Missing teeth
  ≤5316123511142291
  >5145563.89(2.46–6.17)164607.61(3.89–14.88)
 General oral condition
  Good or average31272321682181
  Poor177584.90(3.09–7.78)209725.99(3.00–11.96)

The combined effects of chewing with smoking, alcohol drinking and toothbrush use (as an indicator of oral hygiene) are shown in Table V, for men only. Men who smoked 20 bidi or equivalents per day or more and chewed paan had a 6.7-fold (95% CI 2.5–18.3) increased oral cancer risk. This OR is consistent with a significant negative interaction of smoking and chewing on a multiplicative scale (χ2 = 7.27; p < 0.05). Conversely, the combined effects of chewing and drinking (OR 8.6) and chewing and no use of a toothbrush (OR 11.8) show no significant departure from risk-product multiplicativity.

Table V. Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) of Oral Cavity Cancer According to Various Combinations of Chewing and Smoking, Drinking and Oral Hygiene in Men (309 Cases and 292 Controls, India, 1996–99)1
 Paan chewing
NeverCurrent chewers
Cases/controlsOR2(95% CI)Cases/controlsOR2(95% CI)
  • 1

    Some strata do not add up to the total because of missing values.

  • 2

    Estimates from unconditional regression equations, including terms for age, center, education, oral hygiene, chewing and smoking and drinking habits, as appropriate.

  • 3

    Reference category.

Tobacco smoking
 Never smokers25/1061349/169.19(4.38–19.28)
 Current smokers (cig./day)
  1–1933/551.78(0.93–3.47)35/108.86(3.60–21.83)
  ≥2048/353.69(1.89–7.23)22/86.69(2.45–18.27)
Alcohol drinking
 Never drinker64/1741348/187.31(3.79–14.10)
 Current drinker48/382.83(1.58–5.09)46/138.62(4.12–18.06)
Toothbrush use
 Yes42/1521331/184.65(2.27–9.54)
 No85/802.52(1.49–4.24)89/1911.82(6.15–22.74)

DISCUSSION

In our present case-control study, paan-tobacco chewing was confirmed to be the most important determinant of oral cancer in Southern India. The fraction of the disease attributable to this habit was 49% in men and over 87% in women (Table VI). Among women, tobacco smoking and alcohol drinking have a negligible influence, whereas among men, smoking and drinking accounted for 21 and 26% of oral cancer cases, respectively. A lack of oral hygiene, as indicated by no use of toothbrush, accounted for 32% of oral cancer in men and 64% in women. All together, the factors above seemed to explain 76% of oral cancer in males and 95% in females (Table VI).

Table VI. Percent of Cancer of the Oral Cavity Attributable to Selected Habits by Gender (591 Cases and 582 Controls, India, 1996–99)
FactorAttributable risk percentage (95% CI)1
MenWomen
  • 1

    Estimates from a multiple logistic regression model including terms for gender, age, center, education and the main effects of the factors above. Ranges are in parentheses.

Tobacco smoking21 (−2–44)
Alcohol drinking26 (13–39)
Smoking and drinking35 (15–55)
Paan chewing49 (40–57)87 (83–92)
Paan chewing and smoking68 (53–82)
Poor oral hygiene32 (15–49)64 (47–80)
Chewing and hygiene50 (22–78)95 (91–98)
All above76 (65–86)95 (91–98)

The ORs we found for various levels of smoking and alcohol drinking among men are consistent with those shown before in India10–12, 17–20 and in Europe13 and North America.21 Bidi is confirmed to be at least equally harmful as regular cigarettes. Studies conducted in India have shown that bidis produce more carbon dioxide, nicotine, tar and alkaloids than regular cigarettes.22, 23 Furthermore, the filterless design of the bidi combined with low combustibility may contribute to higher toxin yields than with regular cigarettes.22 It is, however, worth noting that most Indian men in our present study started smoking relatively late, at 20 years or older. Heavy alcohol intake was not common, and the corresponding attributable risk was well below the ones found elsewhere.21, 24 For the combination of drinking and smoking in men, the attributable risk was approximately 80% in the United States and Europe and Latin America14, 21, 24versus 35% in our present study (Table VI).

Our present ORs for paan chewing in men are similar to those reported by Nandakumar et al.10 and Sankaranarayanan et al.11 In the latter study, the fraction of oral cancer attributable to chewing (73%) in Trivandrum in the mid-1980s was greater than in our present investigation, whereas the smoking-attributable fraction was lower (19%, bidi only). In agreement with our findings concerning different types of paan, a study from Pakistan6 showed an OR of 12.5 for paan-tobacco chewing and of 5.2 for chewing paan without tobacco. Interestingly, areca nut, 1 of the main ingredients of paan, is considered the strongest risk factor for oral submucous fibrosis, a precancerous condition very common in India.6, 25 Thus, our findings, albeit based on relatively few exposed subjects, contribute to the evaluation of carcinogenicity of paan without tobacco, which was still deemed to be inadequate in an IARC monograph.3

Women showed substantially higher ORs at any level of paan chewing than men. This difference was found consistently in the 3 participating centers after allowance for town or village of living, in different age groups and when the comparison between men and women was restricted to men who had never smoked or drunk alcoholic beverages. The only 2 Indian studies in which the 2 genders were analyzed separately also showed more elevated ORs in women than men,10, 19 although the difference was less marked than in our present study. In a large cross-sectional study on 927 cases of oral leukoplakia and 47,772 controls, interviewed in the framework of an oral cancer screening trial in the Trivandrum district, tobacco chewers showed an OR of 3.4 (95% CI 2.8–4.1) among men, but 37.7 (95% CI 24.2–58.7) among women.26 A greater susceptibility to the oral damage of pan-tobacco chewing in females is thus possible, as has been reported already for alcohol drinking.26, 27 It is also worth noting that women reported starting chewing on average 2 years earlier than men.

The percentage of ever chewers among female controls in our present study (13%), however, was lower than expected. In the aforementioned oral cancer screening trial, for instance, 22% of 65,792 women 35 years or older were pan-tobacco chewers.28 More than half of control women were chewers in previous case-control studies in Trivandrum28 and Bangalore.10 It is conceivable that the poorest, illiterate women, among whom chewing is commonest, do go to the hospital for advanced oral cancer (stage 3 and 4 in 80% of female cases in our present study), but they seldom attend as outpatients for less severe diseases or go to hospital in order to visit relatives and friends. Such scope for selection bias among female hospital controls should be taken into account in future planning of case-control studies in poor countries.

A gender-related difference was also found in respect to risk related to years of education and, to some extent, oral hygiene, on which our present study provides the first data in an Indian population. The great majority of study participants cleaned their teeth once per day or less, did not use a toothbrush and never had dental check-ups. The number of individuals missing more than 5 teeth or wearing a denture was, however, substantially lower than in studies done with the same protocol in Italy13 and Cuba.14 Among indicators of dental care, the use of a toothbrush, gum bleeding and number of missing teeth were associated with oral cancer risk after adjustment for smoking, drinking and chewing habits. These findings are in agreement with those from the Americas,14, 29 China30 and Europe.13, 31 As in Talamini et al.13 the strongest association emerged for general oral conditions reported by trained interviewers who performed oral inspection. Since inspection was performed before cancer treatment, however, interviewers could not be blinded about case-control status, and results must be interpreted cautiously.

In conclusion, our present study offers an up-to-date picture of major causes of oral cancer in Southern India. Traditional methods for mouth cleaning, such as the use of finger or wooden sticks, seem less effective than the use of a toothbrush. Paan chewing represents the most important cause of oral cancer in men and, most notably, in women. Among men, however, 35% of cases are attributable to the combination of smoking and alcohol drinking. Aggressive campaigns aimed at eliminating paan chewing are thus warranted, in addition to continued efforts to prevent the spread of tobacco smoking. Types of paan that do not include tobacco (e.g., some types of paan-masala) should not be marketed as safe alternatives to paan-tobacco chewing.

Acknowledgements

The authors thank Dr, K. Chaudry for useful comments and Dr. R. Ortiz Reyes and Mrs. A. Arslan for technical assistance.

Ancillary