This study found a lifetime substance use prevalence rate (69.8%) that is significantly higher than the 41% rate found among high school students in Kenya . This implies that substance use rates, in general, increase with age and transition through the education system. This has major policy implications, including the need to focus substance use interventions on younger age-groups such as primary and high school students. Preventing early substance-related problems will reduce the risk of these problems in later adulthood when the magnitude of life stresses is greater.
The prevalence rate of lifetime alcohol use in this study was 51.9%. This rate, though lower than what others have found in similar settings, is still significant. Odek-Ogunde et al  found rates of lifetime alcohol use as high as 84% among students at a private university in Nairobi, Kenya, while Othieno et al  in a more heterogeneous population consisting of outpatients attending several Nairobi primary health care facilities found a lifetime alcohol use prevalence rate of 62%.
The paucity of literature on alcohol use among college students in Africa and other parts of the world was demonstrated by Karama et al  in a 2006 review article in which the authors argued for the need for more studies to be carried out in this area in order to inform interventions. However, studies among high school students in this region report similarly high prevalence rates of alcohol use, ranging from 15% to 57.9% [2, 4, 6].
Almost all (97.6%) of those who had ever used alcohol in this study could be regarded as current users (had an alcoholic drink in the week preceding the study), translating to a 50.7% current alcohol use prevalence rate. In contrast, a Kenyan general population study  found a current alcohol use prevalence rate of 13.4%. The huge difference implies that alcohol use is probably still regarded as being fashionable among college students and very few social sanctions exist to discourage this behaviour. Chances of moving from use to abuse and dependence are therefore heightened, especially considering other factors such as age of onset and frequency of use.
Participants reported having used alcohol on an average of 6 days in the month preceding the study. The only factors significantly associated with this were housing type and government loan. Students residing in hostels drank on average 1.5 days more than those not residing in hostels, suggesting a significant role for peer influence in shaping drinking behaviour. Participants receiving government loan reported drinking almost 2 days more than those not receiving such loans, probably reflecting more readily available funds to sustain drinking behaviour. However, other factors may be responsible for this finding, and further research would be necessary to clarify this. Students in tertiary institutions in Kenya may apply for a Higher Education Loan from the government, and many of them are loaned various amounts of money depending on the course of study and the institution. The loan is recovered once the loanee completes their studies and secures employment or a steady source of income.
The mean number of days the participants reported drinking suggests that they drink between one and two days each week, pointing towards a binge-drinking pattern rather than regular daily intake. In further support for the likelihood of a binge-drinking pattern of alcohol use, almost 60% of the participants in this study reported taking 5 or more drinks per day at least once in the preceding month.
Previous research suggests that although it is difficult to generalise alcohol use patterns in many African countries, binge-drinking is a common phenomenon in the Kenyan general and even clinical population. For instance, a study by Saunders et al among patients at primary health care centres  reported that out of the six countries involved, the Kenyan sample had the highest rates of binge-drinking (episodic heavy drinking). The other countries in the study were the USA, Mexico, Australia, Norway and Bulgaria.
A community based survey by Clausen et al  reported that 27.6% of the Kenyan current drinkers had had 5 or more drinks on at least one occasion in the preceding week, suggesting that binge-drinking was a common phenomenon in the Kenyan drinking population. This pattern of drinking would also explain the high prevalence of reported serious problems associated with alcohol use in our study, such as scuffles, loss and damage to property, medical problems and sexual indiscretions.
An average of 3 drinks per drinking day was found among alcohol users in this study, with students living in hostels, in urban areas and receiving government loan taking more drinks on average than those without those attributes. These findings further illustrate the role of peer influence and urbanisation on alcohol use. The possibility of under-reporting cannot be discounted, given that the participants in this study were students who may be reluctant to reveal their drinking habits, even on an anonymised questionnaire.
In the study by Saunders et al  which was carried out in a primary health care setting, the mean daily number of drinks taken by the 'drinking population' was 9.6, while those classified as alcoholics were taking a mean of 23.4 drinks. Another study in a general population setting  had found a mean weekly drink rate of 9.9 drinks among Kenyan regular alcohol users, many of whom displayed a binge-drinking pattern of alcohol use.
In this study, the mean age at first alcoholic drink was 17.5 years, with the youngest reported age being 11 years. There was no statistically significant difference in male and female age of onset of alcohol use. Otieno and Ofulla  similarly found the highest prevalence of alcohol use among young people aged 16-18 years. Several other studies have reported early age of onset of alcohol use among adolescents and the associated psychological problems in later life [9, 20, 21].
Despite the fact that the legal drinking age in Kenya has been set at 18 years for a long time, it is clear that alcohol is available for sale even to underage drinkers. As has been observed before , failure of enforcement may be part of the problem, as well as the domination of the alcohol market by small-scale often illicit producers and sellers. Partanen  suggests that 80-90% of total alcohol consumed in Kenya comes from 'small-scale production within the informal sector, licit or illicit, using traditional African methods of brewing and local skills of distilling'.
Among some of the problems associated with early age of onset of alcohol use is increased risk of later alcohol abuse and dependence as well as associated social and occupational difficulties [20, 21]. In one study, Grant and Dawson showed that the odds for dependence decreased by 14% while those for abuse decreased by 8% with each increasing year of onset of use . The implication of the high rate of alcohol use and early age of onset in this study is that a large proportion of the respondents are at a high risk of developing alcohol-related disorders as adults.
In the present study, negative effects attributed to alcohol use by the respondents included quarrels and fights, loss and damage to property, regretted sex, unprotected sex, and medical problems. As indicated earlier, most of these problems could be attributed to a binge-drinking pattern of behaviour, rather than regular light use of alcohol. The high risk sexual behaviour is particularly ominous due to the high prevalence of HIV and other sexually transmitted infections in Kenya.
In 2007, Chersich et al  reported an association between binge-drinking and unsafe sex, sexual violence and sexually transmitted infections among Kenyan female sex workers, and the students in the present study are clearly exposed to the same risks. It is clear that unless substance use among adolescents and young adults is addressed, interventions targeting HIV/AIDS, violence and accidents will achieve less than optimal results.
The lifetime prevalence rate of cigarette smoking in this sample was 42.8%, which is higher than what has been reported elsewhere [3, 4, 6, 8]. However, most of these prior reports focused on adolescents in high school, once again raising the possibility that the rates of cigarette use increase with age and academic progression.
In this study, there was a statistically significant difference in the lifetime cigarette use prevalence rates between males and females, with males having a higher rate than females. This is consistent with what has been found in other studies, and may reflect a more tolerant social attitude to male than to female smoking [2–5, 16, 24].
The mean age at first cigarette use in this study was 15.7 years, and the youngest reported age was 10 years. Males had a significantly lower mean age of onset of cigarette use than females. The Kenyan Tobacco Control Act (2007) prescribes the minimum smoking age to be 18 years, yet tobacco products appear to be available to people as young as 10 years old. The same regulatory problems identified in relation to alcohol use are probably also operational with regard to tobacco use in Kenya.
Elsewhere, Kwamanga et al  reported an age of cigarette smoking onset of 12 to 16 years, while Peltzer  reports a mean age of onset of 14.8 years, similar to the findings in this study. Early onset of cigarette use has obvious implications for these young people in social and academic spheres. As a matter of fact, the commonest smoking-attributed problems reported by participants in this study were associated with teachers, parents and physical health. Although it was beyond the scope of this study to identify exactly what sort of problems the smokers encountered with teachers and parents, it may be presumed that they were probably related to discipline and interpersonal conflict. Smoking is outlawed in learning institutions in Kenya, as well as in most public places.
Although only a small proportion (7%) of those who smoked reported using more than 16 cigarettes a day, this is still significant in view of the health risks posed by this amount of exposure to cigarette smoke. It also suggests that this group is already dependent on nicotine, and this may cause problems in important areas of functioning including school, social and family functioning.
The rates of other substance use in this study were low. This may be attributed to either under-reporting or a lack of availability of these substances. It is difficult to ascertain the true reason for the low prevalence rates since the questionnaire used was a self-report format and there was no follow-up of responses. However, this finding appears consistent with the previous studies in the same environment, where reports of illicit drug use have been low [2, 5, 16]. Only cannabis use has been found to be more common in the earlier studies compared to the present one. For instance, Odek-Ogunde  reported a 19.7% lifetime prevalence rate of cannabis use among university students, but found rates less than 5% for drugs such as heroin, cocaine and amphetamines.
Majority of those reporting lifetime substance use indicated that they were influenced by a friend or relatives other than the nuclear family. Similar findings have been reported in other studies, confirming the role of peer pressure and social learning in initiation of substance use [8, 24]. The implication of this finding is that peers and older relatives would also serve as good role models for a substance use intervention program for young people in this setting.
In this study, most respondents indicated that they used substances to relax and relieve stress. Madu and Matla found that majority of drug using adolescents do so when they are bored, tired or stressed up, or at parties . A 1987 Medical Student survey in the US also found similar reasons for most of the drugs used, including to relax, to have a good time, to feel good and even to experiment . This raises the possibility that interventions that help young people to use their time more productively would reduce the incidence of substance use.
This study encountered a number of limitations, chief among them being the relatively small sample size and a study design that precludes generalisation of the results to students in other institutions. Due to the heterogeneity of the institutions and the courses offered in this region, it would take a larger study to fully describe the substance use patterns among college students in Kenya or even within Eldoret Municipality. The information generated from this study may however be useful in the design of such a study.
Another key limitation is the fact that a self-administered questionnaire was used to collect data on current and lifetime substance use. Without any other validating measures it is not possible to conclude with certainty that the information generated is an accurate representation of substance use in this population. However, the WHO model questionnaire has been used elsewhere [2, 14] under similar circumstances, and therefore the results are at least comparable to those found in similar studies.