The burden of diseases such as HIV/AIDS, tuberculosis, and malaria is heaviest in sub-Saharan Africa compared to the rest of the world [1, 2]. For example, sub-Saharan Africa has the highest prevalence and incidence of HIV infection globally. As such, a major of many recent HIV prevention, care and treatment initiatives are being conducted within the region [3, 4]. including most phase I ⁄ IIB HIV-1 vaccine trials [5].

With increasing clinical trials in sub-Saharan Africa to combat these diseases, there is a need for accurate clinical laboratory reference intervals for appropriate participant screening, disease progression monitoring and evaluation of possible clinical trial-associated toxicity and adverse events [6]. Reference intervals are important for guiding patient treatment and management as well as identifying abnormal hematologic values [7]. For example, the complete blood count and CD4 determination are important laboratory tests in HIV-endemic regions [8]. The level of hemoglobin concentration has utility as a prognostic indicator while CD4 is used to make decisions regarding initiation of antiretroviral drugs and to monitor disease progression. These tests require accurate reference intervals for correct interpretation of laboratory results. However, currently used reference intervals in many countries in sub-Saharan Africa are derived from populations in Europe and North America [6, 9]. Since hematologic parameters are affected not only by individual factors such as age, sex and lifestyle, but also by population and ecological factors such as ethnic background, climate, exposure to pathogens and altitude, they vary not only between individuals but also between populations [10]. Thus, there is not a universal definition of ‘normal’ hence it is important to define reference intervals that are suited to the particular population of interest [10]. A few studies conducted in Africa over the last decade have highlighted differences in hematologic parameters between the local population and Caucasian populations in Europe and North America [11–15]. More recently, a study highlighted differences in hematological and biochemistry values between adolescent and young adult males [15]. Despite these recorded inter-population differences in reference values for different geographical regions, few data exist in Africa to provide locally-derived values [7, 12, 13, 16].

Despite these recorded differences, the Division of AIDS (DAIDS) National Institute of Health toxicity tables [17], are still used for grading the severity of adult and pediatric adverse events, whether or not they are considered to be related to the study intervention. This leads to unnecessary exclusion of would be participants misclassified as having abnormal hematologic parameters thereby escalating operational costs especially in phase I safety trials where there may not be a control group [18–20]. This may also lead to improper patient management through misclassification of adverse events. Due to these differences, there is a need to develop and test locally-derived age specific reference intervals within African populations.

While it is desirable to generate reference intervals for different populations, the procedure remains a challenge due to the prohibitive cost involved in performing these studies and the limitation in identifying suitable healthy reference individual. Thus, the recommendation by the Clinical and Laboratory Standards Institute (CLSI) that all diagnostic laboratories must determine and maintain their own reference interval for each laboratory parameter is impractical. CLSI further recommends that if it is not possible to establish the detailed reference studies, then validation of published reference intervals can be performed using own methodology for the population served by the laboratory. Zeh et al have recently established reference intervals for use in western Kenya [15]. These intervals were generated from a study conducted on 13–34 year old, clinically healthy, HIV-seronegative, non-pregnant residents of western Kenya. Because the established reference intervals were from a population in Siaya County in western Kenya, our aim was to validate these established reference intervals for use in Kisumu County of western Kenya. We also retrospectively determined the proportion of participants in an observational prospective cohort study known as the Kisumu Incidence Cohort Study (KICoS), who would be misclassified as having abnormal hematological parameters using the established reference intervals and compared our findings to those obtained using the 2004 NIH DAIDS toxicity tables.

With increasing clinical trials in Africa in an effort to combat tropical diseases [24], a need arises to consider the health status of the likely participants in such studies [12]. In this regard, several African studies have generated reference intervals for use in the respective regions [13, 15, 24, 25]. While it is important to develop locally derived reference intervals that ensure proper assessment of volunteers in clinical trials, monitoring of laboratory-based adverse events and prevention of unnecessary exclusion, it is important to evaluate their use within the local population. To our knowledge, this is one of the first evaluations of established reference intervals reported in sub-Saharan Africa. In this study, we evaluate the use of hematological and biochemistry reference intervals established for western Kenya using specimen drawn from participants in a HIV incidence cohort study in Kisumu. Our values were comparable to those of the established reference intervals for most parameters although our median values were slightly higher for most hematological parameters. This may be so given that the samples for this evaluation were drawn from an urban population that may have had access to better healthcare, clean water and nutrition than the rural population from where the established reference intervals for western Kenya [15] were obtained. Moreover, using the US MGH reference interval, the overall out of range MCV values constituted 54.0% [15] of the study population in rural western Kenya but only 27.6% in this study. Low MCV is an indirect marker of iron deficiency [26]. This is further corroborated by the low eosinophil counts observed in this study. Our eosinophil counts are comparable to those obtained from an urban population of blood donors in Uganda [18] in contrast to higher counts in a similar study in a rural population in the same country [27]. Our values for ALT and creatinine were also lower than those of Zeh et al [15]. Similarly, a study in Cameroon designed to establish reference intervals for biochemical parameters reported statistical differences in biochemistry reference parameters between participants from urban and rural geographic regions [28]. While this might not necessitate the need to establish separate intervals, consideration should be made when applying such intervals within specific populations.

Using the US-based MGH reference intervals to hypothetically select participants in a trial based on Hb, WBC counts, neutrophil counts, eosinophil counts and platelets, 51.5% (n = 491) of the total participants would have been excluded from participating in the study (Table 7). However, using the established reference intervals for western Kenya, only 6.7% (n = 64) of participants would have been excluded from participating in the study. Including BUN in the selection criteria would result in exclusion of over 80% of participants. This was similar in other African studies [7, 13, 15, 27] suggesting that this may result from a common environmental or genetic factor [12, 13]. Thus use of locally established reference intervals would reduce the overall screening to enrollment ratio in this case. This reduces the overall cost of screening and theoretically would reduce the time period for screening by reaching the study target within a shorter time period. Eller et al. have documented similar findings in a study of healthy adult Ugandan blood donors [18]. It is not surprising to see that the adolescent cohort resulted in the most out of range values hence yielded the least enrolled participants using the western-derived reference intervals and toxicity tables. Thus, partitioning of male adolescents needs to be considered in future trials.

Similarly, using the 2004 NIH US DAIDS toxicity grading for screening, 31.0% (n = 295) of participants would have been excluded from the study. Although the table has been revised for some parameters including neutropenia [29], a large proportion of our study participants would still have been excluded based on Hb levels. Moreover, the toxicity table does not take into account the significant difference in red blood cell parameters between males and females, thus, a majority of those excluded would constitute female participants.

The CLSI guidelines recommend the collection of specimen from healthy volunteers for use in validating reference intervals. Thus, a limitation of this study was the failure to screen for possible asymptomatic parasitic infections like malaria and helminthes which are endemic within the study region. However, our Hb values were much higher than those from a study within the same region that screened out malaria infected participants [13]. Moreover, our eosinophil counts were much lower than the two studies within the region which screened participants from a rural population [13, 15]. A second limitation may have been that this was a self-selected population of participants willing to participate in a cohort study. However, this represents a similar population that would be willing to participate in a clinical trial thus provides a good sample to evaluate the use of the locally established reference intervals. Moreover, the HIV prevalence of the study population (12.0%) is comparable to the prevalence within the general population [30].

Given that the number of clinical trials and persons receiving clinical services is expected to increase substantially in sub-Saharan Africa, there is a need for the establishment and evaluation of locally derived clinical laboratory reference values to ensure appropriate general health assessment, treatment monitoring, and efficient implementation of clinical trials. Even more important is the need for the establishment of toxicity grading tables for application in clinical care among Africans based on the documented differences between laboratory reference intervals from African and Caucasian populations. This study confirms that the hematological and biochemistry reference intervals established by Zeh et al. are valid for use in participant recruitment in western Kenya.