**2. Use of reference intervals, consequence of misclassification and selection of a reference population**

#### **2.1. The use of reference intervals**

304 Blood Cell – An Overview of Studies in Hematology

PARAMETER GRADE 1

NEUTROPHILS

clarification August 2009.

assumes a value of zero.

MILD

1.0 –

1.3 x 109 cells/L

abnormal values, how to grade the severity of the abnormality.

GRADE 2 MODERATE

HEMOGLOBIN 10.0 – 10.9 g/dL 9.0 – 9.9 g/dL 7.0 – 8.9 g/dL < 7.0 g/dL

0.999 x 109cells/L

Adult and pediatric values for age >57 days, HIV-negative from the DAIDS toxicity tables version 1.0, December 2004;

**Table 1.** Examples of DAIDS criteria of estimating severity grading based on laboratory parameters.

Reference values, in general, refer to the value or test result obtained by the observation or measurement of a particular type of quantity on an adequate number of persons (reference sample group) selected to represent the general population. Reference values are usually presented as reference intervals which refer to the interval between, and including two reference limits i.e., from the lower reference limit to the upper reference limit defined by a specific percentage (usually 95%). In certain parameters such as absolute counts of monocytes, eosinophils and basophils, only one reference limit (decision limit), more often the upper reference limit is of biological significance hence the lower reference limit

Reference values go hand in hand with toxicity grading or decision limits, which can be defined as specific levels of the analyte that correspond to mild to life threatening clinical situations. Toxicity grading is particularly useful in the decision-making process of interpreting a measured value and assessing the health status of the subject being tested. For this reason reference values or toxicity grading are routinely used in clinical trials at enrollment to determine eligibility, establish baseline measures, and also during the course of the trial to monitor the participants' health. Moreover, several analytes are used either as markers for the possible presence of a disease or as direct evidence for that disease. Reference values, especially hematological and immunologic indices, are influenced by such factors as genetics, dietary patterns, pregnancy, gender, age, ethnic origin and prior exposure to environmental pathogens. Thus, it is important to consider these factors when

0.75 –

GRADE 3 SEVERE

0.5 – 0.749 x 109 GRADE 4 POTENTIALLY LIFE-THREATENING

cells/L

cells/L < 0.5 x 109

Historically, clinical studies as well as routine clinical patient management in most African countries have relied on European-generated automated instrument values, US established reference intervals or the U.S. NIH division of AIDS (DAIDS) toxicity grading tables in assessing clinical parameters in study participants. The US-established reference intervals are obtained from the Massachusetts General Hospital reference values and serve as the standard reference interval comparison for most studies [6]. The DAIDS toxicity tables, also derived from a Caucasian population, are used for grading the severity of adult and pediatric adverse events, whether or not they are considered to be related to the study intervention [7]. DAIDS provides guidelines for estimating severity of adverse events using specific reference intervals (Table 1) as criteria for determining what is 'normal' and among

> Reference intervals are useful both in the clinical and research environment. Medical laboratory reference intervals are primarily used for clinical purposes. They can be used as an indicator of good health. Alternatively, reference intervals/limits can be used to screen for physiological or pathological conditions hence important in routine health assessment,

particularly for screening of anemia, blood disorders and diseases of the immune system. Reference intervals are important for accurate interpretation of laboratory data and provide assistance to the clinician in creating a more comprehensive clinical perspective for diagnosis and management of patients [10]. Of particular importance is the use of reference values as surrogate markers for monitoring disease progression and response to antiretroviral therapy in HIV-infected individuals [11]. For example, decisions to initiate, continue, or change antiretroviral therapy regimens are determined using CD4+ Tlymphocyte cell (CD4) counts, while drug toxicity is monitored using liver function tests, renal function tests, and full blood counts (FBC) [12, 13]. The hemoglobin concentration is used as a marker of anemia. As part of the management of anemia, the clinician conducts additional tests to identify a reversible etiology for anemia (eg, iron deficiency, infection) and if present treats it appropriately. However, in the clinical environment, the statistical definition of reference intervals may not allow certain clinical uses. Because these reference intervals have been derived statistically from a healthy population, they may not be used to rule in or rule out specific medical conditions. The statistically derived 95% reference interval would mean that 5% of normal subjects would have abnormal laboratory values. This is erroneously interpreted that 95% of diseased individuals would fall outside the derived reference interval. It is recommended that the number of diseased individuals who fall outside the defined 95% reference intervals be determined through a study of the distribution of such persons with the target condition [14]. Thus, it is necessary to confirm the validity of the proposed reference intervals with clinicians using a particular test to manage patients.

Laboratory Reference Intervals in Africa 307

these challenges, there is a need to establish accurate, locally derived reference intervals for the target population. Within the last decade, several studies in sub-Saharan Africa have attempted to establish hematological and biochemical reference intervals for use in clinical

The selection of a reference population is as per described in the Clinical Laboratory Standards Institute (CLSI, Wayne, PA, USA) guidelines [21]. The guidelines state that reference individuals selected for the determination of reference intervals should closely resemble the patient population undergoing medical examination and should be of similar age to be clinically significant [21]. The reference individuals should not be hospital or clinic patients unless absolutely necessary. The guidelines describe two selection methods for a reference population: *a priori* and *a posteriori*. *A priori* sampling method involves selection of reference individuals based on well-defined exclusion and partition criteria. The entire selection process takes place before any blood sample is drawn and a sufficient number of reference individuals are targeted to provide statistical validity. *A posteriori* sampling method involves selection of the reference population after the analyte has been tested. The CLSI guidelines recommend a minimum of 120 individuals to allow 90% confidence limits to be non-parametrically calculated for the reference limits [22]. Partitioning of reference intervals either by gender or age is recommended if clinically useful or physiologically well grounded. Even though 120 samples remains the recommended standard, an efficient laboratory, by considering the CLSI revised guideline strategies [8], can determine reference intervals using fewer samples [23]. Alternatively, a laboratory can adopt reference intervals established from another laboratory if the values are verified using the procedures set out in the guidelines.

In our study [9] of adolescents and adults living in rural western Kenya, all participants were screened by a review of medical history, a physical examination, tested for HIV and pregnancy (for females), and treated for any illnesses diagnosed. Participants were included if they were a permanent resident of the study area, between 13 and 34 years of age and able to provide informed consent or assent if a minor. Participants were excluded if they were HIV-seropositive, pregnant, exhibiting febrile symptoms or on any medication. Blood samples were therefore obtained from clinically healthy participants selected to generate hematologic and biochemical reference intervals. Data were partitioned by age (<18 years of age as adolescents and ≥18 as adults) and gender; median and 95th percentile intervals were calculated. The lower 95% reference limit was defined as the 2.5th percentile while the upper limit was defined as the 97.5th percentile. A Wilcoxon rank-sum test was used to test for age and gender differences. We compared our data against reference intervals from the Massachusetts General Hospital (MGH-USA) [6] and the U.S. NIH Division of AIDS (DAIDS) toxicity tables [7] to determine the number of study participants with values outside the MGH ranges or who had any adverse event as graded by the DAIDS criteria. However, while the CLSI guidelines recommend a description of the population from which reference intervals are derived, the DAIDS and Massachusetts General Hospital reference

monitoring and patient management.

**2.3. Selection of a reference population** 

values do not provide such information.

In the research environment, however, the aim is to define a reference population that is as similar as possible to that for which a particular test will be applied with the exception of the presence of the disease. During clinical trials, reference intervals relevant to the study of interest are required to interpret normal values of standard laboratory test results from the target population [15]. This is particularly important during phase I/II safety trials where healthy individuals are assessed without a control group [15-17]. Moreover, clinical reference intervals are necessary in order to accurately assess potential adverse events observed during the course of clinical trials.

#### **2.2. Consequences of misclassification**

A majority of clinically healthy participants have been excluded from several clinical trials in Africa because laboratory hematological and biochemical parameters are classified as abnormal [18-20]. Unnecessary exclusion of potential participants generally results in increased cost for study recruitment to achieve the target sample size. Accurate reference intervals are required for monitoring adverse events during vaccine and drug trials to limit misclassification that might otherwise lead to discontinuation of such trials or erroneous conclusions that the trial interventions are associated with adverse events. A study documented that the expense of adverse event investigation and reporting accounted for at least one-third of the study cost, irrespective of the adverse event grade [18]. To overcome these challenges, there is a need to establish accurate, locally derived reference intervals for the target population. Within the last decade, several studies in sub-Saharan Africa have attempted to establish hematological and biochemical reference intervals for use in clinical monitoring and patient management.

#### **2.3. Selection of a reference population**

306 Blood Cell – An Overview of Studies in Hematology

manage patients.

observed during the course of clinical trials.

**2.2. Consequences of misclassification** 

particularly for screening of anemia, blood disorders and diseases of the immune system. Reference intervals are important for accurate interpretation of laboratory data and provide assistance to the clinician in creating a more comprehensive clinical perspective for diagnosis and management of patients [10]. Of particular importance is the use of reference values as surrogate markers for monitoring disease progression and response to antiretroviral therapy in HIV-infected individuals [11]. For example, decisions to initiate, continue, or change antiretroviral therapy regimens are determined using CD4+ Tlymphocyte cell (CD4) counts, while drug toxicity is monitored using liver function tests, renal function tests, and full blood counts (FBC) [12, 13]. The hemoglobin concentration is used as a marker of anemia. As part of the management of anemia, the clinician conducts additional tests to identify a reversible etiology for anemia (eg, iron deficiency, infection) and if present treats it appropriately. However, in the clinical environment, the statistical definition of reference intervals may not allow certain clinical uses. Because these reference intervals have been derived statistically from a healthy population, they may not be used to rule in or rule out specific medical conditions. The statistically derived 95% reference interval would mean that 5% of normal subjects would have abnormal laboratory values. This is erroneously interpreted that 95% of diseased individuals would fall outside the derived reference interval. It is recommended that the number of diseased individuals who fall outside the defined 95% reference intervals be determined through a study of the distribution of such persons with the target condition [14]. Thus, it is necessary to confirm the validity of the proposed reference intervals with clinicians using a particular test to

In the research environment, however, the aim is to define a reference population that is as similar as possible to that for which a particular test will be applied with the exception of the presence of the disease. During clinical trials, reference intervals relevant to the study of interest are required to interpret normal values of standard laboratory test results from the target population [15]. This is particularly important during phase I/II safety trials where healthy individuals are assessed without a control group [15-17]. Moreover, clinical reference intervals are necessary in order to accurately assess potential adverse events

A majority of clinically healthy participants have been excluded from several clinical trials in Africa because laboratory hematological and biochemical parameters are classified as abnormal [18-20]. Unnecessary exclusion of potential participants generally results in increased cost for study recruitment to achieve the target sample size. Accurate reference intervals are required for monitoring adverse events during vaccine and drug trials to limit misclassification that might otherwise lead to discontinuation of such trials or erroneous conclusions that the trial interventions are associated with adverse events. A study documented that the expense of adverse event investigation and reporting accounted for at least one-third of the study cost, irrespective of the adverse event grade [18]. To overcome The selection of a reference population is as per described in the Clinical Laboratory Standards Institute (CLSI, Wayne, PA, USA) guidelines [21]. The guidelines state that reference individuals selected for the determination of reference intervals should closely resemble the patient population undergoing medical examination and should be of similar age to be clinically significant [21]. The reference individuals should not be hospital or clinic patients unless absolutely necessary. The guidelines describe two selection methods for a reference population: *a priori* and *a posteriori*. *A priori* sampling method involves selection of reference individuals based on well-defined exclusion and partition criteria. The entire selection process takes place before any blood sample is drawn and a sufficient number of reference individuals are targeted to provide statistical validity. *A posteriori* sampling method involves selection of the reference population after the analyte has been tested. The CLSI guidelines recommend a minimum of 120 individuals to allow 90% confidence limits to be non-parametrically calculated for the reference limits [22]. Partitioning of reference intervals either by gender or age is recommended if clinically useful or physiologically well grounded. Even though 120 samples remains the recommended standard, an efficient laboratory, by considering the CLSI revised guideline strategies [8], can determine reference intervals using fewer samples [23]. Alternatively, a laboratory can adopt reference intervals established from another laboratory if the values are verified using the procedures set out in the guidelines.

In our study [9] of adolescents and adults living in rural western Kenya, all participants were screened by a review of medical history, a physical examination, tested for HIV and pregnancy (for females), and treated for any illnesses diagnosed. Participants were included if they were a permanent resident of the study area, between 13 and 34 years of age and able to provide informed consent or assent if a minor. Participants were excluded if they were HIV-seropositive, pregnant, exhibiting febrile symptoms or on any medication. Blood samples were therefore obtained from clinically healthy participants selected to generate hematologic and biochemical reference intervals. Data were partitioned by age (<18 years of age as adolescents and ≥18 as adults) and gender; median and 95th percentile intervals were calculated. The lower 95% reference limit was defined as the 2.5th percentile while the upper limit was defined as the 97.5th percentile. A Wilcoxon rank-sum test was used to test for age and gender differences. We compared our data against reference intervals from the Massachusetts General Hospital (MGH-USA) [6] and the U.S. NIH Division of AIDS (DAIDS) toxicity tables [7] to determine the number of study participants with values outside the MGH ranges or who had any adverse event as graded by the DAIDS criteria. However, while the CLSI guidelines recommend a description of the population from which reference intervals are derived, the DAIDS and Massachusetts General Hospital reference values do not provide such information.
