labNorm: Normalize Laboratory Measurements by Age and Sex

Description

Provides functions for normalizing standard laboratory measurements (e.g. hemoglobin, cholesterol levels) according to age and sex, based on the algorithms described in "Personalized lab test models to quantify disease potentials in healthy individuals" (Netta Mendelson Cohen, Omer Schwartzman, Ram Jaschek, Aviezer Lifshitz, Michael Hoichman, Ran Balicer, Liran I. Shlush, Gabi Barbash & Amos Tanay, doi:10.1038/s41591-021-01468-6). Allows users to easily obtain normalized values for standard lab results, and to visualize their distributions. See more at https://tanaylab.weizmann.ac.il/labs/.

Author(s)

Maintainer: Aviezer Lifshitz aviezer.lifshitz@weizmann.ac.il

Authors:

• Netta Mendelson-Cohen netta.mendelsoncohen@weizmann.ac.il

Other contributors:

• Weizmann Institute of Science [copyright holder]

Available lab names

Description

Names of the labs available in the package.

Usage

LAB_DETAILS

Format

LAB_DETAILS

A data frame with 95 rows and 4 columns:

short_name

Short lab name

long_name

Long lab name

units

a list column with all the units available for the lab

default_units

the default units for the lab

low_male,high_male,low_female,high_female

the reference ranges for the lab, taken from the American Board of Internal Medicine. Can be NA if the lab does not have reference ranges.

Source

American Board of Internal Medicine. ABIM Laboratory Test Reference Ranges — July 2021. https://www.abim.org/~/media/ABIM%20Public/Files/pdf/exam/laboratory-reference-ranges.pdf (2021).

Examples

head(LAB_DETAILS)

Example values of Hemoglobin and Creatinine

Description

Example datasets of Hemoglobin and Creatinine values for testing

Usage

hemoglobin_data

creatinine_data

Format

hemoglobin_data creatinine_data

A data frame with 1000 rows and 3 columns:

age

age of the patient

sex

sex of the patient

value

the lab value for the patient, in the default units for the lab

An object of class data.frame with 1000 rows and 3 columns.

Examples

head(hemoglobin_data)
head(creatinine_data)

Convert values to the default units for the lab

Description

Convert values to the default units for the lab

Usage

ln_convert_units(values, units, lab)

Arguments

Value

the values converted to the default units for the lab

Examples

# emulate a dataset with different units

hemoglobin_diff_units <- hemoglobin_data

# first 50 values will be in mg/ML
hemoglobin_diff_units$value[1:50] <- hemoglobin_diff_units$value[1:50] * 10

# last 50 values will be in mmol/L
hemoglobin_diff_units$value[51:100] <- hemoglobin_diff_units$value[51:100] / 1.61


converted <- ln_convert_units(
    hemoglobin_diff_units$value[1:100],
    c(rep("mg/mL", 50), rep("mmol/L", 50)),
    "Hemoglobin"
)

head(converted)
head(hemoglobin_data$value)

Download high-resolution reference distributions

Description

The data is downloaded to the directory specified by the dir parameter. Note that if you specified a directory different from the default, you will need to set options(labNorm.dir = dir) in order for the package to use the downloaded data in future sessions.
Default directories are:

• Unix: ~/.local/share/LabNorm

• Mac OS X: ⁠~/Library/Application Support/LabNorm⁠

• Win XP (not roaming): ⁠C:\\Documents and Settings\\<username>\\Data\\<AppAuthor>\\LabNorm⁠

• Win XP (roaming): ⁠C:\\Documents and Settings\\<username>\\Local Settings\\Data\\<AppAuthor>\\LabNorm⁠

• Win 7 (not roaming): ⁠C:\\Users\\<username>\\AppData\\Local\\<AppAuthor>\\LabNorm⁠

• Win 7 (roaming): ⁠C:\\Users\\<username>\\AppData\\Roaming\\<AppAuthor>\\LabNorm⁠

Usage

ln_download_data(dir = NULL)

ln_data_downloaded()

Arguments

Value

None.

True if the data was downloaded, false otherwise.

Examples

ln_download_data()


ln_data_downloaded()

Get available units for a lab

Description

Get available units for a lab

Get the default units for a lab

Usage

ln_lab_units(lab)

ln_lab_default_units(lab)

Arguments

Value

a vector of available units for the lab

the default units for the lab

Examples

ln_lab_units("Hemoglobin")

ln_lab_default_units("Hemoglobin")

Normalize lab values to age and sex

Description

Normalize standard laboratory measurements (e.g. hemoglobin, cholesterol levels) according to age and sex, based on the algorithms described in "Personalized lab test models to quantify disease potentials in healthy individuals" doi:10.1038/s41591-021-01468-6.

The "Clalit" reference distributions are based on 2.1B lab measurements taken from 2.8M individuals between 2002-2019, filtered to exclude severe chronic diseases and medication effects. The resulting normalized value is a quantile between 0 and 1, representing the value's position in the reference distribution.

The "UKBB" reference distributions are based on the UK-Biobank, a large-scale population-based cohort study of 500K individuals, which underwent the same filtering process as the "Clalit" reference distributions.

The list of supported labs can be found below or by running LAB_DETAILS$short_name.

Usage

ln_normalize(
  values,
  age,
  sex,
  lab,
  units = NULL,
  reference = "Clalit",
  na.rm = FALSE
)

ln_normalize_multi(labs_df, reference = "Clalit", na.rm = FALSE)

Arguments

Value

a vector of normalized values. If ln_download_data() was not run, a lower resolution reference distribution will be used, which can have an error of up to 5 quantiles (0.05). Otherwise, the full reference distribution will be used. You can check if the high resolution data was downloaded using ln_data_downloaded().
You can force the function to use the lower resolution distribution by setting options(labNorm.use_low_res = TRUE).
If the quantile information is not available (e.g. "Estradiol" for male patients, various labs which are not available in the UKBB data), then the function will return NA.

reference distribution

It is highly recommended to use ln_download_data to download the "Clalit" and "UKBB" reference distributions. If you choose not to download the data, the package will use the demo reference distributions included in the package ("Clalit-demo"), which doesn't include all the labs, and has a resolution of 20 quantile bins and therefore may have an error of up to 5 percentiles (0.05), particularly at the edges of the distribution.

labs

The following labs are supported in the "Clalit" reference (some labs are missing from the UKBB reference):

• WBC

• RBC

• Hemoglobin

• Hematocrit

• Platelets

• MCV

• MCH

• MCHC

• RDW

• MPV

• Large unstained cells, Abs

• Albumin

• Total Cholesterol

• Triglycerides

• BMI

• Iron

• Transferrin

• Ferritin

• PDW

• MPXI

• Total Globulin

• PCT

• HDW

• Fibrinogen

• CH

• Chloride

• Large unstained cells, %

• Macrocytic

• Microcytic

• Hyperchromic

• Hypochromic

• Lymphocytes, Abs

• Lymphocytes, %

• Neutrophils, Abs

• Neutrophils, %

• Monocytes, Abs

• Monocytes, %

• Eosinophils, Abs

• Eosinophils, %

• Basophils, Abs

• Basophils, %

• Microcytic:Hypochromic

• Glucose

• Urea

• Creatinine

• Uric Acid

• Calcium

• Phosphorus

• Total Protein

• HDL Cholesterol

• LDL Cholesterol

• Alk. Phosphatase

• AST

• ALT

• GGT

• LDH

• CPK

• Total Bilirubin

• Direct Bilirubin

• Hemoglobin A1c

• Sodium

• Potassium

• Vitamin D (25-OH)

• Microalbumin:Creatinine

• Urine Creatinine

• Urine Microalbumin

• Non-HDL

• TSH

• T3, Free

• T4, Free

• Blood Pressure, Systolic

• Blood Pressure, Diastolic

• Urine Specific Gravity

• Urine pH

• PT, INR

• PT, sec

• PT, %

• Vitamin B12

• PSA

• ESR

• aPTT, sec

• CRP

• Amylase

• Folic Acid

• Total:HDL

• Hematocrit:Hemoglobin

• Magnesium

• aPTT, ratio

• Indirect Bilirubin

• RDW-SD

• RDW-CV

• LH

• Estradiol

Examples

# Normalize Hemoglobin values to age and sex
hemoglobin_data$quantile <- ln_normalize(
    hemoglobin_data$value,
    hemoglobin_data$age,
    hemoglobin_data$sex,
    "Hemoglobin"
)

# plot the quantiles vs values for age 50-60
library(ggplot2)
library(dplyr)
hemoglobin_data %>%
    filter(age >= 50 & age <= 60) %>%
    ggplot(aes(x = value, y = quantile, color = sex)) +
    geom_point() +
    theme_classic()

# Different units
hemoglobin_diff_units <- hemoglobin_data
hemoglobin_diff_units$value <- hemoglobin_diff_units$value * 0.1
hemoglobin_diff_units$quantile <- ln_normalize(
    hemoglobin_data$value,
    hemoglobin_data$age,
    hemoglobin_data$sex,
    "Hemoglobin",
    "mg/mL"
)

# Multiple units
creatinine_diff_units <- creatinine_data
creatinine_diff_units$value <- c(
    creatinine_diff_units$value[1:500] * 0.011312,
    creatinine_diff_units$value[501:1000] * 11.312
)
creatinine_diff_units$quantile <- ln_normalize(
    creatinine_diff_units$value,
    creatinine_diff_units$age,
    creatinine_diff_units$sex,
    "Creatinine",
    c(rep("umol/L", 500), rep("mmol/L", 500))
)

# Use UKBB as reference
hemoglobin_data_ukbb <- hemoglobin_data %>% filter(age >= 35 & age <= 80)
hemoglobin_data_ukbb$quantile_ukbb <- ln_normalize(
    hemoglobin_data_ukbb$value,
    hemoglobin_data_ukbb$age,
    hemoglobin_data_ukbb$sex,
    "Hemoglobin",
    reference = "UKBB"
)

# plot UKBB vs Clalit
hemoglobin_data_ukbb %>%
    filter(age >= 50 & age <= 60) %>%
    ggplot(aes(x = quantile, y = quantile_ukbb, color = sex)) +
    geom_point() +
    geom_abline() +
    theme_classic()


# examples on the demo data


library(dplyr)
multi_labs_df <- bind_rows(
    hemoglobin_data %>% mutate(lab = "Hemoglobin"),
    creatinine_data %>% mutate(lab = "Creatinine")
)


multi_labs_df$quantile <- ln_normalize_multi(multi_labs_df)


# on the demo data


head(multi_labs_df)

Plot age-sex distribution of a lab

Description

Plot age-sex distribution of a lab

Usage

ln_plot_dist(
  lab,
  quantiles = c(0.03, 0.1, 0.15, 0.25, 0.35, 0.65, 0.75, 0.85, 0.9, 0.97),
  reference = "Clalit",
  pal = c("#D7DCE7", "#B0B9D0", "#8997B9", "#6274A2", "#3B528B", "#6274A2", "#8997B9",
    "#B0B9D0", "#D7DCE7"),
  sex = NULL,
  patients = NULL,
  patient_color = "yellow",
  patient_point_size = 2,
  ylim = NULL,
  show_reference = TRUE
)

Arguments

Value

a ggplot2 object

Examples

set.seed(60427)


ln_plot_dist("Hemoglobin")

# Plot only females
ln_plot_dist("Creatinine", sex = "female", ylim = c(0, 2))

# Set the ylim
ln_plot_dist("BMI", ylim = c(8, 50))

# Project the distribution of three Hemoglobin values
ln_plot_dist("Hemoglobin", patients = dplyr::sample_n(hemoglobin_data, 3))

# Change the quantiles
ln_plot_dist("Hemoglobin",
    quantiles = seq(0.05, 0.95, length.out = 10)
)

# Change the colors
ln_plot_dist(
    "Hemoglobin",
    quantiles = c(0.03, 0.1, 0.25, 0.5, 0.75, 0.9, 0.97),
    pal = c("red", "orange", "yellow", "green", "blue", "purple")
)

# Change the reference distribution
ln_plot_dist("Hemoglobin", reference = "UKBB")


# on the demo data

Compute the lab value for a given quantile

Description

The function ln_quantile_value calculates lab values at a specified quantile, using the default units for that lab. The function ln_patients_quantile_value does the same calculation for a specific group of patients.
Default units for a lab can be obtained using ln_lab_default_units.
If no quantile data is available for a particular lab, age, and sex, the function returns 'NA'.
It should be noted that the values of extreme quantiles (e.g. >0.95 or <0.05 on low resolution, >0.99 or <0.01 on high resolution) may not be reliable, as they may represent outliers in the data.

Note that ln_quantile_value returns values for all combinations of age, sex, and lab, while ln_patients_quantile_value returns values for a specific set of patients, similar to ln_normalize.

Usage

ln_quantile_value(
  quantiles,
  age,
  sex,
  lab,
  reference = "Clalit",
  allow_edge_quantiles = FALSE
)

ln_patients_quantile_value(
  quantiles,
  age,
  sex,
  lab,
  reference = "Clalit",
  allow_edge_quantiles = FALSE
)

Arguments

Value

ln_quantile_value returns a data frame which contains the values for each combination of quantile, age and sex. The data frame has the the following columns:

• age: age in years

• sex: "male" or "female"

• quantile: he quantile

• value: the lab value

• unit: the lab unit

• lab: the lab name

ln_patients_quantile_value returns a vector of value per patient.

Examples

ln_quantile_value(c(0.05, 0.5, 0.95), 50, "male", "WBC")

ln_quantile_value(
    c(0, 0.05, 0.1, 0.4, 0.5, 0.6, 0.9, 1),
    c(50, 60),
    c("male", "female"),
    "Glucose"
)


# on the demo data



hemoglobin_data$quantile <- ln_normalize(
    hemoglobin_data$value,
    hemoglobin_data$age,
    hemoglobin_data$sex,
    "Hemoglobin"
)

hemoglobin_data$value1 <- ln_patients_quantile_value(
    hemoglobin_data$quantile,
    hemoglobin_data$age,
    hemoglobin_data$sex,
    "Hemoglobin"
)
head(hemoglobin_data)