Data whispering: gently taming complex laboratory data for tuberculosis research

Chad Heilig — Division of Tuberculosis Elimination, CDC, 2013-03-14 — 20 slides

This rendering was constructed from an original PowerPoint presentation. Corrections of original text are indicated as “[corrected: X]”, redactions as “[redacted: X]”, and names withheld as “[generalized: X]”. Handouts and other auxiliary sources mentioned in the presentations are not available. The content of each presentation was believed to be correct and valid at the time of the presentation. This content, including references and links within the slides, has not been checked or updated when preparing this site. This rendering was facilitated by Anthropic Claude agentic tools for automated text extraction through computer vision and for assistance with compiling the completed sets.

Setting: Data whispering: gently taming complex laboratory data for tuberculosis research, 2013-03-14. CDC’s Division of Tuberculosis Elimination.

2013-03-14, sunny and cool. Can I bring The Little Prince‘s apprivoiser to tuberculosis data? These data are partial, a little messy, from real laboratories and real illness. I gave a talk, “Data Whispering: Gently Taming Complex Laboratory Data for Tuberculosis Research”. Gave doesn’t capture it, more like declaimed or manifested for the Division of Tuberculosis Delimitation. Opening slide: “No data were harmed in the making of this presentation.” Closing slide: Learn about the data. Learn from the data. Proceed incrementally, iteratively, responsibly. Build up complex logic through small steps. Refine, evolve; document, vet, guard. This talk feels like everything I want my career to be. I can do good things. I can help this division do good things. My branch chief believes it too; he believed it in me before I had a talk to put it in, back when it was just a way I worked. I ride that high. Who wouldn’t.

This presentation draws lessons from 4 DTBE lab-intensive studies: TBTC Study 28, TBTC NAA2m, ID-TB/HIV (Southeast Asia) and Kenya ICF. They are offered as “Some insights, humbly offered” that “Extend beyond lab data”. For each study it pairs the raw laboratory data with the “Goals for analytic data”. “Sample strategies for taming data” lays out a hierarchy of data structures (test → specimen → encounter → across encounters), events dated “by actual date rather than expected schedule”, and constructs refined “to support multiple analytic goals”. An event-history plot of TBTC Study 24 depicts baseline status, endpoints, durations and adverse events in one graphic.

Works referred to

Cain et al. (2010)Elsewhere on the web (opens in a new tab)
Reves et al. (2014)Elsewhere on the web (opens in a new tab)

Related works, not cited

Heilig et al. (2014)Elsewhere on the web (opens in a new tab)
Joloba et al. (2014)Elsewhere on the web (opens in a new tab)
Mac Kenzie et al. (2011)Elsewhere on the web (opens in a new tab)

Full references: this page · whole site, by domain (opens in a new tab) · whole site, by type of work (opens in a new tab)

Slide 1Data whispering: gently taming complex laboratory data for tuberculosis research

Chad Heilig, PhD
Division of Tuberculosis Delimitation
2013-03-14

No data were harmed in the making of this presentation.

Slide 2Insights from lab-intensive studies

Stories from 4 DTBE studies
- TBTC Study 28
- TBTC NAA2m
- ID-TB/HIV (Southeast Asia ICF)
- Kenya ICF
Salient lessons
- Some insights, humbly offered
- Extend beyond lab data
- Design and conduct of research

Abbreviations: DTBE = Division of Tuberculosis Elimination (CDC); HIV = human immunodeficiency virus; ICF = intensified case finding; ID = Improving Diagnosis of TB in People Living with HIV (study); NAA2m = TBTC Study NAA2m: a TBTC–TBRU collaborative study (Uganda, 2008–2010) of nucleic acid amplification tests and other markers of treatment response; TB = tuberculosis; TBTC = Tuberculosis Trials Consortium (CDC-sponsored).

Slide 3TBTC Study 28, phase 2B trial (1)

Laboratory data
433 participants (HRZE, MRZE), ~20 sites
Sputum at weeks 0, 2, 4, 6, 8*, 12, 16, etc
- 1–3 at 0; 2 at 8*; 1 each at other times
Smear, solid culture, liquid culture, DST
- Variation in lab methods, media, contamination
Between and within sites (e.g., switch to MGIT)
- Reference DST at CDC

Abbreviations: CDC = Centers for Disease Control and Prevention; DST = drug susceptibility testing; HRZE = isoniazid (H), rifampin (R), pyrazinamide (Z), ethambutol (E): the standard 4-drug regimen; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); MRZE = moxifloxacin (M), rifampin (R), pyrazinamide (Z), ethambutol (E): moxifloxacin substituted for isoniazid; TBTC = Tuberculosis Trials Consortium (CDC-sponsored).

Slide 4TBTC Study 28, phase 2B trial (2)

Goals for analytic data
Microbiological eligibility
- Mtb+, pansusceptible on culture at baseline
Endpoints
- Mtb± at end of intensive phase
  - 8–10 weeks therapy; 1–2 specimens; multimedia
- Time to first negative, stable conversion
- Treatment failure

Abbreviations: Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); TBTC = Tuberculosis Trials Consortium (CDC-sponsored).

Slide 5TBTC NAA2m, cohort (1)

Laboratory data
50 participants, UG, AFB 3/4+
Sputum at weeks 0 (×2), 2, 4, 6, 8 (×2), 12, etc
Qualitative results
- AFB smear; LJ, 7H10, 7H11, 7H10S, 7H11S; MGIT
Quantitative results
- CFU on LJ, 7H10, 7H11; DTD on MGIT

Abbreviations: 7H10 = Middlebrook 7H10 agar (solid culture medium for mycobacteria); 7H10S = Middlebrook 7H10 agar, selective (with added antibiotics to suppress contaminants); 7H11 = Middlebrook 7H11 agar (solid culture medium for mycobacteria); 7H11S = Middlebrook 7H11 agar, selective (e.g. Mitchison's selective 7H11); AFB = acid-fast bacilli; CFU = colony-forming unit(s); DTD = days to detection (time until a liquid MGIT culture signals positive); LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); NAA2m = TBTC Study NAA2m: a TBTC–TBRU collaborative study (Uganda, 2008–2010) of nucleic acid amplification tests and other markers of treatment response; TBTC = Tuberculosis Trials Consortium (CDC-sponsored); UG = Uganda.

Slide 6TBTC NAA2m, cohort (2)

Goals for analytic data
Qualitative
- Cultures: Mtb+, contaminated, Mtb–
- Specimens: agreement across cultures
Quantitative
- Cultures: detectable (NC, NG, yes); measure
- Specimens: correlation of CFUs, DTD with CFU

Abbreviations: CFU = colony-forming unit(s); DTD = days to detection (time until a liquid MGIT culture signals positive); Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); NAA2m = TBTC Study NAA2m: a TBTC–TBRU collaborative study (Uganda, 2008–2010) of nucleic acid amplification tests and other markers of treatment response; NC = not countable; NG = no growth; TBTC = Tuberculosis Trials Consortium (CDC-sponsored).

Slide 7Figure: latent class model

Plot of sensitivity against specificity: 5 colored points, each inside an ellipse of the same color, with a legend naming 5 media.

Sensitivity and specificity with respect to latent class model (4)

Convex hulls and 99% confidence ellipses for all 5 media

Description

Panels: 1
Elements: 5 open point markers in 5 colors, each surrounded by an ellipse of the same color, depict sensitivity and specificity for 5 media named in a legend at lower left: LJ (blue), 7H10NS (green), 7H11NS (red), 7H10S (purple), 7H11S (orange)
Horizontal: Specificity, 1.0 to 0.3, decreasing from left to right
Vertical: Sensitivity, 0.3 to 1.0; ellipses span about 0.69 to 1.0
Visible pattern: All 5 ellipses sit above sensitivity of about 0.7. Purple (7H10S) and orange (7H11S) ellipses overlap at the upper left, around specificity 0.75 to 0.85 and sensitivity 0.9 to 1.0, Red (7H11NS) ellipse around specificity 0.5. Green (7H10NS) and blue (LJ) ellipses overlap in the middle, around specificity 0.6 and sensitivity 0.7 to 0.9. Every ellipse is wider than it is tall.

Abbreviations: 7H10NS = Middlebrook 7H10 agar, nonselective (without added antibiotics); 7H10S = Middlebrook 7H10 agar, selective (with added antibiotics to suppress contaminants); 7H11NS = Middlebrook 7H11 agar, nonselective; 7H11S = Middlebrook 7H11 agar, selective (e.g. Mitchison's selective 7H11); LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria).

Slide 8Figure: quantitation

Scatter plot of CFU/mL on a log scale against DTD, points colored by week, with black and gray broken lines through the main band.

Description

Panels: 1
Elements: Open and filled circles in 6 colors depict CFU/mL against DTD; the legend at upper right pairs an open and a filled circle for each of Week 0 (orange), Week 2 (olive), Week 4 (green), Week 6 (cyan), Week 8 (lavender) and Week 12 (pink). Fitted line segments run through the main scatter of points, labeled "LJ / 7H10NS / 7H11NS". Below the log scale, rows of jittered points labeled NC and NG.
Horizontal: DTD, 0 to 40 (ticks every 10); axis line ends at 42; points from about 2 to 42, with additional points beyond the right extent.
Vertical: CFU/mL, log scale 10^1 to 10^7, then 2 categorical rows below, NC and NG
Visible pattern: Points scattered in 3 groups: an oval field from top left to center and 2 bands at labels NC and NG. Colors nearly follow hue gradient from left to right. Line segments depict models fitted to values with observed DTD and CFU.

Abbreviations: 7H10NS = Middlebrook 7H10 agar, nonselective (without added antibiotics); 7H11NS = Middlebrook 7H11 agar, nonselective; CFU = colony-forming unit(s); DTD = days to detection (time until a liquid MGIT culture signals positive); LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); NC = not countable; NG = no growth.

Slide 9ID-TB/HIV, cross-sectional (1)

Laboratory data
~2000 participants w HIV; KH/TH/VN
Specimens: 3 sputum, urine, stool, blood, LN
Smear (not blood), LJ (×2), MGIT (TH/VN)
- Scanty smears
- Limited Mtb ID per specimen; some GenProbe

Abbreviations: HIV = human immunodeficiency virus; ID = Improving Diagnosis of TB in People Living with HIV (study); KH = Cambodia; LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); LN = lymph node; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); TB = tuberculosis; TH = Thailand; VN = Vietnam.

Slide 10ID-TB/HIV, cross-sectional (2)

Goals for analytic data
Incremental yield (incomplete 4×7 per person)
- Across detection methods (smear, LJ, MGIT)
- Across anatomic site (sputum, extrapulmonary)
- Occurrence of exceptional Mtb+ (e.g., stool only)
Reference standard for screening/diagnosis algorithm

Abbreviations: HIV = human immunodeficiency virus; ID = Improving Diagnosis of TB in People Living with HIV (study); LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); TB = tuberculosis.

Slide 11ID-TB/HIV incomplete 4×7

	Sput 1	Sput 2	Sput 3	Urine	Stool	Blood	LNA‡
Smear	•	•	•	•	•	—	•
LJ 1	•	•	•	•	•	•	•
LJ 2	•	•	•	•	•	•	•
MGIT	†	†	†	†	†	†	†

Every specimen
Every specimen in TH, VN (but not KH)
Indicated for LN ≥1 cm (≥2 cm inguinal)

Abbreviations: HIV = human immunodeficiency virus; ID = Improving Diagnosis of TB in People Living with HIV (study); KH = Cambodia; LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); LN = lymph node; LNA = lymph node aspirate; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); TB = tuberculosis; TH = Thailand; VN = Vietnam.

Slide 12Kenya ICF, cross-section (1)

Laboratory data
~800 participants w HIV, western KE
Specimens
- Initial: sputum (spot ×2, morning), stool, LN
- Follow-up: sputum (spot ×2)
Tests
- ZN, fluorescent microscopy, MGIT, Xpert
- Vary by specimen, sputum volume, follow-up

Abbreviations: HIV = human immunodeficiency virus; KE = Kenya; LN = lymph node; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); Xpert = Xpert MTB/RIF (Cepheid GeneXpert cartridge nucleic acid amplification test for TB and rifampicin resistance); ZN = Ziehl–Neelsen (acid-fast stain for smear microscopy).

Slide 13Kenya ICF, cross-section (2)

Goals for analytic data
Incremental yield
- Across method (smear, Xpert, MGIT)
- Across specimen multiplicity/type
Reference standard for WHO screening

Abbreviations: MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); WHO = World Health Organization; Xpert = Xpert MTB/RIF (Cepheid GeneXpert cartridge nucleic acid amplification test for TB and rifampicin resistance).

Slide 14[corrected: Kenya ICF 4×7]

	Initial visit					Follow-up visits
	Spot Sput 1	Spot Sput 2	Morn Sput	LNA‡	Stool	Spot Sput 1	[corrected: Spot Sput 2]
Fluor	•		•	•	•	•	•
MGIT	•		•	•	•	•	•
ZN		•	•
Xpert		•	†

Every specimen
If sufficient volume
Indicated for LN ≥1 cm (≥2 cm inguinal)

Abbreviations: ICF = intensified case finding; LN = lymph node; LNA = lymph node aspirate; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); Xpert = Xpert MTB/RIF (Cepheid GeneXpert cartridge nucleic acid amplification test for TB and rifampicin resistance); ZN = Ziehl–Neelsen (acid-fast stain for smear microscopy).

Slide 15Sample strategies for taming data

Hierarchy of data structures
- Within test (smear, culture, Xpert)
- Within specimen, across tests
- Within participant encounter*, across specimens
- Across participant encounters
Date-based events by actual date rather than expected schedule
Define, refine data constructs to support multiple analytic goals (use experience)

Abbreviations: Xpert = Xpert MTB/RIF (Cepheid GeneXpert cartridge nucleic acid amplification test for TB and rifampicin resistance).

Slide 16TBTC Study 24 dates

Timeline chart with one row per participant, showing treatment periods, follow-up, culture results and adverse events over days.

Baseline Mtb; potential Mtb endpoints; Tx durations; AE timing

Description

Panels: 1
Elements: 27 horizontal rows, one per participant, each made of line segments in 3 styles and point markers in 5 symbols, as the legend at upper right names them: thin light gray line, Prestudy treatment; thick gray line, Study treatment; dotted line, Follow-up; filled circle, Positive culture; open circle, Negative culture; diamond with a cross, Treatment failure/relapse; boxed x, Hepatic adverse event; x, Nonhepatic adverse event. Rows are grouped under 4 headings: "Incomplete treatment, incomplete follow-up < 6 months" (8 rows), "Incomplete treatment, incomplete follow-up ≥ 6 months" (2 rows), "Incomplete treatment, complete follow-up" (15 rows), "Complete treatment, late positive culture" (2 rows)
Horizontal: Days since start of TB therapy, -30 to 390, ticks every 30
Vertical: Participant index, 1 to 27 from top to bottom, in the 4 labeled groups; no numeric scale
Visible pattern: Filled circles (positive cultures) cluster at or just before day 0 in nearly every row. Thin light gray lines run from day 0 to about day 60 in most rows, and thick gray study-treatment segments begin around day 30 to 70 and end anywhere from about day 80 to 200. Hepatic adverse event markers fall between about day 70 and 175. Treatment failure/relapse diamonds appear in 3 rows, 1 near day 120 and 2 near day 325.

Text in the picture (summary)

Study participants, renumbered 1 to 27.

Abbreviations: AE = adverse event; Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); TB = tuberculosis; TBTC = Tuberculosis Trials Consortium (CDC-sponsored); Tx = treatment.

Slide 17Generalized goals for taming data

Learn about data: analyze before, during, after
Learn from data: analyze during, after
Proceed incrementally, iteratively, responsibly
- Small steps, build up complex logic
  - Modular: Mtb/NTM/contamination; DST; doses
- Refine, evolve
- Document, vet, guard

Abbreviations: DST = drug susceptibility testing; Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); NTM = nontuberculous mycobacteria.

Slide 18Recurring complex data, beyond lab

More lab and diagnostic data (pediatric)
Ill-defined endpoints (hypersensitivity)
Flexibility in date-based events (SQL!)
- TBTC DOT, initial and continuation?
- TBTC self-administered 12-dose regimen
- TBESC LTBI cohort
- Diagnostic and treatment patterns
- Endpoints in DR TB

Abbreviations: DR = drug-resistant (DR TB = drug-resistant tuberculosis); LTBI = latent tuberculosis infection; SQL = Structured Query Language; TB = tuberculosis; TBESC = Tuberculosis Epidemiologic Studies Consortium (CDC); TBTC = Tuberculosis Trials Consortium (CDC-sponsored).

Slide 19Discussion

A redacted region below the title and the web address.

[redacted: email address]
www.cdc.gov/tb

Slide 20S28 date-based SQL code snippet

Screenshot of code in an editor, with syntax coloring, formatting marks shown and the first line highlighted in light blue.

Description

Text in the picture (summary)

SAS/SQL code with a 15-line comment and 21-line Proc SQL Select statement to determine culture conversion and stable conversion.

Abbreviations: S28 = TBTC Study 28 (moxifloxacin vs isoniazid in the intensive phase of TB treatment); SQL = Structured Query Language.

Abbreviations

7H10 = Middlebrook 7H10 agar (solid culture medium for mycobacteria); 7H10NS = Middlebrook 7H10 agar, nonselective (without added antibiotics); 7H10S = Middlebrook 7H10 agar, selective (with added antibiotics to suppress contaminants); 7H11 = Middlebrook 7H11 agar (solid culture medium for mycobacteria); 7H11NS = Middlebrook 7H11 agar, nonselective; 7H11S = Middlebrook 7H11 agar, selective (e.g. Mitchison's selective 7H11); AE = adverse event; AFB = acid-fast bacilli; CDC = Centers for Disease Control and Prevention; CFU = colony-forming unit(s); DR = drug-resistant (DR TB = drug-resistant tuberculosis); DST = drug susceptibility testing; DTBE = Division of Tuberculosis Elimination (CDC); DTD = days to detection (time until a liquid MGIT culture signals positive); HIV = human immunodeficiency virus; HRZE = isoniazid (H), rifampin (R), pyrazinamide (Z), ethambutol (E): the standard 4-drug regimen; ICF = intensified case finding; ID = Improving Diagnosis of TB in People Living with HIV (study); KE = Kenya; KH = Cambodia; LJ = Löwenstein–Jensen (egg-based solid culture medium for mycobacteria); LN = lymph node; LNA = lymph node aspirate; LTBI = latent tuberculosis infection; MGIT = Mycobacteria Growth Indicator Tube (liquid culture system); MRZE = moxifloxacin (M), rifampin (R), pyrazinamide (Z), ethambutol (E): moxifloxacin substituted for isoniazid; Mtb = Mycobacterium tuberculosis (the bacterium that causes TB); NAA2m = TBTC Study NAA2m: a TBTC–TBRU collaborative study (Uganda, 2008–2010) of nucleic acid amplification tests and other markers of treatment response; NC = not countable; NG = no growth; NTM = nontuberculous mycobacteria; S28 = TBTC Study 28 (moxifloxacin vs isoniazid in the intensive phase of TB treatment); SQL = Structured Query Language; TB = tuberculosis; TBESC = Tuberculosis Epidemiologic Studies Consortium (CDC); TBTC = Tuberculosis Trials Consortium (CDC-sponsored); TH = Thailand; Tx = treatment; UG = Uganda; VN = Vietnam; WHO = World Health Organization; Xpert = Xpert MTB/RIF (Cepheid GeneXpert cartridge nucleic acid amplification test for TB and rifampicin resistance); ZN = Ziehl–Neelsen (acid-fast stain for smear microscopy).

References, compiled for this page

Full citations, with current links, for selected references and related works.

Full references: whole site, by domain (opens in a new tab) · whole site, by type of work (opens in a new tab)

Cain KP, McCarthy KD, Heilig CM, Monkongdee P, Tasaneeyapan T, Kanara N, Kimerling ME, Chheng P, Thai S, Sar B, et al. An algorithm for tuberculosis screening and diagnosis in people with HIV. New England Journal of Medicine. 2010-02-25; 362(8): 707–716. (slide 14)Elsewhere on the web (opens in a new tab)
Reves R, Heilig CM, Tapy JM, Bozeman L, Kyle RP, Hamilton CD, Bock N, Narita M, Wing D, Hershfield E, et al. Intermittent tuberculosis treatment for patients with isoniazid intolerance or drug resistance. International Journal of Tuberculosis and Lung Disease. 2014-05-01; 18(5): 571–580. (slide 16)Elsewhere on the web (opens in a new tab)

Related works, not cited

Heilig CM, Feng P-JI, Joloba ML, Johnson JL, Morgan K, Gitta P, Boom WH, Mayanja-Kizza H, Eisenach KD, Bozeman L, Goldberg SV. How we determined the most reliable solid medium for studying treatment of tuberculosis. Tuberculosis. 2014-05; 94(3): 317–322.Elsewhere on the web (opens in a new tab)
Joloba ML, Johnson JL, Feng P-JI, Bozeman L, Goldberg SV, Morgan K, Gitta P, Boom HW, Heilig CM, Mayanja-Kizza H, Eisenach KD. What is the most reliable solid culture medium for tuberculosis treatment trials? Tuberculosis. 2014-05; 94(3): 311–316.Elsewhere on the web (opens in a new tab)
Mac Kenzie WR, Heilig CM, Bozeman L, Johnson JL, Muzanye G, Dunbar D, Jost KC, Diem L, Metchock B, Eisenach K, et al. Geographic differences in time to culture conversion in liquid media: Tuberculosis Trials Consortium study 28. Culture conversion is delayed in Africa. PLOS ONE. 2011-04-11; 6(4): e18358.Elsewhere on the web (opens in a new tab)