Prof. Lieberman's presentation at the 5th Latin American Future Trends in Medicine meeting describes a new method for testing medicines in the field with paper test cards. The presentation has been shortened by removing some videos.
Seismic Method Estimate velocity from seismic data.pptx
Lieberman paper test card linked in version
1. Finding fake medicines
Encontrar medicamentos falsos
Prof. Marya Lieberman
Department of Chemistry and Biochemistry
University of Notre Dame
Notre Dame, IN 46556
mlieberm@nd.edu
con mi amigo, Google Translate
2.
3. Drug discovery vs. drug delivery
low quality antimalarial
medicine causes 122,000
deaths each year
in kids under 5 ¿POR QUÉ?
Renschler et al. 2015 Am. J. Trop. Med. Hyg.
Genuine coartem cure rate for malaria: 97% (genuina Coartem cura la malaria)
Falsified coartem (top) antimalarial tablet contains <4% API (y falsa no cura nada)
Coartem can be assayed by HPLC in 15
minutes with a compendium assay
El análisis químico (HPLC) puede detectar
todos fármacos de baja calidad
4. • Has any person at your table ever found a
falsified medication? What type of medicine
and how did they find it?
• Hay alguna de las personas en su mesa que se
ha encontrado un medicamento falsificado ?
¿Qué tipo de medicamento y cómo se
encuentran ?
5. if you see a this…there must be a that
si usted ve un “esta” ... debe haber una “que”
Interpol 2014
low risk for manufacturers
profit attracts criminal enterprise
Kenya 2013
Poor regulatory infrastructure
lack of testing capacity
Uganda 2013
supply chain problems
6. constraints and prior solutions
Shimadzu HPLC at Moi U
Broken for past 5 years
sitting unused
in a corner of
Ghanaian FDA
broken instruments
no trained staff
no supply chain
Trying to buy glassware
in Eldoret, Kenya
Constraints for field tests:
fast
easy
cheap
no power
no instruments
no lab equipment
no handling of chemicals
results available/archived
limitaciones de las pruebas:
rápido
fácil
barato
ninguna energía
no hay instrumentos
sin equipo de laboratorio
sin manipulación de productos
químicos
resultados disponibles / archivado
7. Analytical challenges
Many medicines have similar structures
90-120%
343 WHO Essential Medicines
Dosage forms are mixtures (insoluble
excipients, fixed-dose combinations)
Substandard formulations are common
Pills and tablets are solids
8. Wax printing
• hydrophobic wax defines channels
• reagents are stored in paper
• passive pumping through capillary action
• filter, mix, or separate based on channel geometry
Yagoda, H. Ind. Eng. Chem. 1937, 9, 79–82; Carrilho, Martinez, and
Whitesides, Anal. Chem. 2009, 81, 7091–7095
El papel se imprime con tinta cerosa y
después se calentó para fundir la cera.
Es una forma barata de hacer que los
dispositivos de microfluidos.
10. These color tests can identify drugs
that have very similar structures
Estas pruebas de color pueden identificar fármacos que
tienen estructuras muy similares
i. Beta lactam + amide (basic copper test)
ii. Primary amine (ninhydrin)
iii. Phenol (diazo coupling)
13. What’s in a falsified medication?
Active Pharmaceutical
Ingredient (API)
+ →
PharmaceuticalExcipient(s)
• Adulterants/fillers
• chalk
• maize meal
• gypsum
• talcum powder
• No API
• Incorrect dose API
• Wrong API
• Degraded API
14. Please form 12 teams
For each team: Para cado equipo:
• 6 PADs
• pure isoniazid (INH)
• 4 “unknown” samples
• wood sticks (palitos)
• small water dishes (platillos para el agua)
15. 1) Run pure INH and pure water
2) Run 4 unknowns Demo
Test card containing
preloaded dry reagents
Tarjeta de ensayo que
contiene precargado
reactivos secos
Step 1: Crush tablet
and apply to card
You should see powder
in each lane
Paso 1 : se aplica el
polvo a la tarjeta
Usted debe ver el polvo
en cada carril
Step 2: Dip card in water
for 3 minutes
Paso 2 : Coloque el borde
inferior de la tarjeta en
agua durante 3 minutos
Colors develop in
1-3 minutes
colores aparecen
en 1-3 minutos
sample name sample name sample name sample name
Use wooden stick to wipe powder firmly
across the paper
Isoniazid (INH)
Instrucciones de uso del PAD
16. B, G, H: Orange-red (rojo anaranjado
C, F, I: Green (verde)
J: Starch gives black color that does not move
El almidón da color negro que no se mueve
* Does the sample contain INH? (Hay
alguna INH?)
* Does the sample contain anything
that shouldn’t be there? (¿Hay algo
sospechoso?)
17. B, G, H: Orange-red (rojo anaranjado
C, F, I: Green (verde)
J: Starch gives black color that does not move
El almidón da color negro que no se mueve
* Does the sample contain INH? (Hay
alguna INH?)
* Does the sample contain anything
that shouldn’t be there? (¿Hay algo
sospechoso?)
1 INH 30% Starch 70%
2 Antipyrine 50% Rice Flour 50%
3 Paracetamol
4 INH
18. B, G, H: Orange-red (rojo anaranjado
C, F, I: Green (verde)
J: Starch gives black color that does not move
El almidón da color negro que no se mueve
* Does the sample contain INH? (Hay
alguna INH?)
* Does the sample contain anything
that shouldn’t be there? (¿Hay algo
sospechoso?)
1 INH 30% Starch 70%
2 Antipyrine 50% Rice Flour 50%
3 Paracetamol
4 INH
19. Sensitivity: if it’s there, do you see it? Ideal = 100%
Sensibilidad: si está allí , ¿lo ves? Ideal = 100 %
0 10 20 30 40 50 60 70 80 90 100
e- rich phenols
Starch
Talc
Tertiary amines
Baking Soda
Chalk (med-heavy)
Acetaminophen
Pyrazinamide
Isoniazid
Rifamicin
Ethambutol
Ampicillin
Amoxicillin
Beta-lactam
Specificity: if it’s not there, do you not see it? Ideal = 100%
Especificidad: si no está ahí , ¿no lo ves ?
Every pure API detected
with 92-100% sensitivity
and 88-100% specificity
but reading by eye requires
expert readers
Pero esto requiere lectores
expertos
the logistics won’t work
at scale
20. What the user
does not see
2
3
5. Geo-tracking4. Image Analysis
Time
GoodSamples
6. Archiving and
Monitoring
Data collection with cell phones
1
21. Image analysis goals:
• Classification: Assign test images to
correct class based on stored “training”
images
Clasificación : Asigne imágenes de prueba
para corregir clase basada en imágenes de
"entrenamiento “
• Quantification: Measure color ratios or
intensities
Cuantificación : Mida las proporciones de
color o intensidad
• Adaptive learning: Search for patterns in
data sets Aprendizaje adaptativo : Búsqueda
de patrones en conjuntos de datos
22. Fixing bad pictures/corregir malas fotos
“wild type” image
rotation, keystoning,
shadows, color distortion
Image re-sized, aligned,
lanes identified
Prof. Pat Flynn and Prof. Chris Sweet
23. Color bar code
Computer image analysis
Compare unknowns to stored
“authentic” bar codes using neural
network. Asigne imágenes de prueba
para corregir clase basada en
imágenes de "entrenamiento “
• trained on 20 samples each of
Acetaminophen, Acetylsalicylic Acid,
Amodiaquine, Amoxicillin, Ampicillin,
Artesunate, Calcium Carbonate, Corn
Starch, Diethylcarbamazine, Ethambutol,
Isoniazid, Rifampicin, Tetracycline
• Tested with 10 samples each of same
drugs (N=130)
Sandipan Banerjee and Chris Sweet
24. Computer can classify accurately
Counterfeit Drug Detection with Paper Analytical Device Images using Deep
Learning; S. Banerjee, J. Sweet, C. Sweet, WACV 2016 submitted
Acetaminophen
Aspirin
Amodiaquine
Amoxycillin
Ampicillin
Artesunate
CaCO3
Corn Starch
DEC
Ethambutol
Isoniazid
Rifampicin
Tetracycline
Acetaminophen5/10
Aspirin10/10
Amodiaquine10/10
Amoxycillin10/10
Ampicillin3/10
Artesunate3/10
CaCO36/10
CornStarch9/10
DEC10/10
Ethambutol10/10
Isoniazid10/10
Rifampicin10/10
Tetracycline10/10
Actual active ingredient (and number classified correctly)
Howsurewastheneuralnetwork?
25. Color tests can detect “fillers”
Pruebas de color se pueden detectar sólidos insolubles
• talc
eosin red dye cherry
• starch, flour
I2 blue/black
• chalk, baking soda, calcite
Fe(III) Fe2O3
Weaver et al. Analytical
Chemistry, 2013, 85 (13),
6453–6460
26. Pharmaceuticals in “herbal” medicines
Productos farmacéuticos en medicamentos “a base de hierbas"
Samples from Israeli
Ministry of Health—
Division of Enforcement
and Inspection
“Ingredients:Mulberry
leaf extracts, jobstears
seed, medical amylum”
27. A B C D E F G H I J K L
Phenolphthalein
laxative use was
banned in 1999
Samples from Israeli
Ministry of Health—
Division of Enforcement
and Inspection
Sibutramine
anorexiant
banned in 2010
“herbal” medicine
medicamentos “a base de
hierbas"
28. Quantification is hard
• API range 90%-120% “meets standard”
50
70
90
110
130
150
170
190
100%
Chloroquine
(CQ)
70% CQ,
30% chalk
40% CQ,
60% starch
0% CQ,
100% aspirin
0% CQ,
100% chalk
Min Outlier Max Outlier
Chloroquine (CQ) 70% CQ, 30% chalk 40% CQ, 60% starch
0% CQ, 100% aspirin
0% CQ, 100% chalk
Weaver et al, AJTMH 2015
Colorintensity
29. Iodometric titration on a paper card
degradation
products
eg RSH
add known
amount of I2
thiosulfate/starch on
test card performs
back-titration
KOH
20 minAmoxicillin
(este PAD cuantifica el agente de yodación de la sal)
30. Quantification of beta lactams via USP <425>
degradation
products
eg RSH
add known
amount of I2
back-titration with
thiosulfate/starch
KOH
20 minAmoxicillin
31. degradation
products
eg RSH
add known
amount of I2
thiosulfate/starch on
test card performs
back-titration
KOH
20 minAmoxicillin
N. Myers, unpublished data
50% API 75% API 90% API 95% API 100% API
Substandard antibiotics don’t react with all the iodine.
The more I2 is left over, the more dots turn blue
Antibióticos deficientes no reaccionan con todo el yodo .
El más yodo sobra , más puntos se vuelven azules
32. Can we analyze medicines in the real world?
¿Podemos analizar los medicamentos en el mundo real?
33. Question for every table
Pregunta para cada mesa
• If you could test the quality of five medicines
in your country, which five would you pick,
and why?
• Si se pudiera probar la calidad de cinco
medicamentos en su pais, los cuales cinco
elegirías y por qué?
34. Pharmacists at Moi Hospital and staff at Kenyan Pharmacy
and Poisons Board (KPPB) chose ampicillin, amoxycillin,
amoxycillin/clavulanate, ciprofloxacin, and azithromycin
401 brands in Kenya
O.M.G.
Very low quality
missing API
<50% API
substitute API
serious risk
to patients
and public health
Substandard
does not meet
pharmacopeia
standards
risk of harm
to patients
and public health
Good quality
Meets
pharmacopeia
standards
37. PADs run in Kenya
one of these things is
not like the other ones
una de estas cosas no es
como las demás
38. Confirmatory analysis by HPLC:
Amoxycillin 500 mg, clavulanate 125 mg
Suspicious
Normal
Standards
Rebecca Ryan
39. Impacts
Secret shoppers buy
antibiotics at pharmacies in
Western Kenya
Kenyan Pharmacy
and Poisons Board
4 reports to KPPB
2 reports to WHO
Screen medicines at
MTRHAdverse drug reactions
from MTRH clinics
HPLC at ND
167 assays, 57
substandard of which
14 lacked an API
43. Graduate students: Nicholas Myers, Sandipan Banerjee,
James Sweet, Dr. Abigail Weaver, Jamie Luther
Undergraduate students: Kate Girdhar, Margaret Berta, Esseatim Etim
(Winthrop), Sarah Bliese (Hamline), Rebecca Ryan, Steven Froelich,
Hannah Reiser, Kellie Radell, Eliza Herrero, Leah Koenig
Collaborators: Sonak Pastakia, Rahki Kharwa, Mercy Maina, Celia
Ngetich, Phelix Were, and Beatrice Jakait (AMPATH/MTRH Kenya) Chris
Sweet, Center for Research Computing, UND
Cheryl Fleurer and Mark Witkowski, FDA/FCC
Deanna O’Donnell, Hamline U. SBPD Mickey Arieli, IMOH
$$: University of Notre Dame,
Global Alliance for Improved Nutrition,
Lilly Endowment
Bill & Melinda Gates Foundation
US-AID DIV program
Hinweis der Redaktion
artemisinin drugs Nobel 2015
Sub-saharan africa: avg. per capita total health expenditures $32 (US $6,719)
countries that have functional MRAs 135 /218
WHO analysis of 26 MRAs in sub-saharan Africa:
17 countries had a functional QC laboratory, 5 of those labs had documentation of their analysis procedures.
Only 5 countries had a systematic post market testing plan; 7 tested samples when someone complained.
8 countries collected adverse effects reports, but they didn’t act effectively on the results.
More pictures—sad sack pictures
Strengths—Everything is included; orthogonal testing methods (color tests + TLC); small scale so less chemical handling; commercially available, high penetrance/distribution; detects substitute ingredients and lack of API
Weaknesses—not quantitative, need consumable supplies, testing one substance can take hours, high knowledge/skill level required to carry out & interpret tests, ambiguous results common
>4,000 euros + consumable supplies, 1-2 weeks training
Transition: need to analyze solid dosage forms (consider the constraints of the analysis…)
falsified formulations may include unexpected excipients or APIs…
Qualitative analysis can reveal discrepant formulations (eg trout in the milk)
quantitative analysis of API will show more substandards
about this time there was a paper by whitesides describing how to print wax on paper to make sophisticated devices for enzyme assays (DFA)
24 lane PAD
Nothing from praziquantel, mebendazole
simplifies logistics
Sustaining chemistry in the developing world will lead to monitoring over geographical areas over time.
AT least the neural network knows when it knows something!!
transition: so far we have looked at detection of single, pure APIs. What about dosage forms?
ability to detect insoluble excipients is valuable (thoreau’s trout in the milk)
pharmacists at Moi Hospital and KPPB wanted to study five common medicines in Western Kenya
I did a brand survey
81 paracetamol,
108 amoxicillin,
72 amoxy-clav,
89 ciprofloxacin,
51 azithromycin
This is a question about values!
I did a brand survey
81 paracetamol,
108 amoxicillin,
72 amoxy-clav,
89 ciprofloxacin,
51 azithromycin
Minilab—713 in use, started in 1997, 22 very low quality products showcased on their website
458 samples from Kenya sent to US in 2014
vision: make cheap cards that could be used in a low resource setting by untrained people
will produce some visual readout, it will be complicated if we want to find out complex questions, but we can get around that…
Make use of the mobile phone network to archive/interpret data
large scale data collection