sensors

Article A Portable Real-Time Ringdown Breath Analyzer: Toward Potential Diabetic Screening and Management

Chenyu Jiang 1, Meixiu Sun 1, Zhennan Wang 1, Zhuying Chen 1, Xiaomeng Zhao 1, Yuan Yuan 1, Yingxin Li 1,* and Chuji Wang 1,2,*

1 Institute of Biomedical Engineering, Chinese Academy of Medical Sciences & Peking Union Medical College, Tianjin 300192, China; [email protected] (C.J.); [email protected] (M.S.); [email protected] (Z.W.); [email protected] (Z.C.); [email protected] (X.Z.); [email protected] (Y.Y.) 2 Department of Physics and Astronomy, Mississippi State University, Starkville, MS 39759, USA * Correspondence: [email protected] (Y.L.); [email protected] (C.W.); Tel.: +86-22-8789-1131 (Y.L.); +1-662-325-9455 (C.W.)

Academic Editor: Andrea Facchinetti Received: 10 May 2016; Accepted: 15 July 2016; Published: 30 July 2016

Abstract: Breath analysis has been considered a suitable tool to evaluate diseases of the respiratory system and those that involve metabolic changes, such as diabetes. Breath acetone has long been known as a biomarker for diabetes. However, the results from published data by far have been inconclusive regarding whether breath acetone is a reliable index of diabetic screening. Large variations exist among the results of different studies because there has been no “best-practice method” for breath-acetone measurements as a result of technical problems of sampling and analysis. In this mini-review, we update the current status of our development of a laser-based breath acetone analyzer toward real-time, one-line diabetic screening and a point-of-care instrument for diabetic management. An integrated standalone breath acetone analyzer based on the cavity ringdown technique has been developed. The instrument was validated by using the certificated gas chromatography-mass spectrometry. The linear fittings suggest that the obtained acetone concentrations via both methods are consistent. Breath samples from each individual subject under various conditions in total, 1257 breath samples were taken from 22 Type 1 diabetic (T1D) patients, 312 Type 2 diabetic (T2D) patients, which is one of the largest numbers of T2D subjects ever used in a single study, and 52 non-diabetic healthy subjects. Simultaneous blood glucose (BG) levels were also tested using a standard diabetic management BG meter. The mean breath acetone concentrations were determined to be 4.9 ˘ 16 ppm (22 T1D), and 1.5 ˘ 1.3 ppm (312 T2D), which are about 4.5 and 1.4 times of the one in the 42 non-diabetic healthy subjects, 1.1 ˘ 0.5 ppm, respectively. A preliminary quantitative correlation (R = 0.56, p < 0.05) between the mean individual breath acetone concentration and the mean individual BG levels does exist in 20 T1D subjects with no ketoacidosis. No direct correlation is observed in T1D subjects, T2D subjects, and healthy subjects. The results from a relatively large number of subjects tested indicate that an elevated mean breath acetone concentration exists in diabetic patients in general. Although many physiological parameters affect breath acetone, under a specifically controlled condition fast (<1 min) and portable breath acetone measurement can be used for screening abnormal metabolic status including diabetes, for point-of-care monitoring status of ketone bodies which have the signature smell of breath acetone, and for breath acetone related clinical studies requiring a large number of tests.

Keywords: cavity ringdown breath analyzer; breath acetone; high data throughout; GC-MS validation; elevated mean acetone concentration

Sensors 2016, 16, 1199; doi:10.3390/s16081199 www.mdpi.com/journal/sensors Sensors 2016, 16, 1199 2 of 15

1. Introduction Breath analysis by testing the volatile organic compounds (VOCs) of exhaled breath, one of the three key clinical diagnostic techniques that have been used in Eastern Medicine for more than three thousand years, provides a potential non-invasive method for disease diagnosis, therapeutic monitoring, and metabolic status monitoring [1–5]. To date, there are more than 2000 VOCs in low concentrations, from parts per million (ppm) to parts per billion (ppb) or parts per trillion (ppt), identified to be present in exhaled human breath. Breath acetone has long been known as a breath biomarker for diabetes [6–17]. Effort on breath acetone analysis using various methods and technologies toward diabetes diagnostics and monitoring has been made over the past 50 years. However, huge variations exist among the results from different studies [6]. One of the main reasons is that those studies used a limited number of human subjects or samples, which is partially due to high costs of breath analysis using the convention analytical methods such as GC-MS and long testing times resulting from sophisticated sample preparation such as sample pre-concentration as required by the methods. To this end, it is highly desirable to have a high data throughput instrument or method that is able to test a sufficiently large number of diabetic subjects in a relatively short experimental time. Recent efforts in pursuits of a real-time on-line breath analyzer with high data throughput are fruitful [9,11,18–28]. Several relatively new MS-based analytical techniques have been developed and tested for on-line breath analysis, including proton transfer reaction mass spectrometry (PTR-MS) [22,23], vacuum-free ion mobility spectroscopy (IMS) [24,25], and selected ion flow tube mass spectrometry (SIFT-MS) [9,11,26]. Breath analysis is also conducted by using electrical sensors, which are comparatively inexpensive and smaller in size [27,28], but the issue with detection selectivity and requirement of frequent baseline calibration remain to be further addressed. More recently, to pursue a large scale of clinical testing in near-real time, on-line toward high data throughput, we have designed and constructed a breath acetone analyzer based on the cavity ringdown spectroscopy (CRDS) technique [29–35]. As a laser spectroscopy-based technique, CRDS has been successfully used for trace gas analysis including breath analysis. Compared to the use of the GC-MS method, the CRDS method is well accepted for trace gas analysis on account of its advantages of high sensitivity, high accuracy, real-time response without need of sample pre-process, and relatively low cost, which make the technique both scientifically and economically attractive for breath analysis, particularly when a large number of subjects have to be tested. Now a ringdown breath acetone analyzer has been developed recently and applied to measure breath acetone concentration of human subjects in a laboratory and a clinic. Breath acetone in human, canine, and muridae subjects has been measured using the CRDS technique in the previous studies [35–41]; however, none of the large volume of data (1000 or more) was measured by a real-time on-line ringdown breath acetone analyzer. In this work, we update the status of the development of a standalone ringdown breath acetone analyzer and its clinical application in measurements of breath acetone from 22 hospitalized Type 1 diabetic (T1D) patients, 312 hospitalized Type 2 diabetic (T2D) patients, and 52 non-diabetic healthy subjects under various situations (fasting or post-meals). Thousands of data collected by the real-time on-line standalone ringdown breath analyzer in a short experimental period are reported here. Potential applications of the real-time portable breath acetone analyzer (LaserBreath-001) in diabetes screening and management are discussed.

2. Experimental

2.1. Ringdown Breath Acetone Analyzer The integrated standalone breath acetone analyzer (LaserBreath-001) developed in our lab is illustrated in Figure1. All displays, including the ringdown decay waveform and real-time ringdown signals, are controlled via the touch screen. Figure2 shows a schematic diagram of the optical layout of in the breath acetone analyzer. A Q-switch Nd:YAG laser (Changchun New Industries Optoelectronics, China) equipped with a compact laser head (21.0 mm ˆ 88 mm ˆ 74 mm) was employed as the Sensors 2016, 16, 1199 3 of 15 Sensors 2016, 16, 1199 3 of 15 Sensors 2016, 16, 1199 3 of 15 lightemployed source as in the the light breath source analyzer. in the The breath laser analyzer operated. The at 266laser nm operated with a repetitionat 266 nmrate with of a 1 repetition kHz and employedsingle-pulserate of 1 kHz as energyandthe lightsingle-pulse of source 4.5 µJ. in energy Except the breath of for 4.5 an analyzerµJ. iris Except and. The twofor anlaser reflective iris operated and two mirrors atreflective 266 (Thorlabs nm mirrorswith PF05-03-F01, a repetition(Thorlabs rateNewton,PF05-03-F01, of 1 kHz NJ, USA)andNewton, single-pulse used NJ, to directUSA) energy theused laser of to 4.5 beam,direct µJ. Except thethe outputlaser for beam,an of iris the andthe laser outputtwo was reflective coupled of the lasermirrors into thewas ringdown(Thorlabs coupled PF05-03-F01,cavityinto the without ringdown Newton, using cavity any NJ, mode-matching USA) without used using to direct optical any th mode-matching parts.e laser The beam, gas the cell, optical output was comprised parts.of the Thelaser of gas awas 50-cm-long cell, coupled was intostainlesscomprised the steelringdown of a pipe 50-cm-long with cavity an innerstainlesswithout diameter steelusing pipe ofany 3.81 with mode-matching cm an (CRD inner Optics diameter optical Inc., of Lompoc, 3.81parts. cm CA,The (CRD USA),gas Optics cell, one Inc.,was pair comprisedofLompoc, mirror CA, mounts of aUSA), 50-cm-long (CRD one Opticspair stainless of Inc.,mirror Lompoc, steel mounts pipe CA, (CRDwith USA), an Optics inner and Inc., onediameter Lompoc, pair of of high-reflectivity 3.81CA, USA),cm (CRD and Optics UV one mirrors pair Inc., of Lompoc,(Loshigh-reflectivity Gatos CA, Research, USA), UV onemirrors San pair Jose, (Losof CA,mirror Gatos USA, mounts Research, R = 99.99%, (CRD San Optics radiusJose, CA, Inc., of curvatureUSA, Lompoc, R = 99.99%, =CA, 1 m). USA), Eachradius and mirror of one curvature pair mount of high-reflectivityheld= 1 m). a mirror Each mirror whose UV mirrorsmount position held (Los could aGatos mi berror adjustedResearch, whose in positionSan multiple Jose, could CA, dimensions USA, be adjusted R = via99.99%, threein multiple radius alignment of dimensions curvature screws. =Asvia 1 shownm).three Each alignment in Figuremirror2 ,mountscrews. three portsheld As shown werea mirror located in Figurewhose at this 2,position three gas cell, ports could one were be was adjusted located used as atin a gasthismultiple inletgas cell, anddimensions one another was viatwoused three were as a alignment gasgas outlets, inlet andscrews. which another wereAs shown two connected were in Figure gas to a outlets, vacuum2, three whichports pump werewere (Oerlikon, located connected SC5D, at this to Cologne, gasa vacuum cell, Germany) one pump was usedand(Oerlikon, aas micro a gasSC5D, pressure inlet Cologne, and manometer another Germany) two (MKS, wereand 870B, a gas micro Andover,outlets, pressure which MA, manometer USA)were usedconnected (MKS, to monitor 870B,to a vacuumAndover, the inside pump MA, cell (Oerlikon,pressure.USA) used In SC5D, orderto monitor Cologne, to ensure the Germany) theinside gas cellcell and protectedpressure. a micro from Inpressure order contaminated, manometerto ensure high-purity the (MKS, gas 870B,cell protectedAndover, (>99.99%, fromMA, USA)DaFangcontaminated, used Special to high-puritymonitor Gas, Tianjin, the nitrogen inside China) cell(>99.99%, was pressure. used DaFang to flush In order theSpecial gas to cellGas,ensure each Tianjin, the time China)gas after cell one was protected breath used to sample fromflush contaminated,wasthe gas tested. cell Aeach miniaturehigh-purity time after photomultiplier nitrogen one breath (>99.99%, sample tube DaFang (PMT)was tested. (R74000U-09, Special A miniature Gas, Tianjin, Hamamatsu, photomultiplier China) Japan) was used wastube to used (PMT) flush as the(R74000U-09, gas detector cell each to Hamamatsu, capture time after the Japan) ringdownone breath was decay usedsample as waveform thewa sdetector tested. that Ato was miniaturecapture input the tophotomultiplier ringdown an oscilloscope decay tube (Tektronixwaveform (PMT) (R74000U-09,TDS3012B,that was input Beaverton, Hamamatsu, to an oscilloscope OR, USA)Japan) for(Tektronix was display. used as TDS3012B, the detector Beaverton, to capture OR, the USA) ringdown for display. decay waveform that was input to an oscilloscope (Tektronix TDS3012B, Beaverton, OR, USA) for display.

Figure 1. Schematic of the integrated standalone CRDS breath acetone analyzer (LaserBreath-001). Figure 1. Schematic of the integrated standalone CRDS breath acetone analyzer (LaserBreath-001). Figure 1. Schematic of the integrated standalone CRDS breath acetone analyzer (LaserBreath-001).

Figure 2. Schematic diagram of the optical layout in the ringdown breath analyzer. Figure 2. Schematic diagram of the optical layout in the ringdown breath analyzer. Figure 2. Schematic diagram of the optical layout in the ringdown breath analyzer. A ringdown waveform was digitalized to 10,000 data points and transferred to the in-house electronicA ringdown portion waveformof the instrument. was digitalized The in-house to 10,000 developed data points ringdown and softwaretransferred was to used the in-houseto obtain electronica ringdown portion time by of fittingthe instrument. the ringdown The in-housedata points developed to a single ringdown exponential software decay. was used to obtain a ringdown time by fitting the ringdown data points to a single exponential decay.

Sensors 2016, 16, 1199 4 of 15

A ringdown waveform was digitalized to 10,000 data points and transferred to the in-house electronic portion of the instrument. The in-house developed ringdown software was used to obtain a ringdown time by fitting the ringdown data points to a single exponential decay.

2.2. Measuring Method The background subtraction method, described in our previous works [39–43], was applied to obtain breath acetone concentrations. In this method, the optical cavity loss caused by the laboratory atmosphere at 1 atm is defined as the effective absorbance of the atmosphere, as described by the equation: d 1 1 A “ σ nd “ ´ (1) atm 266 c τ τ ˆ atm 0 ˙ ´20 2 where σ266 is the absorption cross-section of acetone at 266 nm, i.e., 4.5 ˆ 10 cm /molecules at atmospheric pressure and room temperature, n is the acetone concentration, d is the distance between the two mirrors, c is the speed of light in a vacuum, τatm and τ0 are the ringdown times when the gas cell is filled with laboratory air to 1 atm and under vacuum, respectively. Likewise, the absorbance of breath acetone is obtained by

d 1 1 A “ σ nd “ ´ (2) breath 266 c τ τ ˆ breath 0 ˙ where Abreath is the absorbance of the breath gas, τbreath is the ringdown time detected when the gas cell is filled with breath gas. In this study, by using the laboratory atmosphere as the background, the absolute concentration of breath acetone (the upper limit) in the breath gas is obtained by

∆A “ Abreath ´ Aatm “ σ pνq nd (3) where ∆A is the absorbance difference. It should be noted that the breath acetone concentrations obtained by the background subtraction method are the upper limits of acetone concentration in the breath samples. The rationale of the background subtraction method includes an assumption that the absorbance difference is attributed to the absorption of acetone alone. The reliability of this assumption has been rigorously evaluated in a previous study by investigating the possible absorbance from other atmospheric molecules and high abundance breath volatile organic compounds (VOCs) including several atmospheric molecules (H2O, N2,O2, and CO2), high concentration breath gases from 0 to 10 ppm (NH3, CH4, NO, N2O and CO), and high abundance breath VOCs (C5H8, CH3OH, C2H5OH, C3H8O, C2H6 and C5H12). The contribution of the highest abundance VOC, isoprene (C5H8), to the absorption at 266 nm is about 200 times smaller than that of acetone in normal human breath [37,38,42]. Other breath VOCs mentioned above have no or negligible absorption compared with acetone at 266 nm.

2.3. Breath Analysis Volunteers and Patients In this study, all healthy participants were volunteers. T1D and T2D subjects were hospitalized patients in the Metabolic Diseases Hospital, Tianjin Medical University, China. The research procedures and activities followed the research protocols approved by the Institutional Review Board (IRB approval number: IRB2013-053-01) of human subject research in Tianjin. All volunteers and patients agreed to perform the experiments (signed an informed consent form). The more detailed information including process was shown in the supplementary material [41]. A subject took a deep breath and breathed a single breath via a disposable mouthpiece into a fluorinated ethylene propylene (FEP) breath-gas-collection bag (~1 L), which was cleaned using high-purity nitrogen (>99.99%) prior to use. Simultaneous BG levels were measured using a standard diabetic-management blood glucometer (Roche, Accu-Chek Performa, Basel, Switzerland). The breath Sensors 2016, 16, 1199 5 of 15 sample collected in the FEP breath bag was introduced into the gas cell through a section of quarter-inch tubing connected to the gas inlet of the analyzer. It was tested that the FEP bag can keep breath gas fresh for up to 6 h. No additional procedure was used to handle excessive moisture in the exhaled Sensorsbreath 2016 since, 16, water1199 molecules have no absorption at 266 nm. 5 of 15

3.3. Results Results and and Discussion Discussion

3.1.3.1. Performance Performance of of the the Instrument Instrument AfterAfter the the breath breath analyzer analyzer was was constructed, constructed, tests tests were were conducted conducted to to check check its its properties properties including including stability,stability, reproducibility, reproducibility, response response time, time, and and the the limit limit of detection. of detection. The ringdown The ringdown baseline baseline scans were scans obtainedwere obtained in order in to order test tothe test instrument’s the instrument’s stability. stability. The instrument The instrument stability stability is defined is definedas σ/ τ where as σ/τ σwhere and στand areτ theare thestandard standard deviation deviation and and average average of ofringdown ringdown time, time, respectively. respectively. The The internal internal pressurespressures of of the the gas gas cell cell were were set set at at low low and and high high levels, levels, i.e., i.e., 0 0 Torr Torr and and 730 730 Torr, Torr, by by introducing introducing differentdifferent quantities quantities of of laboratory laboratory air air into into the the gas gas ce cell.ll. As As shown shown in inFigure Figure 3, 3a, typical a typical baseline baseline scan scan of theof theLaserBreath-001 LaserBreath-001 was was obtained obtained with with each each data data point point to be to generated be generated from from an average an average over over100 ringdown100 ringdown events. events. The Thetypical typical baseline baseline stability stability is better is better than than 0.5%. 0.5%. The The mirror mirror reflectivity reflectivity was was characterizedcharacterized toto bebe 99.9% 99.9% by by measuring measuring ringdown ringdown time time in an in empty an empty cell. Figure cell.4 Figureshows the4 shows analyzer’s the analyzer’sresponse to response laboratory to airlaboratory at 730 Torr. air The at 730 sharp Torr. rising The and sharp falling rising edges and of falling the signals edges indicate of the the signals rapid indicateresponse the of therapid analyzer. response This of testthe demonstratesanalyzer. This the test good demonstrates reproducibility the good and fast reproducibility response of the and breath fast responseacetone analyzer.of the breath The decrease acetone in analyzer. ringdown The time de whencrease nitrogen in ringdown and laboratory time when air fills nitrogen a vacuumed and laboratorycavity is due air to fills the a Rayleigh vacuumed scattering cavity atis 266due nm, to the this Rayleigh Rayleigh scattering lossat 266 is self-correctednm, this Rayleigh using scatteringthe background loss is subtractionself-corrected method using [the37]. bac Givenkground the known subtraction absorption method cross-section [37]. Given of the acetone known at ´20 2 absorptionatmospheric cross-section pressure, σ of266nm acetone= 4.5 atˆ atmospheric10 cm /molecule, pressure, σ using266nm =the 4.5 ringdown× 10−20 cm2 baseline/molecule, stability, using the0.5%, ringdown and the baseline mirror reflectivity,stability, 0.5%, 99.99%, and the the mirror theoretical reflectivity, detection 99.99%, limit the ofthe theoretical analyzer detection for acetone limit is ofdetermined the analyzer to for be 30acetone ppb (parts is determined per billion) to be based 30 ppb on the(parts 3 ´ perσ criterion.billion) based on the 3 − σ criterion.

Figure 3. The typical instrument stability in terms of the ringdown time obtained. Each data point is Figure 3. The typical instrument stability in terms of the ringdown time obtained. Each data point is generatedgenerated from from averaging averaging 100 100 ringdown ringdown events, events, a a stability stability σσ//ττ of of0.17% 0.17%was wasmeasured. measured.

It should be noted that the breath acetone analyzer gave consistent results in the vacuum It should be noted that the breath acetone analyzer gave consistent results in the vacuum conditions, measurements of nitrogen, and the laboratory air. However, the successive conditions, measurements of nitrogen, and the laboratory air. However, the successive measurements measurements of a breath samples showed a consistent decrease in acetone concentrations (increase of a breath samples showed a consistent decrease in acetone concentrations (increase in obtained in obtained ringdown times) (in Figure 4). The instability came primarily from the humidity variation ringdown times) (in Figure4). The instability came primarily from the humidity variation of the breath of the breath sample. Due to the humidity loss of a breath sample after it was collected and stored in sample. Due to the humidity loss of a breath sample after it was collected and stored in a breath bag, a breath bag, the obtained ringdown times increased in the three successive measurements, and this the obtained ringdown times increased in the three successive measurements, and this phenomenon phenomenon was more obvious when the breath sample was tested right after collection, as can be was more obvious when the breath sample was tested right after collection, as can be seen in Figure5. seen in Figure 5. Three bags of breath sample were collected from the same subject and tested Three bags of breath sample were collected from the same subject and tested immediately, which immediately, which were marked as breath-1, breath-2, and breath-3 in Figure 5. The first measured were marked as breath-1, breath-2, and breath-3 in Figure5. The first measured results of these three results of these three samples were all dramatically higher (lower in terms of ringdown times) than those that followed, and then reached stable values gradually. In order to confirm the effect of humidity on measured results, cylinder nitrogen, cylinder acetone samples (acetone helium mixture), and atmospheric air were tested. Cylinder nitrogen, laboratory air, and hallway air were tested using the instrument directly, while the cylinder acetone samples were transferred into three bags and tested immediately. The obtained results show that measured ringdown times of nitrogen and acetone samples were consistent, respectively. However, the ringdown times of laboratory air were

Sensors 2016, 16, 1199 6 of 15 samples were all dramatically higher (lower in terms of ringdown times) than those that followed, and then reached stable values gradually. In order to confirm the effect of humidity on measured results, cylinder nitrogen, cylinder acetone samples (acetone helium mixture), and atmospheric air were tested. Cylinder nitrogen, laboratory air, and hallway air were tested using the instrument directly, while the cylinder acetone samples were transferred into three bags and tested immediately. The obtained results showSensors that 2016 measured, 16, 1199 ringdown times of nitrogen and acetone samples were consistent, respectively.6 of 15 However,Sensors 2016 the, 16 ringdown, 1199 times of laboratory air were longer than those of hallway air because6 of 15 the longer than those of hallway air because the laboratory humidity is lower due to the air cooling and laboratory humidity is lower due to the air cooling and the dehumidification system. Further note that thelonger dehumidification than those of hallway system. airFurther because note the that labora the toryeffect humidity of humidity is lower on the du acetonee to the airmeasurements cooling and the effect of humidity on the acetone measurements is not due to absorption of water, instead it is due isthe not dehumidification due to absorption system. of water, Further instead note it isthat due the partially effect of to humidity the influence on the of theacetone breath measurements moisture on partiallytheis not reflectivity to due the to influence absorption of the ringdown of of the water, breath mirrors. instead moisture it is due on partially the reflectivity to the influence of the ringdownof the breath mirrors. moisture on the reflectivity of the ringdown mirrors.

Figure 4. The instrument shows good reproducibility in terms of the ringdown times obtained at Figure 4. The instrument shows good reproducibility in terms of the ringdown times obtained at differentFigure 4. pressures: The instrument 0 Torr, shows730 Torr good nitrogen, reproducibility and the laboratory in terms atmosphere.of the ringdown times obtained at different pressures: 0 Torr, 730 Torr nitrogen, and the laboratory atmosphere. different pressures: 0 Torr, 730 Torr nitrogen, and the laboratory atmosphere.

1.2 1.2 s)  s)  1.1 N N 1.1 2 2 N N2 2

lab air lab air

Ringdown time ( cylinder 1.0 cylinder cylinder acetone-2 acetone-3 breath-3 Ringdown time ( cylinder breath-2 1.0 acetone-1cylinder cylinder acetone-3 hallway air breath-3 acetone-1 acetone-2 breath-2 hallway air breath-1 0.9 breath-1 0.9 Test sequence Test sequence Figure 5. The instrument response to various samples, such as cylinder nitrogen, cylinder acetone heliumFigure mixture,5. The instrument laboratory response air, hallway to various air, and samples, breath samples. such as cylinder nitrogen, cylinder acetone Figure 5. The instrument response to various samples, such as cylinder nitrogen, cylinder acetone helium mixture, laboratory air, hallway air, and breath samples. heliumFor mixture,the purpose laboratory of reducing air, hallway high humidity air, and in breath the breath samples. samples, a membrane filter with a pore size ofFor 0.22 the µmpurpose was connectedof reducing to high the humidityinstrument in theinlet breath port. samples,Figure 6 ashows membrane its humidity filter with removal a pore effectiveness.sizeFor theof 0.22 purpose µm Four was ofbreath reducingconnected samples highto thewere humidity instrument collected in frominlet thebreath port.the same Figure samples, subject 6 shows aand membrane itstested humidity right filter afterremoval with the a pore sizebreath ofeffectiveness. 0.22 gasµm collection. was Four connected breath The samplessame to humidity the were instrument collected effect was inletfrom observed the port. same Figurewhen subject a6 breathshows and tested sample its humidity right (marked after removal theas breath-1) was tested without using a membrane filter, but the results of another three samples effectiveness.breath gas Fourcollection. breath The samples same humidity were collected effect was from observed the same when subject a breath and sample tested (marked right after as the (marked as breath-2, breath-3, and breath-4) were different from the breath-1 sample yet consistent breathbreath-1) gas collection. was tested The without same using humidity a membrane effect wasfilter, observed but the results when of a breathanother sample three samples (marked as among(marked the as three breath-2, since breath-3, a membrane and filterbreath-4) was used.were different from the breath-1 sample yet consistent breath-1) was tested without using a membrane filter, but the results of another three samples (marked among the three since a membrane filter was used. as breath-2, breath-3, and breath-4) were different from the breath-1 sample yet consistent among the three since a membrane filter was used.

Sensors 2016, 16, 1199 7 of 15 Sensors 2016, 16, 1199 7 of 15

Sensors 2016, 16, 1199 7 of 15 2.4 2.4 2.2 s)  2.2 s)

 N2 2.0 N2 lab air lab air

lab air N2 2.0 N2 breath-2 lab air lab air 1.8 breath-3 lab air with filter breath-4 with filter

Ringdown time ( time Ringdown breath-2 with filter 1.8 breath-3 breath-1 with filter breath-4 with filter

Ringdown time ( time Ringdown 1.6 without filter with filter breath-1 1.6 without filter Test sequence Figure 6. The instrument response to various Testsamples sequence and the effectiveness of humidity removal Figure 6. The instrument response to various samples and the effectiveness of humidity removal using using a membrane filter. a membraneFigure 6. filter.The instrument response to various samples and the effectiveness of humidity removal 3.2. Comparisonusing a membrane of Acetone filter. Values from CRDS to GC-MS Values 3.2. Comparison of Acetone Values from CRDS to GC-MS Values 3.2. ComparisonIn order to of investigateAcetone Values the fromaccuracy CRDS of to the GC-MS analyzer, Values six breath samples from four individual healthyIn order subjects to investigate and five thebreath accuracy samples of from the analyzer,three T2D six subjects breath were samples tested from using four both individual the In order to investigate the accuracy of the analyzer, six breath samples from four individual healthyringdown subjects breath and acetone five breath analyzer samples and a from certified three GC T2D-MS subjects facility. In were this testedexperiment, using breath both thesamples ringdown healthy subjects and five breath samples from three T2D subjects were tested using both the breathincluding acetone fasting analyzer (14-h and overnight a certified fast), GC-MSafter breakfas facility.t, after In thislunch, experiment, and after dinner breath from samples the healthy including ringdownand the T2D breath patients acetone were analyzer taken. The and GC-MS a certified analys GCis-MS was facility. conducted In this in the experiment, State Key Laboratorybreath samples on fasting (14-h overnight fast), after breakfast, after lunch, and after dinner from the healthy and the includingOdor Pollution fasting Control, (14-h overnight Tianjin fast),Academy after ofbreakfas Environmentalt, after lunch, Science and (TAES),after dinner which from is athe certified healthy T2Dandfacility patients the T2Dfor were trace patientstaken. chemical were The taken.species GC-MS The analysis. GC-MS analysis As analys wasthe Compendiumis conducted was conducted in Method the in State the TO-15, State Key Key a Laboratory cryogenic Laboratory pre- on on Odor PollutionOdorconcentrator Pollution Control, (ENTECH TianjinControl, Academy Instruments,Tianjin Academy of Inc., Environmental 7100A, of Envi Simironmental ScienceValley, CA,Science (TAES), USA) (TAES), which was employed iswhich a certified is to a removecertified facility for tracefacilityN chemical2, CO for2, and trace species H2 O.chemical analysis. species As analysis. the Compendium As the Compendium Method TO-15, Method a cryogenicTO-15, a cryogenic pre-concentrator pre- (ENTECHconcentratorThe Instruments, VOCs-enriched (ENTECH Inc., Instruments, sample 7100A, was Simi introducedInc., Valley, 7100A, CA, into Simi USA) th eValley, GC-MS was employedCA, system USA) operated towas remove employed in the N 2selected, COto remove2, andion H 2O. NmonitoringThe2, CO VOCs-enriched2, and (SIM)H2O. mode sample with wasa higher introduced sensitivit intoy and the GC-MSbetter signal-noise system operated ratio. The in theoperating selected ion monitoringconditionsThe (SIM)VOCs-enriched and mode analytical with sample conditions a higher was sensitivity ofintroduced the GC-MS and into system better the GC-MS signal-noisecan be system seen elsewhere. ratio.operated The Itin operating took the selected about conditions one ion andmonitoring analyticalhour to test conditionsone(SIM) sample. mode ofConsidering with the GC-MSa higher high system sensitivittesting can costsy be and including seen better elsewhere. the signal-noise long testing It took ratio. time, about Thethe one measured operating hour to test oneconditions sample.acetone concentrationsConsidering and analytical using high conditions the testing methods of costs the (theGC-MS including analyzer system theand can long the be GC-MS) testingseen elsewhere. were time, only the It up measuredtook to about2.5 ppm acetoneone that covers the major range of acetone concentration in healthy and T2D subjects. concentrationshour to test one using sample. the methods Considering (the high analyzer testing and costs the including GC-MS) the were long only testing up totime, 2.5 the ppm measured that covers acetoneFigure concentrations 7 shows the using results the and methods they agree (the with analyzer each other, and thewith GC-MS) a linear werefitting only coefficient up to 2.5of 0.98. ppm the major range of acetone concentration in healthy and T2D subjects. thatThe covers slope (0.99)the major suggests range that of acetonethe obtained concentration acetone concentrations in healthy and via T2D both subjects. methods are consistent. Figure7 shows the results and they agree with each other, with a linear fitting coefficient of 0.98. ThisFigure GC-MS 7 showsvalidation the resultstest proves and they that agreethe ringdo with eachwn breath other, acetonewith a linear analyzer fitting is reliablecoefficient and of fast, 0.98. TheThe slopewhich slope (0.99)can (0.99) be suggestsused suggests for subsequent that that the the obtained obtainedmeasurements acetoneaceton ofe breathconcentrations concentrations samples. via via both both methods methods are areconsistent. consistent. ThisThis GC-MS GC-MS validation validation test test proves proves thatthat thethe ringdownringdown breath breath acetone acetone analyzer analyzer is reliable is reliable and andfast, fast, 3.0 whichwhich can can be usedbe used for for subsequent subsequent measurements measurements of breath samples. samples. Pearson's r=0.98 2.5 y=0.99x-0.16 3.0 P<0.05 2.0 Pearson's r=0.98 2.5 y=0.99x-0.16 1.5 P<0.05 2.0 1.0 1.5 0.5

Acetone concentration (GC-MS) (ppm) (GC-MS) concentration Acetone 1.0 0.0 0.5 0.51.01.52.02.53.0 Acetone concentration (CRDS) (ppm) Acetone concentration (GC-MS) (ppm) (GC-MS) concentration Acetone 0.0 Figure 7. The ringdown breath analyzer’s0.51.01.52.02.53.0 performance validated by a certified GC-MS facility. Acetone concentration (CRDS) (ppm)

Figure 7. The ringdown breath analyzer’s performance validated by a certified GC-MS facility. Figure 7. The ringdown breath analyzer’s performance validated by a certified GC-MS facility.

Sensors 2016, 16, 1199 8 of 15

3.3. Breath Tests in Human Subjects Using the Validated Breath Acetone Analyzer 52 non-diabetic healthy subjects, 22 T1D, and 312 T2D patients participated in this study over a period of two months. Table1 lists the detailed information on the 1257 breath samples collected from the 386 subjects, for example, in the 22 T1D subjects, the number of the subjects who provided 1, 2, 3, and 4 breath samples is 2, 2, 2, and 16, respectively. In total, 76 breath samples were collected from the T1D subjects. Table2 lists the statistical information and test results (n1, n2, and n3 designate T1D, T2D, and healthy subjects, respectively). It should be pointed out that the subtraction approach is feasible only if inspired/room air concentrations compared to exhaled concentrations are low (<5%). As a matter of fact, acetone concentrations in the atmosphere air ranged from 0 to 40 ppb in accordance with Toxicological Profile for acetone (U.S. Department of Health and Human Services, Agency for Toxic Substances and Disease Registry, 1994), which depends on the local air quality. The concentration is about 0.6%, 3%, and 2% of the mean breath acetone concentrations in T1D subjects (4.9 ˘ 16 ppm), T2D subjects (1.5 ˘ 1.3 ppm), and healthy subjects shown in Table2, correspondingly.

Table 1. Breath samples collected from the 386 human subjects.

Subject Number T1D (n = 22) T2D (n = 312) Healthy (n = 52) Sample Number Per Subject 1 2 57 9 2 2 4 4 3 2 56 5 4 16 195 34 Total number of samples 76 1013 168

In this study, up to a maximum of four types of breath samples from each individual subject (average 3.2 per subject) was collected and tested under four different conditions: fasting, 2 h-breakfast, 2 h post-lunch, and 2 h post-dinner. Figure8 shows the mean breath acetone concentrations for the non-diabetic healthy subjects and diabetic subjects (T1D and T2D) under both fasting and 2 h post-meals. The breath acetone concentrations of the 52 non-diabetic subjects ranged from 0.1 to 2.0 ppm and the global mean over the 168 samples collected from the 52 healthy subjects was 1.1 ˘ 0.5 ppm. As shown in Table2 and Figure8, the ranges of breath acetone concentrations for the non-diabetic healthy subjects under the four different conditions were 0.3–1.9, 0.1–2.0, 0.1–2.0, and 0.3–1.4 ppm, respectively. The mean breath acetone concentrations for the healthy subjects under the four different conditions were determined to be 1.3 ˘ 0.3, 0.9 ˘ 0.5, 1.0 ˘ 0.6, and 1.1 ˘ 0.4 ppm, respectively. In general, the mean breath acetone concentration in the healthy subjects studied in this work is consistent with the published results, which have a range from 0.39 to 1.03 ppm [6]. It is normal that breath acetone concentrations in a few samples were 2.0 ppm. In fact, breath acetone concentrations can be affected on hourly and daily time scales by some factors including diet. After initiating a calorie restriction diet, breath acetone concentrations rise for 3–8 days (in subjects who were losing fat) before reaching a new steady state. In this work, all healthy subjects were not under the condition of the same diet. Whether the cause of the few higher breath acetone concentrations is due to the different diets or not remains to be studied. Sensors 2016, 16, 1199 9 of 15

Table 2. Statistical information and test results of the 386 human subjects.

Factor T1D (n = 22) T2D (n = 312) Healthy (n = 52) p Value Age 28.1 ˘ 13.4 (13.0~61.0) 57.0 ˘ 12.9 (14.0~83.0) 27.0 ˘ 4.8 (20.0~48.0) <0.0001 Gender male 8 (34.8%) 178 (57.1%) 22 (42.3%) 0.931 female 14 (65.2%) 134 (42.9%) 30 (57.7%) Years (year) 5.5 ˘ 7.4 (0.1~23.0) 11.7 ˘ 8.0 (0.0~36.0) — — Weight (kg) 54.8 ˘ 6.3 (46.0~69.0) 72.8 ˘ 14.0 (44.0~117.0) 62.2 ˘ 11.4 (46.0~91.0) <0.0001 Height (cm) 163.1 ˘ 5.4 (156.0~171.0) 165.3 ˘ 8.6 (144.0~186.0) 167.5 ˘ 8.0 (153.0~183.0) 0.1 BMI (kg/m2) 20.6 ˘ 2.5 (18.4~27.2) 26.5 ˘ 3.8 (17.9~38.2) 22.1 ˘ 3.1 (17.1~33.9) <0.0001 Acetone (ppm) Fasting (n1 = 22, n2 = 312, n3 = 52) 6.9 ˘ 21.7 (0.7~103.7) 1.7 ˘ 0.7 (0.1~19.8) 1.3 ˘ 0.3 (0.3~1.9) 0.3 2h post-breakfast (n1 = 20, n2 = 255, n3 = 40) 6.3 ˘ 19.6 (0.7~89.5) 1.5 ˘ 1.1 (0.1~7.1) 0.9 ˘ 0.5 (0.1~2.0) <0.05 2h post-lunch (n1 = 16, n2 = 195, n3 = 41) 4.0 ˘ 10.3 (0.3~42.5) 1.4 ˘ 1.0 (0.1~7.2) 1.0 ˘ 0.6 (0.1~2.0) 0.4 2h post-dinner (n1 = 18, n2 = 251, n3 = 35) 1.5 ˘ 0.7 (0.6~2.9) 1.5 ˘ 1.3 (0.1~10.6) 1.1 ˘ 0.4 (0.3~1.4) 0.2 Total (n1 = 76, n2 = 1013, n3 = 168) 4.9 ˘ 16 (0.3~103.7) 1.5 ˘ 1.3 (0.1~19.8) 1.1 ˘ 0.5 (0.1~2.0) <0.01 BGL (mg/dL) Fasting (n1 = 22, n2 = 312, n3 = 38) 203.3 ˘ 53.9 (86.4~324.0) 152.1 ˘ 41.0 (63.0~304.2) 91.8 ˘ 7.3 (79.2~108.0) <0.0001 2h post-breakfast (n1 = 20, n2 = 255, n3 = 35) 252.9 ˘ 70.4 (124.2~433.8) 206.5 ˘ 59.8 (73.8~392.4) 96.8 ˘ 13.6 (48.6~124.2) <0.0001 2h post-lunch (n1 = 16, n2 = 195, n3 = 32) 207.9 ˘ 70.4 (106.2~324.0) 189.0 ˘ 57.2 (73.8~405.0) 105.3 ˘ 14.0 (79.2~131.4) <0.0001 2h post-dinner (n1 = 18, n2 = 251, n3 = 34) 235.6 ˘ 93.3 (93.6~394.2) 181.4 ˘ 52.3 (90.0~397.8) 118.3 ˘ 12.4 (100.8~138.6) <0.0001 Total (n1 = 76, n2 = 1013, n3 = 168) 224.6 ˘ 74.7 (86.4~433.8) 180.6 ˘ 56.2 (63.0~405.0) 100.0 ˘ 14.5 (48.6~138.6) <0.0001 NOTE Values presented as mean ˘ SD (min~max) with analysis of Kruskal-Wallistest or n (%) with Pearson chi-square test. Sensors 2016, 16, 1199 10 of 15 Sensors 2016, 16, 1199 10 of 15

30 6.9 T1D 6.3 T2D healthy 4.9 20

4.0

10

1.7 1.5 1.4 1.0 1.5 1.5 1.5 Mean acetone concentration (ppm) 1.3 0.9 1.1 1.1 0

fasting 2h-post 2h-post 2h-post total breakfast lunch dinner Figure 8. Mean breath acetone concentrations in 22 T1D subjects, 312 T2D subjects, and 52 non- Figure 8. Mean breath acetone concentrations in 22 T1D subjects, 312 T2D subjects, and 52 non-diabetic diabetic subjects, under four different conditions: fasting, 2 h post-breakfast, 2 h post-lunch, and 2 h subjects, under four different conditions: fasting, 2 h post-breakfast, 2 h post-lunch, and 2 h post-dinner. post-dinner. The error bar corresponds to one standard deviation. The error bar corresponds to one standard deviation.

The breath acetone concentrations of the 22 T1D and 312 T2D subjects involved in this work are shownThe in breathFigure acetone8, it ranged concentrations from 0.3 to 103.7 of the ppm, 22 T1D and and 0.1 to 312 19.8 T2D ppm, subjects respectively. involved The in thismean work values are areshown 4.9 ± in 16 Figure ppm 8(T1D),, it ranged and from1.5 ± 1.3 0.3 ppm to 103.7 (T2D), ppm, which and 0.1are to about 19.8 ppm,4.5 and respectively. 1.4 times of The the mean mean values value ˘ ˘ ofare acetone 4.9 16 concentration ppm (T1D), and in the 1.5 52 1.3non-diabetic ppm (T2D), he whichalthy human are about subjects, 4.5 and respectively. 1.4 times of theFour mean different value rangesof acetone of breath concentration acetone concentrations in the 52 non-diabetic for the T1 healthyD subjects human under subjects, the four respectively. different conditions Four different are 0.7–103.7,ranges of breath0.7–89.5, acetone 0.3–42.5, concentrations and 0.6–2.9 ppm, for the resp T1Dectively. subjects The under mean the breath four differentacetone concentrations conditions are under0.7–103.7, the four 0.7–89.5, different 0.3–42.5, conditions and 0.6–2.9 for the ppm, 22 T1D respectively. subjects are The 6.9 mean ± 21.7, breath 6.3 ± acetone19.6, 4.0 concentrations± 10.3, and 1.5 ±under 0.7 ppm. the fourSimilarly, different the conditionsfour ranges for of thebreath 22 T1D acetone subjects concentrations are 6.9 ˘ 21.7, for 6.3the ˘31219.6, T2D 4.0 are˘ 10.3,0.1–19.8, and 0.1–7.1,1.5 ˘ 0.7 0.1–7.2, ppm. Similarly,and 0.1–10.6 the fourppm, ranges respectively; of breath and acetone the four concentrations mean breath for acetone the 312 concentrations T2D are 0.1–19.8, for the0.1–7.1, T2D 0.1–7.2,are 1.7 ± and 0.7, 0.1–10.6 1.5 ± 1.1, ppm, 1.4 respectively;± 1.0, and 1.5 and ± 1.3 the ppm, four correspondingly. mean breath acetone For concentrationsthe T1D and T2D for subjectsthe T2D studied are 1.7 ˘in0.7, this 1.5 work,˘ 1.1, the 1.4 mean˘ 1.0, values and 1.5of all˘ 1.3diabetic ppm, subjects correspondingly. under fasting For or the 2 T1Dh post-meals and T2D aresubjects all higher studied than in that this in work, the non-diabetic the mean values heal ofthy all human diabetic subjects, subjects respectively. under fasting The or result 2 h post-meals suggests thatare allthere higher is an than elevation that in of the mean non-diabetic breath acetone healthy concentration human subjects, for diabetic respectively. subjects The no result matter suggests which ofthat the there four is conditions an elevation they of are mean under. breath acetone concentration for diabetic subjects no matter which of the fourFigure conditions 9a–c respectively they are under.show the Figure breath9a–c acetone respectively concentration show the versus breath the acetone concurrent concentration BG level ofversus the 22 the T1D concurrent subjects, BG312 levelT2D ofsubj theects, 22 T1Dand 52 subjects, non-diabetic 312 T2D healthy subjects, subjects and 52 involved non-diabetic in this healthy work. Nosubjects direct involved correlation in this between work. No the direct measured correlation individual between breath the measured acetone individualconcentrations breath and acetone the individualconcentrations BG levels and the exists individual in the 22 BG T1D levels subjects, exists 312 in theT2D 22 subjects, T1D subjects, and 52 312 non-diabetic T2D subjects, healthy and subjects52 non-diabetic studied in healthy this work, subjects as shown studied inin Figu thisre work,9a–c, correspondingly. as shown in Figure However,9a–c, correspondingly. for the 20 T1D subjects,However, all for T1D the subjects 20 T1D subjects,excluding all two T1D subjects subjects with excluding ketoacidosis, two subjects a linear with correlation ketoacidosis, between a linear the meancorrelation breath between acetone theconcentrations mean breath and acetone the mean concentrations BG levels was and observed. the mean A BG linear levels fitting was yielded observed. R =A 0.56 linear and fitting p < yielded0.05, as R shown = 0.56 andin Figurep < 0.05, 10. as The shown results in Figure are consistent10. The results with are the consistent results reported with the previouslyresults reported in the previously literatures in based the literatures on the group based methods on the group [17,35]. methods The cause [17,35 of]. Thesome cause individuals of some deviatingindividuals from deviating the general from trend the general needs trendto be further needs to investigated. be further investigated. In fact, neither a single study nor the collective results from over the last 50 years of study can give a conclusive answer to whether there is a quantitative relation between breath acetone and BG or not to date, even though grouped breath acetone concentrations based on the BG level and HbA1c showed some critical meaning in previous studies. The number of T1D or T2D subject accounts for only about 1% of the more than 3000 human subjects recruited collectively for quantitative breath acetone measurements based on the published literature [6]. Consequently, the number of diabetic subjects used in breath analysis in the literature is still obviously insufficient for the study of breath acetone concentrations in diabetic people and its correlation with blood glucose level. On top of that, there are huge variations among the results from different studies on account of no “gold-standard

Sensors 2016, 16, 1199 11 of 15 method” for breath acetone measurements due to technical problems of sampling and analysis. With the remarkable progress of analytical techniques and sampling methodology, a real-time on-line ringdown breath analyzer, the LaserBreath-001, a validated breath analyzer, was applied for measuring 1257 breath acetone samples from 52 non-diabetic healthy subjects and 334 diabetic patients (22 T1D and 312 T2D), which represented a relatively large number of diabetic subjects used ever in a single study. The results suggest that elevated mean breath acetone concentrations exist in diabetic subjects. In 20 T1D subjects with no ketoacidosis, a linear correlation between the mean individual breath acetone concentration and the mean individual BG levels is observed. No direct (without grouping and considering the group mean) correlation was found for between the breath acetone and the BG level for T1D subjects, T2D subjects, and healthy subjects in this work. As can be seen in Table 2, there are significant differences in the weight and body mass index (BMI) among the T1D group, the T2D group, and the non-diabetic healthy group. Both parameters were found to correlate to the breath acetone concentration in the previous studies. Comparatively, the relationship between a breath acetone level and a BG level and other bioinformatic parameters remains to be further investigated with a large number of subjects (e.g., 1000 s more). This type of breath analyzer, whichSensors 2016 will, 16 be, 1199 further developed into an ideal tool for breath acetone measurement, is the key11 of to 15 obtaining a large pool of clinical data on breath acetone.

Figure 9. The measuredmeasuredbreath breath acetone acetone concentrations concentrations versus versus BG BG levels, levels, (a) ( 22a) T1D22 T1D subjects; subjects; (b) 312 (b) T2D312 T2Dsubjects; subjects; and (andc) 52 (c healthy) 52 healthy subjects. subjects.

6

y=0.21x-0.72 R=0.56, P<0.05 4

2 Meanacetone concentration (ppm) Meanacetone concentration (ppm) 0 100 150 200 250 300 350 Mean BG level (mg/dL) Figure 10. Observed correlation of the mean breath acetone concentration (ppm) with the mean blood Figure 10. Observed correlation of the mean breath acetone concentration (ppm) with the mean blood glucose level in all T1D subjects with no ketoacidosis. The two samples with ketoacidosis are marked glucose level in all T1D subjects with no ketoacidosis. The two samples with ketoacidosis are marked in dashed circles. in dashed circles.

In fact, neither a single study nor the collective results from over the last 50 years of study can give a conclusive answer to whether there is a quantitative relation between breath acetone and BG or not to date, even though grouped breath acetone concentrations based on the BG level and HbA1c showed some critical meaning in previous studies. The number of T1D or T2D subject accounts for only about 1% of the more than 3000 human subjects recruited collectively for quantitative breath acetone measurements based on the published literature [6]. Consequently, the number of diabetic subjects used in breath analysis in the literature is still obviously insufficient for the study of breath acetone concentrations in diabetic people and its correlation with blood glucose level. On top of that, there are huge variations among the results from different studies on account of no “gold-standard method” for breath acetone measurements due to technical problems of sampling and analysis. With the remarkable progress of analytical techniques and sampling methodology, a real-time on-line ringdown breath analyzer, the LaserBreath-001, a validated breath analyzer, was applied for measuring 1257 breath acetone samples from 52 non-diabetic healthy subjects and 334 diabetic patients (22 T1D and 312 T2D), which represented a relatively large number of diabetic subjects used ever in a single study. The results suggest that elevated mean breath acetone concentrations exist in diabetic subjects. In 20 T1D subjects Sensors 2016, 16, 1199 12 of 15 with no ketoacidosis, a linear correlation between the mean individual breath acetone concentration and the mean individual BG levels is observed. No direct (without grouping and considering the group mean) correlation was found for between the breath acetone and the BG level for T1D subjects, T2D subjects, and healthy subjects in this work. As can be seen in Table2, there are significant differences in the weight and body mass index (BMI) among the T1D group, the T2D group, and the non-diabetic healthy group. Both parameters were found to correlate to the breath acetone concentration in the previous studies. Comparatively, the relationship between a breath acetone level and a BG level and other bioinformatic parameters remains to be further investigated with a large number of subjects (e.g., 1000 s more). This type of breath analyzer, which will be further developed into an ideal tool for breath acetone measurement, is the key to obtaining a large pool of clinical data on breath acetone.

3.4. Toward Diabetes Screening and Management Currently, diabetes screening still relies on the use of a pharmacy-based blood glucose meter (BGM). Readings from a commercial BGM have a large standard deviation and typically a BGM has an error bar up to 20% or larger when it is used for BG measurements, depending on how the testing steps are followed. Accordingly, the observed deviation in the BG levels of both healthy and diabetic subjects could be partially attributed to the measurement uncertainty of the used diabetic-management blood glucometer. Blood A1C test is therefore recommended as a further confirmation of the BGM results. In addition to the pain and inconvenience from the required blood sampling when using a BGM, collective costs of using a BGM cannot be underestimated, approximately one USD per measurement [44]. For diabetes management, daily frequent check of BG levels would cost much more. A low cost and noninvasive (no blood draw) diabetes screening and or diabetic status monitoring is highly desirable, yet such as device is still not available. Issues in using the breath acetone analyzer include: (1) Many physiological parameters and conditions affect breath acetone, such as fasting, excise, diets, exposure to high acetone industrial environments. This fact requires specific control when using the breath acetone analyzer to measure breath acetone concentration to ensure that is only from diabetic disease; (2) Although diabetic patients, particularly T1D, have elevated mean breath acetone concentration in general, an individual breath acetone measurement is not sufficient for diabetes screening because of the high false positive rate that is partially due to the situations in (1). Our statistic analysis from the data collected from 22 T1D patients under the condition of fasting shows that the screening accuracy based on an individual breath acetone measurement using the breath analyzer is about 74%. Nevertheless, this breath acetone analyzer still holds significant values in the following situations. (1) For T1D human subjects under a particular control, e.g., fasting, convenient and noninvasive breath acetone measurements can provide diabetes pre-screening with about 74% accuracy or better. This is also particularly valued for babies or children who have been diagnosed as T1D; breath acetone measurements as a complementary means can significantly reduce pain, cost, and inconvenience when using a BGM many times a day is required; (2) For diabetic patients with ketone bodies, breath acetone is a reliable indicator of the physiological symptoms and breath acetone level is strongly correlated to the blood β-hydroxybutyrade (BHB) level [41]; this breath analyzer can be a useful instrument for status monitoring; (3) For human subjects under intensive exercise, the analyzer can conveniently measure pre- and post-exercise breath acetone to evaluate the effect of exercise on burning fatty acids.

4. Summary A GC-MS validated breath acetone instrument based on the CRDS technique has been used for breath acetone measurements in 386 human subjects. More than 1200 breath samples collected from 334 diabetic patients (22 T1D subjects and 312 T2D subjects) in a clinic and 52 non-diabetic healthy subjects in the research lab under different situations (fasting or post-meals) were measured by the validated breath acetone analyzer, LaserBreath-001; and blood glucose levels were simultaneously tested. The mean value of breath acetone concentration for the diabetic subjects (T1D and T2D) under Sensors 2016, 16, 1199 13 of 15 fasting, 2 h post-breakfast, 2 h post-lunch, or 2 h post-dinner is all higher than that in the non-diabetic healthy subjects correspondingly and so is the total mean values. This result indicates that both T1D and T2D subjects have elevated mean breath concentrations. There was a linear correction (R = 0.56, p < 0.005) between the mean breath acetone concentration and the mean BG levels for 20 T1D subjects. No correlation was found between an individual breath acetone concentration and an individual BG level in T1D subjects, T2D subjects, and healthy subjects in this work. The quantitative information about breath acetone levels was obtained by using a relatively large number of diabetic subjects involved in this work, which benefited from using the ringdown breath acetone analyzer with advantages of fast response, on-line measurement, and relatively low operation cost over the conventional GC-MS method. These features enable further clinical testing that may require a larger number of diabetic patients under various situations. The clinical application of breath acetone in pre-screening diabetes depends on improvement of the positive rate, which is about 74% currently for T1D. For diabetes with ketone bodies, the breath acetone concentration is a reliable and complimentary means in monitoring the blood BHB level. In this update, some data and findings are from our previous studies, some are from the most recent and unreported progresses, such as the latest version of the breath acetone analyzer, LaserBreath-001. The performance of the instrument on the measurement reproducibility (>95%) was improved by using a moisture filter in the breath sample instruction inlet. Further development efforts will be focused on more clinical testing; developments of algorithms to reduce interferences from other physiological parameters and of optimal conditions for specifications of measured breath acetone from diabetes only; and reduction of the instrument size, weight, and cost toward a POC device that can be used as a noninvasive means complementary to BG measurements in diabetes screening and management.

Acknowledgments: This work was supported by National Science Foundation of China (Grant No. 81471701). The authors thank the participation of numerous volunteers who provided breath gas samples. Author Contributions: Yingxin Li and Chuji Wang supervised the research project and revised the manuscript. Chenyu Jiang, Meixiu Sun, Zhennan Wang Xiaomeng Zhao, and Yuan Yuan performed the experiments. Chenyu Jiang, Zhennan Wang and Zhuying Chen processed the data. Chenyu Jiang wrote the manuscript. Conflicts of Interest: The authors declare no conflicts of interest.

References

1. Jansson, B.O.; Larsson, B.T.J. Analysis of organic compounds in human breath by gas chromatography-mass spectrometry lab. Clin. Med. 1969, 74, 961–966. 2. Pauling, L.; Robinson, A.B.; Cary, P. Quantitative analysis of urine vapor and breath by gas-liquid partition chromatography. Proc. Natl. Acad. Sci. USA 1971, 68, 2374–2376. [CrossRef][PubMed] 3. Gordon, S.M.; Szidon, J.P.; Krotoszynski, B.K.; Gibbons, R.D.; O’Neill, H.J. Volatile organic compounds in exhaled air from patients with lung cancer. Clin. Chem. 1985, 31, 1278–1282. [PubMed] 4. Risby, T.H.; Solga, S.F. Current status of clinicalbreath analysis. Appl. Phys. B 2006, 85, 421–426. [CrossRef] 5. Mürtz, M. Breath diagnostics using laser spectroscopy. Opt. Photonics News 2005, 16, 30–35. [CrossRef] 6. Wang, Z.; Wang, C. Is breath acetone a biomarker of diabetes—A historical review on breath acetone measurements. J. Breath Res. 2013, 7.[CrossRef][PubMed] 7. Rooth, G.; Ostenson, S. Acetone in alveolar air, and the control of diabetes. Lancet 1966, 288, 1102–1105. [CrossRef] 8. Nelson, N.; Lagesson, V.; Nosratabadi, A.; Ludvigsson, J.; Tagesson, C. Exhaled isoprene and acetone in newborn infants and in children with diabetes mellitus. Pediatr. Res. 1998, 44, 363–367. [CrossRef][PubMed] 9. Diskin, A.M.; Španˇel,P.; Smith, D. Time variation of , acetone, isoprene and ethanol in breath: A quantitative SIFT-MS study over 30 days. Physiol. Meas. 2003, 24, 107–109. [CrossRef][PubMed] 10. Smith, D.; Spanˇel,P.; Fryer, A.A.; Hanna, F.; Ferns, G.A. Can volatile compounds in exhaled breathbe used to monitor control in diabetes mellitus? J. Breath Res. 2011, 5.[CrossRef][PubMed] 11. Turner, C.; Španˇel,P.; Smith, D. A longitudinal study of ammonia, acetone and propanol in the exhaled breath of 30 subjects using selected ion flow tube mass spectrometry, SIFT-MS. Physiol. Meas. 2006, 27, 321–337. [CrossRef][PubMed] Sensors 2016, 16, 1199 14 of 15

12. Teshima, N.; Li, J.; Toda, K.; Dasgupta, P.K. Determination of acetone in breath. Anal. Chim. Acta 2005, 535, 189–229. [CrossRef] 13. Anderson, J.C.; Lamm, W.J.; Hlastala, M.P. Measuring airway exchange of endogenous acetone using a single-exhalation breathing maneuver. J. Appl. Physiol. 2006, 100, 880–889. [CrossRef][PubMed] 14. Natale, C.D.; Paolesse, R.; Martinelli, E.; Capuano, R. Solid-state gas sensors for breath analysis: A review. Anal. Chim. Acta 2014, 824, 1–17. [CrossRef][PubMed] 15. Reyes-Reyes, A.; Horsten, R.C.; Urbach, H.P.; Bhattacharya, N. Study of the exhaled acetone in type 1 diabetes using quantum cascade laser spectroscopy. Anal. Chem. 2015, 87, 507–512. [CrossRef][PubMed] 16. Anderson, J.C. Measuing breath acetone for minitoring fat loss: Review. Obesity 2015, 23, 2327–2334. [CrossRef][PubMed] 17. Li, W.; Liu, Y.; Lu, X.; Huang, Y.; Liu, Y.; Cheng, S.; Duan, Y. A cross-sectional study of breath acetone based on diabetic metabolic disorders. J. Breath Res. 2015, 9.[CrossRef][PubMed] 18. Nording, M.L.; Yang, J.; Hegedus, C.M.; Bhushan, A.; Kenyon, N.J.; Davis, C.E.; Hammock, B.D. Endogenouslevels of five fatty acid metabolites in exhaled breath condensate to monitor asthma by high-performance liquid chromatography: Electrospray tandem mass spectrometry. IEEE Sens. J. 2010, 10, 123–130. [CrossRef][PubMed] 19. Ulanowska, A.; Trawinska, E.; Sawrycki, P.; Buszewski, B. Chemotherapy control by breath profile with application of SPME-GC/MS method. J. Sep. Sci. 2012, 35, 2908–2913. [CrossRef][PubMed] 20. Yuan, H.; Mester, Z.; Lord, H.; Pawliszyn, J. Automated in-tube solid phase micro extraction coupled with liquid chromatography-electrospray ionization mass spectrometry for the determination of selected benzodiazepines. J. Anal. Toxicol. 2000, 24, 718–725. [CrossRef][PubMed] 21. Smith, D.; Wang, T.; Španˇel,P. On-line, simultaneous quantification of ethanol, some metabolites and water vapour in breath following the ingestion of alcohol. Physiol. Meas. 2002, 23, 477–479. [CrossRef][PubMed] 22. Harrison, G.R.; Critchley, A.D.; Mayhew, C.A.; Thompson, J.M. Real-time breath monitoring of propane and its volatile metabolites during surgery using a novel mass spectrometric technique: A feasibility study. Br. J. Anaesth. 2003, 91, 797–799. [CrossRef][PubMed] 23. Moser, B.; Bodrogi, F.; Eibl, G.; Lechner, M.; Rieder, J.; Lirk, P. Mass spectrometric profile of exhaled breath-field study by PTR-MS. Respir. Physiol. Neurobiol. 2005, 145, 295–300. [CrossRef][PubMed] 24. Ruzsanyi, V.; Baumbach, J.; Litterst, P.; Westhoff, M.; Freitag, L. Detection of human metabolites using multi-capillary columns coupled to ion mobility spectrometers. J. Chromatogr. A 2005, 1084, 145–151. [CrossRef][PubMed] 25. Allafchian, A.R.; Majidian, Z.; Lelbeigi, V.; Tabrizchi, M. A novel method for the determination of three volatile organic compounds in exhaled breath by solid-phase microextraction-ion mobility spectrometry. Anal. Bioanal. Chem. 2016, 408, 839–847. [CrossRef][PubMed] 26. Smith, D.; Španˇel,P. Selected ion flow tube mass spectrometry (SIFT-MS) for on-line trace gas analysis. Mass Spectrom. Rev. 2005, 24, 661–700. [CrossRef][PubMed] 27. Lei, J.; Hou, C.; Huo, D.; Luo, X.; Bao, M.; Li, X.; Yang, M.; Fa, H. A novel device based on a fluorescent cross-responsive sensor array for detecting lung cancer related volatile organic compounds. Rev. Sci. Instrum. 2015, 86.[CrossRef][PubMed] 28. Tisch, U.; Haick, H. Chemical sensors for breath gas analysis: The latest developments at the breath analysis summit 2013. J. Breath Res. 2014, 8.[CrossRef][PubMed] 29. Wang, C.; Sahay, P. Breath analysis using laser spectroscopy techniques: Breath biomarkers, spectral fingerprints, and detection limits. Sensors 2009, 9, 8230–8262. [CrossRef][PubMed] 30. Wang, C. Biomedical applications of cavity ringdown spectroscopy: Present and perspective. In Proceedings of the 57th Southeast/61st Southwest Joint Regional Meeting of the American Chemical Society, Memphis, TN, USA, 1–4 November 2005. 31. O’keefe, A.; Deacon, D.A. Cavity ring-down optical spectrometer for absorption measurements using pulsed laser sources. Rev. Sci. Instrum. 1988, 59, 2544–2551. [CrossRef] 32. Berden, G.; Peeters, R.; Meijer, G. Cavity ring-down spectroscopy: Experimental schemes and applications. Int. Rev. Phys. Chem. 2000, 22, 565–607. [CrossRef] 33. Wang, C.; Srivastava, N.; Jones, B.A.; Reese, R.B. A novel multiple species ringdown spectrometer for in situ measurements of , , and carbon isotope. Appl. Phys. B 2008, 92, 259–270. [CrossRef] Sensors 2016, 16, 1199 15 of 15

34. Berden, G.; Engeln, R. Cavity Ring-Down Spectroscopy: Techniques and Applications; Wiley-Blackwell: Chichester, UK, 2009. 35. Wang, C.; Mbi, A.; Shepherd, M. A study on breath acetone in diabetic patients using a cavity ringdown breath analyzer: Exploring correlations of breath acetone with blood glucose and glycohemoglobin A1C. IEEE Sens. J. 2010, 10, 54–63. [CrossRef] 36. Ciaffoni, L.; Hancock, G.; Harrison, J.J.; van Helden, J.-P.H.; Langley, C.E.; Peverall, R.; Ritchie, G.A.D.; Wood, S. Demonstration of a mid-infrared cavity enhanced absorption spectrometer for breath acetone detection. Anal. Chem. 2013, 85, 846–850. [CrossRef][PubMed] 37. Wang, C.; Surampudi, A.B. An acetone breath analyzer using cavity ringdown spectroscopy: An initial test with human subjects under various situations. Meas. Sci. Technol. 2008, 22.[CrossRef] 38. Wang, C.; Mbi, A. A new acetone detection device using cavity ringdown spectroscopy at 266 nm: Evaluation of the instrument performance using acetone sample solutions. Meas. Sci. Technol. 2007, 18, 2731–2741. [CrossRef] 39. Sun, M.; Chen, Z.; Gong, Z.; Zhao, X.; Jiang, C.; Yuan, Y.; Wang, Z.; Li, Y.; Wang, C. Determination of breath acetone in 149 type 2 diabetic patients using a ringdown breath-acetone analyzer. Anal. Bioanal. Chem. 2015, 407, 1641–1650. [CrossRef][PubMed] 40. Wang, Z.; Wang, C.; Lathan, P. Breath acetone analysis of diabetic dogs using a cavity ringdown breath analyzer. IEEE Sens. J. 2013, 14, 1117–1123. [CrossRef] 41. Sun, M.; Zhao, X.; Yin, H.; Wang, Z.; Jiang, C.; Liu, W.; Chen, Z.; Yuan, Y.; Li, Y.; Wang, C. Study of breath acetone and its correlations with blood glucose and blood beta-hydroxybutyrate using an animal model with lab-developed type 1 diabetic rat. RSC Adv. 2015, 5, 71002–71010. [CrossRef] 42. Sahay, P.; Scherrer, S.; Wang, C. Measurements of the weak UV absorptions of isoprene and acetone at 261–275 nm using cavity ringdown spectroscopy for evaluation of a potential portable ringdown breath analyzer. Sensors 2013, 13, 8170–8187. [CrossRef][PubMed] 43. Gong, Z.; Sun, M.; Jiang, C.; Wang, Z.; Kang, M.; Li, Y.; Wang, C. A ringdown breath acetone analyzer: Performance and validation using gas chromatography-mass spectrometry. J. Anal. Bioanal. Tech. 2014, S7. [CrossRef] 44. Yeaw, J.; Lee, W.C.; Aagren, M.M.; Christensen, T. Cost of self-monitoring of blood glucose in the United States among patients on an regimen for diabetes. J. Manag. Care Pharm. 2012, 18, 21–32. [CrossRef] [PubMed]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).