HTST pasteurization. Is it time to raise statutory time / temperature conditions to destroy Mycobacterium avium subsp. paratuberculosis (MAP)?
Note this article has been updated to reflect the published consensus of researchers and clinicians at the conference on MAP in the US in 2017 that Mycobacterium avium subsp. paratuberculosis (MAP) is a human pathogen (Kuenstner et al, 2017).
The use of a high temperature short time heat treatment (HTST) of 72°C for 15 seconds to destroy pathogenic bacteria in milk, reduce the number of spoilage organisms and increase shelf life is well established (Cerf and Condron, 2006; Codex Alimentarius (2004); Juffs and Deeth, 2007).
The history of pasteurization (pasteurisation is also valid) is fascinating and is notable for its public health success and for the insights of many scientists and engineers. Prior to the introduction of pasteurization, consumption of raw cow milk was a major source of infection by bacteria causing tuberculosis. Pasteurization has eliminated heat-treated-milk as a source of infection. Regrettably raw milk and raw milk products remain a major source of new cases of bovine tuberculosis.
This article calculates the effect of HTST treatment on the number of log reductions of major milk pathogens and discusses the temperature milk should be pasteurized if Mycobacterium avium subsp. paratuberculosis (MAP) was designated as a human pathogen. The log reductions refer to log10 or decimal (10 fold) reductions in the concentration of viable bacteria. The article does not discuss the effects of heat on the functional properties or the nutritive quality of milk.
Current indicator pathogens
The initial work on milk pasteurization by scientists was on heat treatments to eliminate the pathogens Mycobacterium tuberculosis and M. bovis. It soon became apparent that there was another pathogen, Coxiella burnetii, which was even more resistant to heat. C. burnetii is the causal agent of Q-fever.
Currently C. burnetii is recognised as the most heat-resistant non-sporulating pathogen likely to be present in milk. C. burnetii was selected as the heat treatment indicator organism for pasteurization before the significance of Listeria monocytogenes and other pathogens was known. International regulatory agencies require that HTST pasteurization of milk is designed to ensure at least a 5 log reduction of M. tuberculosis, M. bovis and C. burnetii in whole milk (4% milkfat) e.g. Codex Alimentarius, (2004), PMO (2011). Note some strains of Listeria monocytogenes can have a higher thermal resistance than C. burnetii in milk.
The 15 seconds at 72°C requirement is the minimum heat treatment required and many companies in Britain and Ireland use 25 second holding times and higher temperatures e.g. 75°C for 25 seconds (O'Reilly et al., 2004) for pasteurization for added safety.
Emerging potential pathogens that are more heat resistant than current indicator pathogens
There is increasing concern over the presence of M. avium subsp. paratuberculosis (MAP) in milk since it has been linked to Crohn’s Disease (CD) and a range of other inflammatory diseases. MAP is the causal agent of Johne’s disease, a chronic, currently incurable although there are now experimental treatments, inflammatory bowel disease of cattle and other ruminants.
Because the symptoms of these human and animal diseases are similar and that MAP has been isolated from gut-lesions combined with the detection of a unique MAP-DNA insertion sequence (IS900) in intestinal tissue from many patients and the elimination of the disease in some patients by prolonged antibiotic treatment, a link between Crohn’s disease and MAP has been suggested. This has still to be conclusively proven but the evidence for this linkage is increasing. In a recent review Bach (2015) concluded that "although evidence shows a potential association of MAP as a causative agent of CD, it has not been elucidated yet whether MAP is the primary cause of the disease, or whether it exploits the mucosal inflammation already developed and just exacerbates the disease".
Following a conference at Temple University in the US (Kuenstner et al , 2017) the conference participants reached consensus on several issues relating to MAP and concluded that the accumulating information now strongly supports the theory that MAP is a zoonotic bacterium; can cause disease in humans. A majority of the conferees (72%) noted that MAP present in dairy products and meat causes disease in some humans and thus poses a public health threat. Note the evidence presented suggested that a significant number of CD cases were the result of infection by MAP.
MAP has been isolated from commercial calf replacers (CMRs) suggesting that feeding CMRs to calves may constitute a method of horizontal transfer of this pathogen. Even more concerning, is that significant concentrations of MAP have been isolated from commercial powdered infant milk formulae. The isolation of MAP from these commercial milk products suggests either a high thermal resistance of this organism and/or post process contamination.
If , probably when, MAP receives formal international recognition as a human pathogen it will become the most thermally resistant, vegetative (non-sporulating) pathogen in milk and it will be necessary to increase the statutory time-temperature requirements used in pasteurisation to reduce the numbers of this organism by 5 log cycles.
What is meant by a 5 log reduction?
One log reduction reduces initial numbers by 10. Assuming we have 100 infective doses of C. burnetii per ml of milk held at 72°C and that the D 72°C for this pathogen is 1.88 seconds then we can detail the effect of log reductions as follows.
- One log reduction will reduce the concentration of C. burnetii to 10 cells/ml. Holding time 1.88 seconds.
- Two log reductions gives 1 cell / ml (After 1.88 x 2 = 3.76 seconds)
- Three log reductions gives 0.1 cell / ml (After 1.88 x 3 = 3 5.64 seconds
- Four log reductions gives 0.01 cell / ml (After 1.88 x 4 = 7.54 seconds)
- Five log reductions gives 0.001 cell / ml (After 1.88 x 5 = 9.4 seconds)
This calculated decline in numbers is valid providing the fat content of the milk does not exceed 4% and that sugar and other solids have not been added. In milk containing more than 4% fat and/or added sucrose and other solids the temperature and or holding time of pasteurization would need to be increased to obtain a satisfactory kill of C. burnetii.
As discussed earlier, internationally-recognised standards for HTST milk pasteurization require milk to be treated at 72°C for 15 seconds in such a way that all particles of milk are subjected to this time / temperature combination.
Note heat treatments giving five or more log reductions do not fully eliminate pathogens from milk; they reduce the number of viable cells to virtually undetectable levels (e.g. 1 / 1000 litres) and below the infective dose! This is somewhat similar to the concept of "commercial sterility" familiar in canning except that the number of pathogen-log reductions is much lower. Because the number of viable cells present after pasteurization also depends on their initial concentration it is obviously important that raw milk should contain only low concentrations of pathogens. Hence the importance of high animal health, hygienic production of milk and processes such as bactofugation and microfiltration for lowering microbial numbers.
Importance of turbulent flow
In order for all particles of milk to be held for at least 15 seconds at 72°C the flow of milk through the pasteurizer must be adjusted so that all particles receive the minimum required heat treatment. This is also reflected in relevant heat treatment legislation. This is easier achieved if the flow velocity is high enough to give turbulent flow as opposed to laminar flow. In laminar flow, the velocity of the fastest particle is typically twice that of the average particle potentially resulting in some particles getting about half the required holding time if the holding tube has not been 'sized' correctly!
The type of flow depends on the geometry of the holding-tube, flow rate of the fluid, and the viscosity and density of the fluid. Reynold’s number, NRe, is frequently is used to distinguish the type of flow. It is accepted that the flow pattern is laminar for NRe, < 2,100. Turbulent flow exists when NRe, exceeds 4,000. Note 4,000 is not a particularly high value and higher values are recommended (Mullan, 2018).
Reynolds Number can be calculated using equation 1.
Equation 1. NRe = ρ x v x D
ρ = density, kg/m3
v = velocity of flow, m/s
D = diameter, m
µ= viscosity of product at the heat treatment temperature, Pa s.
Even under high turbulent flow conditions there will be a range of particle residence times. Commercial plants have a responsibility to periodically undertake flow velocity tests on pasteurizers to verify that the pasteurizer meets legislative and food safety standards. To allow for the range of particle velocities the length of the holding tube, and hence holding time, should be adjusted to ensure that the fastest travelling particles are held for the correct holding period.
Note that the minimum holding time is normally determined using an electrical conductivity test and salt solutions (e.g. Lang and Jordan, 1958). While this works well for salt solutions and milk the results obtained are not readily transferable to more concentrated, viscous, milk products e.g. ice cream mixes. The holding time determined by mathematically calculating the residence time in the holding tube using the tube dimensions and dividing by the flow rate, yields the average flow rate. While this is usually close to minimum holding time under high turbulent flow conditions Re > 11,000 it should be divided by 2 under laminar flow to ensure that the fastest particles have received the minimum holding time.
Mathematical basis for log reductions and determination of process time
The principles of the calculations used in the following sections have been explained, with worked examples, in the Ebook, "Thermal processing of acid fruit and vegetable products. Significant microorganisms, recommended processing time / temperatures, and public health significance of spoilage" by Mullan (2012). The article "Modelling the destruction of Mycobacterium avium subsp. paratuberculosis" also provides an explanation of the mathematics involved and access to free On Line calculators.
Theoretical determination of the number of log reductions of current pathogens following HTST pasteurization of milk at 72°C for 15 seconds
We will now calculate the number of log cycles that populations of C. burnetii and L. monocytogenes should be reduced following HTST pasteurization. The mathematics have been explained previously. To do this we need to obtain the decimal reduction values (D values) of the most heat resistant strains in milk at 72°C. The D-value is the time taken, at a given temperature, to achieve a 10-fold (or a 1 log) decrease in the number of viable organisms. D values are widely available in the scientific literature. The FDA (FDA/CFSAN, 2000) and the ICMS (ICMSF. 1996) have published thermal resistance data for many organisms. More recently, Juffs and Deeth (2007) have compiled a series of D and Z values for major milk borne pathogens as part of a review of the effectiveness of milk pasteurization. The Lemgo D- and z-value database for food (Schwarzer et al., 2007) is also an excellent resource. Note D and Z values show significant variation and caution in selecting values for thermal resistance modelling is required. D-values of C. burnetii and Listeria monocytogenes are shown in table 1.
|Table 1. Selected decimal reduction times of major pathogens found in milk.|
|Coxiella burnetii||1.88||Cerf and Condron (2006)|
|Listeria monocytogenes||2.9||Mackey and Bratchell (1989)|
To calculate the number of log reductions of C. burnetii that HTST pasteurization should give we divide the number of seconds milk is held at 72°C by the D72°C of the organism. Note the D72°C value must be in the same time units as holding time at 72°C.
This gives 15 = 7.98 or almost 8 log reductions.
Clearly this is significantly more than the minimum 5 log reductions required to meet the requirements for pasteurization. In reality the number of log reductions would be expected to be higher since we have not accounted for the killing effects of the “come up” time to 72°C and the killing that would be expected to also take place during the initial part of cooling. Note. The lethality during heating and cooling, which is rarely calculated, can be significant and the limited data available suggests that it might constitute in the region of 25% of the overall process-lethality (Gut et al., 2004).
The infective dose of C. burnetii for healthy non-immuno-compromised people is generally expected to be around 1. So unless the number of infective cells was very high, detection of viable cells of C. burnetii would not be expected in HTST-pasteurized milk.
The number of log reductions of L. monocytogenes is calculated in a similar way.
=15 = 5.2
The infective dose of most strains of L. monocytogenes for people who are not immuno-compromised, or very young or old, is normally considered to be high, ranging from a few hundred to perhaps a million or more per gram. However, there is an argument for treating L. monocytogenes in a similar way to salmonella and requiring foods to contain <1 viable cell per 25 grammes. As discussed earlier the actual log reduction would be expected to be higher indicating that HTST pasteurization would be predicted to reduce the probability of survival of this pathogen to almost insignificant levels providing high quality milk was used.
Theoretical calculation of the holding time required for pasteurization at 72°C
Let’s assume that you were designing a HTST pasteurization process for milk for the first time. How might you design the process given that you were using 72°C as the holding temperature?
Knowing the requirement for a minimum of a 5 fold log reduction of all pathogens, the minimum process time is determined by multiplying the number of decimal reductions required by the D72°C value of the most heat resistant pathogen. Analysis of the data in Table 1 indicates that L. monocytogenes is more heat resistant than C. burnetii so using L. monocytogenes as the indicator, then the minimum holding time would be 5 x 2.99 = 14.5 seconds.
Theoretical determination of the number of log reductions of the potential pathogen Mycobacterium avium subsp. paratuberculosis following HTST pasteurization of milk at 72°C for 15 seconds
Unlike the situation with Clostridium botulinum and F0 calculations, there is no one scientifically-validated reference D72°C or Z-value value for MAP for use in modelling the destruction of MAP during milk pasteurization. Following a review by the UK Food Standards Agency, values of D 72°C for MAP of 12 to 14 seconds have been widely cited within the industry (Food Standards Agency, 2002). However, even the 14-second value is low compared with data for a strain, ATCC 19698, which had a D 71°C of 16.5 seconds (Sung and Collins, 1998). Other workers have also reported high D values e.g. Keswani and Frank (1998). Taking 15 seconds as a conservative estimate of the D 72°C for MAP, then the holding time at 72°C to obtain a 5-log reduction would be 5 x 15 = 75 seconds. This is unrealistic.
With a D 72°C of 15 seconds, no reduction in viable numbers would be expected following conventional pasteurization for 15 seconds and there would be a minimal reduction after 25 seconds:
25 = 1.7, so less than 2 log reductions.
Note many milk processors in Ireland and the UK are voluntarily using significantly higher time / temperature combinations to combat the potential threat from MAP.
We will next calculate the number of MAP predicted in pasteurized milk following heat treatment. There is no consensus concerning the concentration of MAP in raw milk. Lynch et al. (2007) reported < 1 CFU/mL in milk from clinically infected herds. Grant et al. (1999) have indicated that the concentration of MAP in raw milk could be as high as 104 CFU/mL due to faecal contamination. Sung and Collins (1998) have suggested that a MAP concentration of 106 CFU/mL should be used when modelling MAP destruction for safety reasons.
Lets assume a concentration of MAP of 104 CFU/mL in raw milk. We will calculate the concentration of viable MAP following pasteurization at 72°C for 25 seconds. To do this we will assume a D72°C of 15 seconds.
The number of microbial survivors, N2, following the heat treatment of a microbial population, N1, at a defined temperature for a particular time, T, can be calculated using equation 2 providing the decimal reduction time of the organism, D, is known. While D is usually expressed in minutes or seconds it must be expressed in the same time units as T.
Equation 2. T = D (log N1- log N2); Stumbo (1973).
Equation 2, can be rewritten to give N2, the number of survivors, as shown in equation 3.
Equation 3. log N2=logN1 - t/D
N2 is obtained by finding the antilog of log N2. This calculation can be performed using the free calculator at https://www.dairyscience.info/map/test-d.asp .
This calculation indicates that 2.15 × 102 CFU/mL of MAP from a population of 104 CFU/mL would survive a heat treatment of 25 seconds at 72 °C.
If the concentration of MAP in the raw milk was 1 CFU / 10 ml then the small reduction in viable numbers following heat treatment would probably be sufficiently low to render MAP non-detectable for all but specialist laboratories since MAP is demanding to culture.
Theoretical calculation of the holding time required for a 5 log kill for Mycobacterium avium subsp. paratuberculosis
Clearly obtaining a 5 log kill of MAP at 72°C using holding times of 15 or 25 seconds is impossible if a D72°C value >14 seconds is used. To obtain more realistic reductions in numbers we need to use higher holding temperatures and/or extended holding times.
Using the free On Line tools provided by the author the D78°C value of MAP can be calculated (assuming a Z value of 8.246°C and a D72°C of 14 seconds) as 2.60 seconds. Note as discussed earlier higher values for D and Z for MAP could have been used would give higher predicted D values. Using this D78°C value the holding time at 78°C to obtain a 5-log reduction would be 5 x 2.6 = 13 seconds.
The predicted number of log reductions for heat treatment at 78°C can now be calculated:
=15 = 5.8 log reductions
If we were to add the thermal destruction predicted during "come up" and "come down" during heat treatment a log reduction factor of around 6 would be expected.
Using a holding time of 25 seconds would give further assurance resulting in additional killing:
= 25 = 9.6 log reductions
This would suggest that if the pasteurization temperature for HTST milk was increased by an additional 6°C, to 78°C, this would provide a 5-10 log reduction for MAP depending on holding time. However, work by Grant et al. (2005) found that pasteurization at 78.5°C for either 15 or 25 seconds resulted in only a 4.0 to 5.2 log reduction of MAP. This is significantly lower than would be expected.
There are several reasons why this discrepancy may have occurred. Estimates of D values at higher temperatures may be subject to error (Mullan, 2012) and the uncertainty concerning a realistic D72°C (value too low) value for MAP may also have had an effect. The published data for the Z-value of MAP show some variation and the use of higher or lower values would have given different estimates of D at higher temperatures. MAP strains have a tendency to clump and when clumped have a higher thermal resistance e.g. Keswani and Frank (1998). It is possible that clumping of MAP during pasteurization contributed to the reduced log reductions following pasteurization.
Nevertheless the demonstration of a 5 log reduction for MAP after 25 seconds at 78.5°C by Grant et al. (2005) equates with the general requirement for a pasteurization-type heat treatment to reduce the numbers of pathogens by 5 log cycles. However, this was achieved using a temperature and a holding time markedly above the currently accepted values. Hence this author believes that a case exists for a new minimum time temperature combination for the HTST treatment of milk.
We will now consider the effects of raising the temperature to 80°C and using the most thermally resistant strain of MAP reported to date.
Taking a D 71°C of 16.5 seconds as an alternative value for MAP, a D80°C of 1.34 can be calculated. Using this value the minimum holding time at 80°C to obtain a 5-log reduction would be 5 x 1.34 = 6.7 seconds. Keeping the holding time at 15 seconds, this additional temperature rise should give 15/1.34 = 11.2 log reductions, a significant improvement in the thermal destruction of MAP over 72°C for 15 seconds.
While there is occassional comment about the importance of holding time versus temperature on microbial destruction, increasing temperature particularly in 10° C increments has a major effect on microbial destruction.
An even higher safety margin would be obtained by using time / temperature heat treatments intermediate between HTST and ultra high heat treatment, typically 120°C to 135°C for 1 - 4 seconds. These conditions are used in extended shelf life products designated ESL products.
Taking a D 71°C of 16.5 seconds we can calculate a D 121°C of 1.4 x 10-5 seconds. Using a holding time of just 1 second would be predicted to give 1 / 1.4 x 10-5 = 10,000 log reductions of MAP. Even allowing for linearity errors and variations in D values this level of microbial destruction would be expected to give a high level of safety.
Changing the statutory requirements for pasteurization
There is now significant data indicating that commercial milk pasteurization at 72°C for 15 seconds can not be relied upon to give a 5 log reduction of MAP. As a consequence the industry in Ireland and the UK have reacted by voluntarily increasing the severity of heat treatment used in pasteurization and work is being done to control Johne’s disease in cattle.
While Johne’s disease is notifiable in Ireland and some other countries it is not a notifiable animal disease in the UK. This difference in notification policy is probably down to the greater national importance of food manufacturing and export in Ireland.
As stated earlier many milk-processors have increased the heat treatment of milk for retail sale but that is not the case for all. Additionally not all farmers are participating in Johne’s eradication schemes.
While there is an ongoing debate about the role of MAP as the primary causal agent of Crohn’s disease there is now a consensus that it is a human pathogen, however, this has still be to be formalised globally. In the meantime and until MAP can be eliminated from the milk supply it would be prudent to apply the precautionary principle and increase the statutory time / temperature conditions for the processing of milk and to introduce compulsory, national, Johne’s eradication schemes.
Thanks to Professors Joe Frank and Horacio Bach for making copies of their work on MAP available.
Bach,H.(2015).What role does Mycobacterium avium subsp. paratuberculosis play in Crohn’s Disease? Curr Infect Dis Rep. 17, 3-11.
Cerf, O. and Condron, R. (2006). Coxiella burnetii and milk pasteurization: an early application of the precautionary principle? Epidemiol. Infect. 134, 946–951.
Codex Alimentarius (2004). Code of Hygienic Practice for Milk and Milk Products, CAC/RCP 57–2004 (Amended 2009). Page 22. Available from: <http://www.codexalimentarius.net/download/ standards/10087/CXP_057e.pdf>. Accessed 7th December, 2015.
FDA/CFSAN, (2000). Kinetics of Microbial Inactivation for Alternative Food Processing Technologies - Overarching Principles: Kinetics and Pathogens of Concern for All Technologies. Available from: <http://www.fda.gov/food/foodscienceresearch/safepracticesforfoodprocesses/ucm100198.htm>.
Food Standards Agency (2002). Strategy for the control of Mycobacterium avium subspecies paratuberculosis (MAP) in cows milk. [On-line] Available from: http://tna.europarchive.org/20111030113958/http://www.food.gov.uk/multimedia/pdfs/map_strategy.pdf.
Gut, J. A. W., Fernandes, R., Tadini, C. C., and Pinto, J. M. (2004). HTST Milk Processing: Evaluating the thermal lethality inside plate heat exchangers. International Conference Engineering and Food. ICEF9. Montpellier, France. Available from: <http://sites.poli.usp.br/pqi/lea/docs/icef2004a.pdf> .
Grant, I. R., Ball, H. J. and Rowe, M. T. (1999). Effect of higher pasteurisation temperatures, and longer holding times at 72°C, on the inactivation of Mycobacterium paratuberculosis in milk. Lett. Appl. Microbiol., 28, 461- 465.
Grant I.R., Williams A.G., Rowe M.T. and Muir D.D. (2005). Efficacy of various pasteurization time-temperature conditions in combination with homogenization on inactivation of Mycobacterium avium subsp. paratuberculosis in milk. Appl. and Environ. Microb., 71, 2853–2861.
ICMSF. 1996. Microorganisms in foods 5 - Characteristics of microbial pathogens. New York: Blackie Academic & Professional.
Juffs, H. and Deeth, H. (2007). Scientific evaluation of pasteurisation for pathogen reduction in milk and milk products. Canberra and Wellington: Food Standards Australia New Zealand. Available from: <http://www.foodstandards.gov.au/code/proposals/documents/Scientific%20Evaluation.pdf>. Accessed 5 December 2015.
Keswani, J. and Frank, J. F. (1998). Thermal inactivation of Mycobacterium paratuberculosis in milk. J. of Food Prot., 61, 974-978.
Kuenstner et al. (2017). The Consensus from the Mycobacterium avium ssp.paratuberculosis (MAP) Conference 2017. Frontiers in Public Health. Available from: https://doi.org/10.3389/fpubh.2017.00208
Lang, G.I.A. and Jordan, W.K. (1958). Comparison of standard methods of measuring holding time in H.T.S.T. pasteurizers. J. Dairy Science. 41, 769-775.
Lynch, D., Jordan,K. N., Kelly, P. M., Freyne,T. and Murphy, P. M. (2007). Heat sensitivity of Mycobacterium avium ssp. paratuberculosis in milk under pilot plant pasteurization conditions. Int. J. of Dairy Technol. 60,98–104.
Mackey, B.M. and Bratchell, N. (1989). The heat resistance of Listeria monocytogenes. Lett. Appl. Microbiol. 9, 89-94.
Mullan, W.M.A. (2008). Modelling the destruction of Mycobacterium avium subsp. paratuberculosis. [On-line]. Available from: <https://www.dairyscience.info/index.php/thermal-processing/161-map.html>. Accessed: 7th December, 2015. Updated May 2013. 2015.
Mullan,W.M.A. (2012).Thermal processing of acid fruit and vegetable products. Significant microorganisms, recommended processing time / temperatures, and public health significance of spoilage. Ebook available from: <http://www.dairyscience.info/index.php/thermal-processing.html>.
Mullan, W.M.A. (2018). UHT processing. Lethality, chemical effects and use of temperature-time-integrators. [On-line]. Available from: https://www.dairyscience.info/index.php/thermal-processing/325-uht-processing.html . Accessed: 10 April, 2018.
O'Reilly, C.E., O'Connor, L., Anderson, W., Harvey, P., Grant, I.R., Donaghy, J., Rowe, M. and O'Mahony, P. (2004). Surveillance of bulk raw and commercially pasteurised cows’ milk from approved Irish liquid milk pasteurising plants for Mycobacterium paratuberculosis. Appl. Environ. Microb. 70, 5138 – 5144.
PMO (2011). Grade A Pasteurized Milk Ordinance. Food and Drug Administration, College Park, MD. Available from: Regulation/GuidanceDocumentsRegulatoryInformation/Milk/ucm2007966.htm. Accessed 7th December, 2015.
Stumbo, C. R. (1973). Thermobacteriology in food processing, 2nd ed. Academic Press, New York.
Schwarzer, K., Schneider, J., Becker, B. and Müller, U. (2007). Lemgo D- and z-value database for food. Available from: <http://www.hs-owl.de/fb4/ldzbase/>.
Sung, N. and Collins, M.T. (1998). Thermal tolerance of Mycobacterium paratuberculosis. Appl. Environ. Microb., 64, 999–1005.
How to cite this article
Mullan, W.M.A. (2015). [On-line]. Available from: https://www.dairyscience.info/index.php/food-model/277-htst-pasteurization.html . Accessed: 20 June, 2018. Updated October, 2017; April 2018.