Review Reports - Analytical Methods for the Characterisation of Aroma Compounds in Milk

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

General comments

The manuscript provides a comprehensive overview of milk aroma analysis. However, it currently leans towards being descriptive rather than critical. To achieve the level of a high-impact review, the authors must transition from summarizing literature to providing expert guidance and addressing the fragmentation between chemical and sensory data.

Abstract

The abstract provides a general overview of the topic but remains largely descriptive. The authors should more clearly define the scope of the review, specify the criteria used for comparing analytical techniques, and highlight the key conclusions or recommendations derived from the analysis to improve clarity and impact.

Introduction

The Introduction provides relevant background on milk aroma analysis and volatile extraction techniques; however, it lacks a clearly defined research objective and critical positioning within existing review literature. The authors should better articulate the purpose and novelty of this review, clarify what gap it addresses compared to previous studies, and improve the logical flow by reducing repetition and strengthening the connection between concepts.

The methodology for literature selection and analysis is not sufficiently described. The authors should clarify the criteria used for selecting studies, including databases searched, inclusion/exclusion criteria, and any systematic approach applied.

Chemical Drivers of Milk Aroma

Sections 2.3 & 2.4. There is insufficient focus on emerging processing technologies. While UV-C and HHP are mentioned, the manuscript does not provide a comparative evaluation of their specific volatile markers relative to conventional treatments such as UHT. A clearer differentiation in terms of aroma impact is needed. Section 2.5. The authors identify flavored milk as a “large research gap”; however, this statement is not further explored. As a review article, the manuscript should attempt to explain the underlying reasons for this gap, such as potential masking effects of added ingredients or the lack of standardized extraction methodologies for high-sugar and high-solids dairy systems.

Sample Preparation Methods

This is one of the strongest sections of the manuscript; however, it remains overly descriptive. In Section 3.1.1.2, the authors report a wide temperature range (10–60°C) for SPME without providing guidance. The authors should include a “best practice” summary (e.g., table or schematic) outlining optimal conditions (fiber type, temperature, extraction time) for different milk systems (e.g., raw vs. UHT), with consideration of artifact formation and reproducibility.

Volatile Profiling Techniques

The discussion of comprehensive two-dimensional GC (GC×GC) is too brief relative to its importance. Additional detail should be provided on modulation techniques (e.g., thermal vs. flow modulation) and their influence on sensitivity and detection of trace aroma compounds. Please, expand the discussion on column orthogonality (e.g., polar vs. non-polar phases) and its relevance for resolving key compound classes in milk, particularly lactones and sulfur-containing volatiles.

Sensory Analysis and Instrumental Correlation

The discussion of electronic noses and electronic tongues lacks critical evaluation. The authors should clearly state that these tools are primarily suited for classification and pattern recognition, and cannot replace human sensory analysis.

Recommendation: Strengthen the role of GC-Olfactometry (GC-O) by emphasizing that certain key odorants may be present below instrumental detection limits (MS) but remain perceptible to the human olfactory system.

Data Analysis and Chemometrics

Figure 2 (analytical workflow) is informative but remains generic and does not provide sufficient methodological insight. The authors should suggest specific and appropriate statistical or machine learning models (e.g., PLS-R, Random Forest) for correlating instrumental and sensory data, rather than listing techniques without guidance.

Future Perspectives

The Future Perspectives section remains too general and does not provide concrete or actionable directions for advancing the field. The authors should propose specific future research directions, including the development of standardized methodologies for milk volatile analysis, improved integration of sensory and instrumental data, and the application of advanced analytical and machine learning approaches.

Conclusion

The Conclusion section is largely a summary of the reviewed content and does not sufficiently emphasize the key takeaways or the overall contribution of the manuscript. The authors should provide a more concise and focused synthesis of the most important insights from the review, clearly highlighting which analytical approaches are most effective under specific conditions. The section should avoid repeating descriptive details and instead emphasize the main contributions, practical relevance, and added value of the work.

Author Response

Comment 1: The manuscript provides a comprehensive overview of milk aroma analysis. However, it currently leans towards being descriptive rather than critical. To achieve the level of a high-impact review, the authors must transition from summarizing literature to providing expert guidance and addressing the fragmentation between chemical and sensory data.

Response 1: Thank you very much for taking the time to review this manuscript. We greatly appreciate your feedback on the general tone and writing style, as well as your specific recommendations to improve the overall quality of this review.

Comment 2: The abstract provides a general overview of the topic but remains largely descriptive. The authors should more clearly define the scope of the review, specify the criteria used for comparing analytical techniques, and highlight the key conclusions or recommendations derived from the analysis to improve clarity and impact.

Response: Thank you for the feedback on how we can improve the abstract for our manuscript. We have redone the abstract to more clearly define the scope and highlight the key conclusions. We have not discussed how the analytical techniques were compared, as this varies by technique and depends on what is available in the literature. The modified abstract can be seen on page 1, paragraph 1, lines

Comment 3: The Introduction provides relevant background on milk aroma analysis and volatile extraction techniques; however, it lacks a clearly defined research objective and critical positioning within existing review literature. The authors should better articulate the purpose and novelty of this review, clarify what gap it addresses compared to previous studies, and improve the logical flow by reducing repetition and strengthening the connection between concepts.

Response 3: Thank you for your feedback on the writing style used in the introduction. We appreciate the detailed feedback you have given us to improve the introduction to better explain the purpose of this review and the gaps it addresses. We have rewritten the introduction to better reflect the research objectives and positioning within the current body of literature. Changes occur from page 1 until page 3, in all paragraphs. Specific line changes occur on lines 36, 48-49, 51-52, 54-58, 60-62, 64-78, 87-111.

Comment 4: The methodology for literature selection and analysis is not sufficiently described. The authors should clarify the criteria used for selecting studies, including databases searched, inclusion/exclusion criteria, and any systematic approach applied.

Response 4: Thank you for your recommendation to include a methodology for how we selected the studies. We did not include this because the review was written as a comprehensive, narrative review rather than a systematic review. As such, while we did have selection criteria, they were not strictly adhered to, as deviations from the criteria were needed to identify sufficient literature to adequately compare the methods. For your reference, we have added the methodology below, but we have not included it in the revised manuscript.
A detailed literature review was conducted using the EBSCO database to identify articles. The following criteria were used to screen and select articles: 1) studies involving bovine dairy milk, 2) studies that focus on analytical chemistry techniques for the analysis of volatiles, and 3) sensory data used as supporting information for instrumental approaches. The search strategy for this review involved using keywords such as “dairy”, “milk”, “volatile*”, “aroma”, “flavo*”, “analytical chemistry”, “GC—MS”, “GC×GC”, “SPME” and “olfacto*”. Studies were included in this review if they were published in the last 15 years and were peer-reviewed. Studies were excluded if they focused on other animal milks (such as sheep, horse, or human milk) or fermented dairy products, or if they relied exclusively on sensory methods without instrumental support. Some studies are included as exceptions to the exclusion criteria because of their high applicability.

Comment 5: Sections 2.3 & 2.4. There is insufficient focus on emerging processing technologies. While UV-C and HHP are mentioned, the manuscript does not provide a comparative evaluation of their specific volatile markers relative to conventional treatments such as UHT. A clearer differentiation in terms of aroma impact is needed.

Response 5: Thank you for highlighting this limitation in the discussion of how processing can impact the flavour of milk products. We appreciate your feedback on how to improve this section by providing a comparison of the volatility profiles for each manufacturing type. We have not included this, as comparing the various methods to one another is complicated by studies that use inconsistent methods, do not include quantification, or omit sensory analysis. A fair comparison of processing methods cannot be made due to these inconsistencies.

Comment 6: Section 2.5. The authors identify flavored milk as a “large research gap”; however, this statement is not further explored. As a review article, the manuscript should attempt to explain the underlying reasons for this gap, such as potential masking effects of added ingredients or the lack of standardized extraction methodologies for high-sugar and high-solids dairy systems.

Response 6: Thank you for your comment on the gap surrounding flavoured milk studies. We do point out in the text that only one study was identified that considered flavoured milk matrices, indicating a significant gap. We did not make any assertions about why there is a gap, as there is insufficient literature to make assumptions about its drivers. We have added a brief explanation of possible causes on page 9, paragraph 2, lines 376-380.

Comment 7: This is one of the strongest sections of the manuscript; however, it remains overly descriptive. In Section 3.1.1.2, the authors report a wide temperature range (10–60°C) for SPME without providing guidance. The authors should include a “best practice” summary (e.g., table or schematic) outlining optimal conditions (fiber type, temperature, extraction time) for different milk systems (e.g., raw vs. UHT), with consideration of artifact formation and reproducibility.

Response 7: Thank you for your comment here on how to improve the quality of the review by giving best-practise advice for these sections. We have included best-practice tables for both HS-SPME and SAFE as they have the highest variation. The HS-SPME table can be found on page 14, with a descriptive paragraph from line 590-608.

Comment 8: The discussion of comprehensive two-dimensional GC (GC×GC) is too brief relative to its importance. Additional detail should be provided on modulation techniques (e.g., thermal vs. flow modulation) and their influence on sensitivity and detection of trace aroma compounds. Please, expand the discussion on column orthogonality (e.g., polar vs. non-polar phases) and its relevance for resolving key compound classes in milk, particularly lactones and sulfur-containing volatiles.

Response 8: Thank you for the recommendation on improving the GC×GC discussion in the manuscript. We agree that it is an incredibly important method for volatile analysis, and that modulation and column orthogonality are crucial components of this technique. However, only one study was identified that employed GC×GC to analyse milk volatiles, so any discussion of modulation and orthogonality would be entirely theoretical. Such a discussion would be more relevant in a review of GC×GC methods, as this review is particularly interested in the application of analytical techniques to milk. We cannot adequately discuss how column selection and modulation methods would improve volatile analysis without additional comparative studies.

Comment 9: The discussion of electronic noses and electronic tongues lacks critical evaluation. The authors should clearly state that these tools are primarily suited for classification and pattern recognition, and cannot replace human sensory analysis.

Response 9: Thank you for pointing out that we had not adequately represented that electronic sensor arrays cannot replace human sensory analysis methods. We completely agree and have clarified that these sensors can only be used for basic pattern recognition and that human sensory analysis is required to draw any conclusions about aroma and flavour. We have added this modification to page 27-28, paragraph 2, lines 1226-1230.

Comment 10: Recommendation: Strengthen the role of GC-Olfactometry (GC-O) by emphasizing that certain key odorants may be present below instrumental detection limits (MS) but remain perceptible to the human olfactory system.

Response 10: Thank you for the recommendation to strengthen our discussion of GC—O. We have included an explicit statement on page 28, paragraph 2, lines 1232-1235, to emphasise that certain aromas are undetectable to human perception.

Comment 11: Figure 2 (analytical workflow) is informative but remains generic and does not provide sufficient methodological insight. The authors should suggest specific and appropriate statistical or machine learning models (e.g., PLS-R, Random Forest) for correlating instrumental and sensory data, rather than listing techniques without guidance.

Response 11: Thank you for your feedback on how we can improve the figure. We have modified it to be more informative about the recommended workflow for conducting a correlation study that incorporates instrumental and sensory data. Additionally, we have included an extra explanation of the three machine learning algorithms on page 29, paragraph 1, lines 1299-1315, that indicates how each method should be used for analysis.

Comment 12: The Future Perspectives section remains too general and does not provide concrete or actionable directions for advancing the field. The authors should propose specific future research directions, including the development of standardized methodologies for milk volatile analysis, improved integration of sensory and instrumental data, and the application of advanced analytical and machine learning approaches.

Response 12: Thank you for the recommendations on how we can improve the future perspectives section of the manuscript. We have rewritten the entire section to provide clearer, actionable research directions to guide future work analysing volatiles in dairy matrices. The changes are on pages 32-33, paragraphs 1-7, lines 1433-1490.

Comment 13: The Conclusion section is largely a summary of the reviewed content and does not sufficiently emphasize the key takeaways or the overall contribution of the manuscript. The authors should provide a more concise and focused synthesis of the most important insights from the review, clearly highlighting which analytical approaches are most effective under specific conditions. The section should avoid repeating descriptive details and instead emphasize the main contributions, practical relevance, and added value of the work.

Response 13: Thank you for the recommendations on how to improve our conclusion. We have rewritten it to clearly indicate the key takeaways relevant to researchers seeking to understand current methods for analysing volatiles in milk. The Modifications start on page 34, paragraphs 5, lines 1515-1561.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript provides a well-documented and up-to-date review of analytical methods used for the characterization of aroma compounds in milk, clearly highlighting the relationship between processing, volatile composition, and sensory perception. The structure is logical, and the literature integration is strong, particularly in the sections discussing HS-SPME and UHT processing.

As a suggestion for improvement, some methodological sections could be slightly condensed, as certain passages becom overly detailed and repeat similar points regarding extraction temperatures, times, and methodological limitations. Readability could be improved by adding more comparative summary tables or conceptual figures. In addition the discussion on the correlation between instrumental data and consumer perception could be expanded with more concrete examples of statistical or chemometric approaches used in recent studies. The section on emerging methodologies (multi-omics, AI/data fusion) would also benefit from a more practical discussion of current applications and limitations.

Author Response

Comment 1: The manuscript provides a well-documented and up-to-date review of analytical methods used for the characterization of aroma compounds in milk, clearly highlighting the relationship between processing, volatile composition, and sensory perception. The structure is logical, and the literature integration is strong, particularly in the sections discussing HS-SPME and UHT processing.

Response 1: Thank you very much for taking the time to review this manuscript. We appreciate your comments and feedback on our review.

Comment 2: As a suggestion for improvement, some methodological sections could be slightly condensed, as certain passages become overly detailed and repeat similar points regarding extraction temperatures, times, and methodological limitations. Readability could be improved by adding more comparative summary tables or conceptual figures. In addition, the discussion on the correlation between instrumental data and consumer perception could be expanded with more concrete examples of statistical or chemometric approaches used in recent studies. The section on emerging methodologies (multi-omics, AI/data fusion) would also benefit from a more practical discussion of current applications and limitations.

Response 2: Thank you very much for your recommendations to improve the readability of our manuscript. To address your feedback, we have included additional summary tables in the main text. The included tables are designed to provide summaries that can inform the selection of optimal experimental conditions for different extraction methodologies. The first table on HS-SPME recommendations has been added on page 11, with an extra paragraph added to explain the table in detail from lines 483-501. The second added table presents recommended experimental conditions for solvent-assisted flavour evaporation (SAFE), with an explanatory paragraph on page 16 starting at line 756. We have also added an additional image, providing a simple visual method for differentiating between pasteurised and UHT milk. This figure (Figure 1) is added on page 5, with the following statement added on page 5, line 231-232: “Figure 1 below shows the standard temperatures and holding times used for traditional pasteurisation and UHT processing of bovine milk.”

Thank you very much for your helpful comments and suggestions to elaborate and link instrumental with sensory data. While we would have liked to include some examples from the recent scientific literature, no such linkage was conducted, highlighting one of the key gaps discussed in this manuscript. While recent research links physicochemical properties to sensory outcomes, this has not been extended to volatile data and, therefore, is not included in our manuscript. We also appreciate your comment on a more practical discussion of emerging technologies and their applications in the analysis of milk. We have included a discussion of how the three major machine learning algorithms should be used, along with their respective limitations. The following statement has been added on page 31, lines 1425-1441 “The three key MLAs, SVM, CART, and RF, have different applications and drawbacks when correlating volatile analysis with sensory outcomes. SVM can be used for both regression and classification, using volatile data to predict sensory outcomes or to classify samples according to sensory scores. However, SVMs do not provide information about which volatiles drive the sensory outcomes, limiting actionable product insights. In contrast, CART can also be used for regression and classification analyses, but provides a clearer link between volatile data and sensory scores. CART is also capable of revealing threshold effects, determining which sensory outcomes are influenced by specific volatile concentrations. CART can be limited by high overfitting, especially with large volatile datasets and small sensory datasets. CART can also overlook multi-compound interactions, instead preferring single-factor predictors of sensory outcomes. RF is more robust to overfitting than CART and provides relative importance rankings, which can be very important for identifying driver compounds behind sensory outcomes. Variable importance may be biased to more abundant compounds, not necessarily those that carry the greatest sensory impact. Additionally, RF cannot account for thresholds and synergistic interactions among volatiles when determining sensory outcomes.”

Reviewer 3 Report

Comments and Suggestions for Authors

Abstract - no remarks.

Introduction - line 35 - ESL is considered to be part of pasteurized milks. It is good to include this new technology, to consider the impact of temperature on the flavor profile of the milk. Nothing has been said about ISI technology.

In my opinion, there are too few literary sources reviewed in the introduction.

Line 72-79 - It should be emphasized that the subject of this review is exclusively milk that is from bovine and that "meets" all organoleptic, physicochemical and microbiological requirements, because the development of unwanted side microflora can significantly affect VOCs.

Line 81-84 - in my opinion, this should be cleared up, because nowhere is the influence of fat content on VOCs considered. Otherwise, where are the other ingredients of milk, the change of which would result in a reflection in VOCs? I did not find a part in the article that treats the influence of each ingredient of milk separately on these substances. Only other factors such as diet, technological processes and others are found.

2.2. - UHT is a type of sterilization and should be removed from this point.

point 2.4. - What about lactose-free milks - how does the VOC profile change with them? Line 235 - Is transglutaminase treatment allowed in practice?

Besides temperature treatment, how does mechanical treatment - cleaning, homogenization - affect? There is a lack of data on the "deodorization" process. It significantly affects VOCs.

This point does not describe where the respective types of pasteurization find further application. The answer to this question can be discussed by paying attention to how the respective temperature treatment also affects VOCs in the resulting product - drinking milk, cheese, fermented milks, etc.

In this section, for pasteurization, different pasteurization regimes should be considered - with different times and retention times, as well as the inclusion of ESL technology. In sterilization, one and two-stage sterilization, UHT and ISI technology should be considered.

Why is ultrasonic processing not considered? Why is the influence of packaging on the VOC profile not considered?

Section 2.5. - to separate the influences of natural and artificial flavors added to milks.

Point 2.5. - in my opinion, the article should be limited to bovine milk. The consideration of breast milk and alternatives to dairy products are completely different cases. This should be removed from the article.

The entire second point lacks any type of visualization of the presented data, for the purpose of easier reading and comprehension - table, figure, graph.

I have no comments on the analytical part. point 7 - line 1270 full cream should be replaced with full fat milk. Conclusion - To emphasize what has not been done and what are the future directions for work based on the review in one sentence.

Author Response

Comment 1: Introduction - line 35 - ESL is considered to be part of pasteurized milks. It is good to include this new technology, to consider the impact of temperature on the flavor profile of the milk. Nothing has been said about ISI technology.

Response 1: Thank you very much for bringing this to our attention. We were not referring to ESL as its own type of milk, but rather to indicate that longer-shelf-life products are increasingly desired. We have clarified the language to emphasise this difference by adding the following statement on page 1, line 53: “As dairy markets increasingly rely on products with a longer shelf life such as pasteurised and ultra‑high‑temperature (UHT) milk”

Comment 2: In my opinion, there are too few literary sources reviewed in the introduction.

Response 2: Thank you very much for your comments regarding improving the number of references in the introduction. As part of our response to another reviewer, we have reworded the introduction and added additional references within the text.

Comment 3: Line 72-79 - It should be emphasised that the subject of this review is exclusively milk that is from bovine and that "meets" all organoleptic, physicochemical and microbiological requirements, because the development of unwanted side microflora can significantly affect VOCs.

Response 3: Thank you very much for your helpful suggestions about emphasising that this comprehensive review is focused on bovine milk. To improve clarity, we have refined the content throughout the manuscript to focus solely on bovine milk, except for the work by High et al. (2019), which provides a key comparison of multiple extraction techniques in a milk matrix, particularly sheep’s milk.

Comment 4: Line 81-84 - in my opinion, this should be cleared up, because nowhere is the influence of fat content on VOCs considered. Otherwise, where are the other ingredients of milk, the change of which would result in a reflection in VOCs? I did not find a part in the article that treats the influence of each ingredient of milk separately on these substances. Only other factors such as diet, technological processes and others are found.

Response 4: Thank you very much for your comment on improving the clarity of our manuscript. We have now removed our statement around fat from page 2, paragraph 5, lines 82-83, and clarified that we are not discussing the impact of individual physicochemical properties on aroma.

Comment 5: 2.2. - UHT is a type of sterilization and should be removed from this point.

Response 5: Thank you for bringing this to our attention. We have now removed mention of UHT processing from paragraph 3 on page 3, in line 128.

Comment 6: Point 2.4. - What about lactose-free milks - how does the VOC profile change with them?

Response 6: Thank you for noting the lack of analysis regarding lactose-free milk. We acknowledge the importance of these products and have included a discussion on the volatiles in lactose-free milk compared to conventional milk. The following statement has been added on page 8-9, lines 350-371: “Lactose-hydrolysed, or lactose-free, milk is a key product to consider given the rising prevalence of lactose intolerance in consumers. While these products are becoming increasingly popular, research into the volatile changes in these milks remains limited. Jansson et al. (2014) investigated the volatiles present in UHT-processed lactose-free milk and compared them with those in conventional UHT and filtered (low-lactose) UHT milk. This study revealed that conventional UHT milk had the highest ketone abundance, which is responsible for stale notes in UHT milk [39]. Aldehydes and ketones, as products of the Maillard reactions, are dependent on the sugar content within the milk matrix during processing. The removal of lactose from the milk will likely limit these reactions, resulting in the decreased ketone concentration observed in the lactose-free products. The reduced ketone content is supported by Jensen et al. (2015), who compared conventional UHT milk to lactose-free milk processed using direct and indirect UHT technologies. This study revealed the importance of UHT processing technologies: direct UHT led to an increase in ketone abundance in lactose-free milk during storage, whereas indirect UHT started with a higher ketone abundance that continued to decrease during storage [40]. This investigation revealed that lactose-free milk, regardless of direct or indirect UHT, had a higher bitterness intensity than conventional UHT milk, linked to higher levels of primary amines resulting from proteolytic reactions. The higher bitterness in lactose-free milk was linked to the activity of the β-galactosidase enzyme added to hydrolyse lactose, suggesting a need to identify more effective methods for removing lactose to maintain sensory outcomes.”

Comment 7: Line 235 - Is transglutaminase treatment allowed in practice?

Response 7: Thank you very much for your helpful comments. We had not discussed the investigation of the use of transglutaminase enzymes in commercial milk products. We have also not addressed this in the text, as it is beyond the scope of our review; however, within Australia, transglutaminase is approved for use in fermented dairy and dairy analogues.

Please find the report at the following link: https://www.foodstandards.gov.au/sites/default/files/2024-03/A1275%20Approval%20Report.pdf

Comment 8: Besides temperature treatment, how does mechanical treatment - cleaning, homogenization - affect? There is a lack of data on the "deodorization" process. It significantly affects VOCs.

Response 8: Thank you very much for pointing out this gap in our comprehensive review. We appreciate your feedback on how we can improve the discussion of processing impacts on volatiles. After refining the scope of this review to focus on bovine milk drinks, we conducted an additional literature search to identify studies that examine the impact of mechanical treatment on milk volatiles. In the literature, VOC analysis has been conducted only on cheese, not directly on milk prior to fermentation. As such, we have not included any discussion of these processes.

Comment 9: This point does not describe where the respective types of pasteurization find further application. The answer to this question can be discussed by paying attention to how the respective temperature treatment also affects VOCs in the resulting product - drinking milk, cheese, fermented milks, etc.

Response 9: Thank you for identifying a gap in the discussion. In response to some of your earlier comments, we have refined the scope to focus on liquid bovine milk intended for drinking, rather than other applications. As such, we have not modified the manuscript to discuss the use of pasteurised milk, as we consider it the final product.

Comment 10: In this section, for pasteurization, different pasteurization regimes should be considered - with different times and retention times, as well as the inclusion of ESL technology. In sterilization, one and two-stage sterilization, UHT and ISI technology should be considered.

Response 10: Thank you very much for your recommendations to improve our discussion of pasteurisation and UHT processing. While these technologies are important to understand, the current body of scientific literature lacks studies investigating how these processing techniques affect the volatiles in bovine milk. As such, we are unable to include a discussion of these methods, as the literature focuses more on other physicochemical properties than on volatile aroma compounds.

Comment 11: Why is ultrasonic processing not considered? Why is the influence of packaging on the VOC profile not considered?

Response 11: Thank you very much for mentioning the use of ultrasonic processing. We have included a discussion of this technique. We appreciate your feedback on including this technique, as it has added valuable information to the manuscript.

The following statement has been added on page 8, lines 324-347 “Ultrasonic processing is one of the most popular non-thermal processing techniques for bovine milk. It is widely used for its ability to improve processing efficiency and maintain the functional properties of milk products. While limiting milk exposure to high thermal loads, ultrasonic processing does induce some volatile changes, mostly linked to oxidation reactions driven by free radicals generated by ultrasonic disruption of microorganisms within the matrix [35]. Cheng et al. (2025) investigated how high-intensity ultrasonication affected the interactions between volatiles and β-lactoglobulin, which is known to bind various chemical compounds and possibly impact flavour. This investigation indicated that β-lactoglobulin is capable of binding volatile compounds, affecting aroma retention in milk, with ultrasonic treatment altering the binding mechanics and leading to a higher number of ketones within the matrix [36]. However, this work did not determine whether these were naturally occurring ketones in the milk or oxidation products from ultrasonic processing. Chouliara et al. (2010) investigated how ultrasonic processing interacted with different thermal processing methods, and the resulting impact on microbial, chemical and sensory properties. The investigation found that ultrasonic treatment resulted in taste scores equal to or lower than those of untreated milk for all milk samples [37]. When considering lipid oxidation factors, it was determined that ultrasonication did significantly increase the rate of lipid oxidation; however, still lower than the limit of becoming sensorially unacceptable [38]. Additionally, it was determined that secondary oxidation products, namely aldehydes, increased during storage in ultrasonicated samples, whilst untreated milks remained relatively stable in aldehyde abundances [37]. These investigations indicate that while ultrasonic processing may be an effective non-thermal processing technique, the sensory and volatile impact may be a limiting factor for consumer acceptance.”

We have included a discussion of packaging materials and their impact on milk volatiles. Thank you for advising us on including this discussion.

The following statement has been added on page 10, lines 430-444: “Another key factor driving volatility during storage is the specific packaging used to bottle the milk. Simon and Hansen (2001) evaluated how different packaging types impacted the taste of milk refrigerated for 15 weeks. Standard cardboard packaging had a higher rate of flavour deterioration than packaging with a foil barrier [52]. Additionally, cardboard packaging imparted a notable cardboard flavour, while the foil barrier imparted a more pronounced cooked flavour. The cardboard flavour continued to intensify, while the cooked flavour deteriorated. Karatapanis et al. (2006) investigated a wider range of packaging materials, including clear glass, clear PET, coloured PET, three-layer co-extruded HDPE, coloured monolayer HDPE and cardboard. The analysis revealed that clear packaging exhibited higher volatility, whereas coloured packaging reduced the observed rate of change [6]. This is likely due to the amount of light being blocked by the coloured packaging, reducing the influence of photo-oxidation, as indicated by the absence of dimethyl sulphide, a key photo-oxidative product, in milk stored in darkly coloured bottles. The selection of appropriate packaging materials will affect the volatile profiles during storage and the sensory acceptance of these products.”

Comment 12: Point 2.5. - in my opinion, the article should be limited to bovine milk. The consideration of breast milk and alternatives to dairy products are completely different cases. This should be removed from the article.

Response 12: Thank you for your advice on refining the scope of our manuscript. We have edited our review to reflect this change and have removed breast milk and dairy alternatives from our discussion. Discussion of breast milk was removed from page 6, paragraph 2, with edits made to lines 244-247 and 249-251. We have removed the discussion of UV-C processing of coconut milk on page 7, paragraph 2, lines 262-265. In addition, we have removed plant-based milks that were analysed using PTR—TOF—MS on page 23, paragraph 1, lines 1018-1019.

Comment 13: The entire second point lacks any type of visualization of the presented data, for the purpose of easier reading and comprehension - table, figure, graph.

Response 13: Thank you very much for your comment. We did not include any visual components here due to the amount of data that would need to be communicated in a figure, table or graph. Any visual that included a comparison of all the drivers discussed was too cluttered and difficult to read at a glance, defeating the purpose of including one.

Comment 14: Point 7 - line 1270 full cream should be replaced with full fat milk.

Response 14: Thank you for pointing out our oversight regarding the language. We have edited the line from reading “full-cream milk” to “full-fat milk”.

Comment 15: Conclusion - To emphasize what has not been done and what are the future directions for work based on the review in one sentence.

Response 15: Thank you very much for your kind feedback on how we could improve our conclusion. Another reviewer had additional comments, which we responded to. This also included a similar statement that we needed to emphasise what was analysed in this review and key future directions. The new conclusion addresses both their comments and yours.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for submitting the revised version of your manuscript and for the effort invested in addressing the reviewer comments. The manuscript has improved in several areas, particularly regarding the restructuring of the Introduction, the expansion of the Future Perspectives section, and the inclusion of practical guidance in the HS-SPME discussion.

However, after carefully evaluating the revised manuscript and the rebuttal responses, I remain concerned that several major conceptual issues have not been adequately resolved.

The primary concern raised in the initial review was that the manuscript remained predominantly descriptive and lacked sufficient critical synthesis expected from a high-impact review article. Although some sections were expanded, the revised version still largely summarizes existing literature rather than providing a deeper interpretative framework, expert guidance, or critical integration of findings across studies.

In particular, several key reviewer comments were not substantively addressed:

Methodological transparency remains insufficient. While the authors provided literature selection criteria in the rebuttal letter, this information was intentionally omitted from the manuscript itself. Even for a narrative review, readers require a clear description of the literature search strategy, databases consulted, and inclusion/exclusion criteria to evaluate the scope and reliability of the review.
The discussion of emerging processing technologies remains limited. The response emphasizes inconsistencies across studies as justification for avoiding comparative evaluation. However, the purpose of a review article is to critically synthesize heterogeneous literature and provide interpretative insight despite methodological variation.
The discussion of GC×GC remains underdeveloped relative to its analytical importance. The rebuttal states that further discussion would be “entirely theoretical”; however, theoretical discussion of modulation strategies, column orthogonality, and implications for resolving milk aroma compounds would still provide substantial value to readers and is appropriate within a review article.
Several responses in the rebuttal rely on the absence of sufficient literature as justification for not expanding discussion points. While this limitation is understandable, the manuscript would benefit from stronger expert interpretation regarding the implications of these gaps and the methodological challenges contributing to them.

Overall, while the manuscript has improved, I do not believe the revision sufficiently addresses the major conceptual concerns raised during the first review round.

Author Response

Comment 1: Dear Authors,

However, after carefully evaluating the revised manuscript and the rebuttal responses, I remain concerned that several major conceptual issues have not been adequately resolved.

Response 1: Thank you very much for taking the time to review our changes and provide feedback on the rebuttal. We appreciate the effort and time you have put into helping us improve our manuscript. We appreciate the identification of improvements in the previous revisions and the recommendations for further improvements to the overall writing and to specific sections.

Comment 2: The primary concern raised in the initial review was that the manuscript remained predominantly descriptive and lacked sufficient critical synthesis expected from a high-impact review article. Although some sections were expanded, the revised version still largely summarizes existing literature rather than providing a deeper interpretative framework, expert guidance, or critical integration of findings across studies.

Response 2: Thank you for highlighting that we have not fully addressed your original comment about improving the amount of guidance and interpretation in the literature. We have made several modifications throughout the manuscript to improve the critical interpretation of the literature, with changes highlighted for easy identification. As other comments address specific sections within the manuscript, the modifications highlighted in this response are in the sensory methods discussion, in response to the previous round of comments that identified a lack of critique in that section. The included additions are also listed below for your reference.

The following statement was added to page 28, lines 1263-1271: “The inclusion of e-tongue techniques should be considered only a routine instrumental quality control approach, not a replacement for sensory evaluation or analytical flavour research. At this stage, e-tongue systems provide a rapid screening tool for monitoring product consistency, detecting batch-to-batch variation, and identifying potential deviations from expected quality profiles. They may be useful for complementing detailed physicochemical tests and non-volatile analyses of milk, particularly in quality assurance settings where rapid, repeatable measurements are required. However, any assessment of sensory perception must be conducted using human panels, including trained panels for descriptive analysis and consumer studies to determine sensory acceptance. The rapid screening capacity of e-tongue systems may support quality assurance when linked with comprehensive sensory testing, but their outputs should be interpreted as instrumental indicators rather than direct measures of human flavour perception. Therefore, e-tongue methods should not be used in isolation to conduct flavour research or to make conclusions about sensory quality without validation against human sensory data.”

The following statement was added to page 30, lines 1361-1366: “Overall, while AEDA is a well-established technique for aroma analysis, the limitations of a small number of sniffers, olfactory fatigue, and long analysis sessions with multiple dilutions likely necessitate multi-day experiments. AEDA has many potential sources of error. As such, aroma analysis should be conducted using MF techniques to improve statistical significance and reduce experimental load, thereby limiting between-day variation in samples.”

The following statement was added to page 31, lines 1397-1406: “However, this study highlights the lack of integration between sensory and instrumental analysis. While a link between compound identities and aromatic character is attempted, the statistical analysis relies exclusively on the aroma data and their changes. No attempt is made to quantify the volatiles and associate their continuous change during storage with the changing aroma character. By linking quantifiable changes in volatiles to observed aroma effects, it may be possible to identify multifactor interactions that shape aroma rather than merely assuming that single compounds are responsible for complex aroma character. This modification would enable the identification of mechanisms driving aroma change, creating actionable insights for manufacturers.”

Comment 3: In particular, several key reviewer comments were not substantively addressed:

Response 3: Thank you for the further clarification on the importance of including the literature search strategy in a narrative review. We have incorporated the previous response into the methodology section of the revised manuscript for clarity. We have highlighted the addition for easy identification.

The following statement is included on pages 2-3, line 89-100: “2. Methodology

A literature review was conducted using the EBSCO database to identify articles. The following criteria were used to screen and select articles: 1) studies involving bovine dairy milk, 2) studies that focus on analytical chemistry techniques for the analysis of volatiles, and 3) sensory data used as supporting information for instrumental approaches. The search strategy for this review involved using keywords such as “dairy”, “milk”, “volatile*”, “aroma”, “flavo*”, “analytical chemistry”, “GC—MS”, “GC×GC”, “SPME” and “olfacto*”. Studies were included in this review if they were published in the last 15 years and were peer-reviewed. Studies were excluded if they focused on other animal milks (such as sheep, horse, or human milk) or fermented dairy products, or if they relied exclusively on sensory methods without instrumental support. Some studies are included as exceptions to the exclusion criteria because of their high applicability.”

Comment 4: The discussion of emerging processing technologies remains limited. The response emphasizes inconsistencies across studies as justification for avoiding comparative evaluation. However, the purpose of a review article is to critically synthesize heterogeneous literature and provide interpretative insight despite methodological variation.

Response 4: Thank you for noting the limited analysis of differences in the volatile profiles of milk exposed to different processing techniques. We have included significant modifications to the discussion to include comparative analysis between the various processing techniques. The additions have been highlighted in text for easy identification and listed below for reference.

The following statement was added to pages 4-5, lines 190-196: “Pasteurisation represents the mildest range of changes to the volatiles and, therefore, aroma of milk for commercial consumption. The fresh, green aromas and slight sulphurous notes of raw milk are still present; however, the addition of Maillard products alters the profile in detectable ways. Namely, the increasing number of Strecker aldehydes leads to more cooked, malty and sweaty notes that can impact consumer acceptance. Additionally, the ketone content increases in pasteurised milk, promoting floral and sweet aromas within these processed milks.”

The following statement was added to page 6, lines 257-261: “The key changes observed in UHT milk products is an increase in cooked, stale and burnt notes, linked to the aldehydes and sulphur compounds produced by the Maillard reactions facilitated by higher temperatures. The acid content of these UHT milks is also higher than pasteurised milk, giving rise to more detectable sweaty and cheesy notes.”

The following statement was added to page 7, lines 287-293: “While sensory analysis has not been conducted for these HHP milk products, the increased aldehyde, ketone, and sulphur content would have similar effects on the overall aroma as observed with thermal processing techniques. There would likely be increased levels of burnt and cooked notes associated with aldehydes, with ketones imparting various aromas depending on the specific ketones formed. However, as these compounds are typically associated with thermal processing, the overall aroma impact will likely be less than that detected in pasteurised and UHT milks.”

The following statement was added to page 7, lines 302-306: “The more complex volatile profile of UV-C processed milk will impart a range of aromas to the final product, indicating a need for comprehensive sensory analysis. The possible link to lipid oxidation will likely promote off-aromas, including burnt, cardboard and stale notes linked to increased aldehyde and ketone content.”

Comment 5: The discussion of GC×GC remains underdeveloped relative to its analytical importance. The rebuttal states that further discussion would be “entirely theoretical”; however, theoretical discussion of modulation strategies, column orthogonality, and implications for resolving milk aroma compounds would still provide substantial value to readers and is appropriate within a review article.

Response 5: Thank you for your comment; it will help us improve the technical discussion of GC×GC, particularly modulation and orthogonality. We have overhauled the entire discussion with multiple modifications. All additions and changes have been highlighted in the revised manuscript for easy reference, with the changes listed below.

The following statement was added to page 22, line 1023: “with initial reports dating back to the [84].”

The following statement was added to pages 23-24, line 1029-1095: “different, and ideally orthogonal, selectivity.”

The following statements were added to page 23, lines 1036-1051: “which is then further separated in a long (typically 30 m) second dimension (2D) [86]. A microfluidic device, commonly known as the Dean's switch (DS) in honour of its inventor and marketed under trade names such as Capillary Flow Technology (CFT, Agilent Technologies) and SilFlow (Trajan Scientific and Medical), is used to connect the ¹D and ²D columns [87]. The device contains narrow internal pathways that minimise peak dispersion, while flow switching between the two columns is controlled by an electronic pressure controller (EPC). This arrangement allows narrow effluent bands (typically spanning a few minutes) to be diverted from the ¹D column and further separated on a full-length analytical column, thereby substantially increasing peak capacity. The trade-offs, however, are longer overall analysis times and a more complex experimental setup: heart-cut windows must be determined in advance, the GC oven typically requires cooling between the ¹D and ²D separations before being ramped again, and a cryotrap is often needed to re-focus the transferred band and prevent peak broadening. Although H/C MDGC has been widely applied in flavour analysis, its use in milk analysis remains limited, with most studies focusing on fatty acid methyl esters (FAMEs) and contaminants such as polychlorinated biphenyls (PCBs) [88-91].”

The following statements were added to page 23, lines 1036-1051: “Although several modulator types have been developed, such as thermal, valve-based, and flow modulators [90], thermal modulation is widely regarded as the most effective modulator for GC×GC [93], and is therefore the most common modulation method used in food analysis [94]. Cryogenic thermal modulators, commercially available under various trade names such as the LECO Dual Jet and the Zoex loop modulator, use liquid N2 or CO2 jets to trap and re-focus narrow bands of ¹D effluent before pulsing them into the ²D column. This produces the sharp, high-amplitude peaks that enhance GC×GC sensitivity, thereby improving library matching and compound identification. However, the consumable cost and operational complexity associated with cryogen supply have driven the development of consumable-free alternatives, including solid-state thermal modulators (e.g., resistively heated trap designs [95, 96], commercialised as the SepSolve INSIGHT modulator) and flow modulators such as forward-fill/flush [97] (LECO FLUX modulator), reverse-fill/flush [98] (LECO Paradigm Shift modulator), and multi-loop splitter-based devices [99]. Flow modulators are attractive for routine food laboratories given their lower running costs and simpler operation, as they require no cryogen logistics; however, they typically yield broader ²D peaks and reduced sensitivity for trace odour-active compounds. This is a rel-evant trade-off in milk aroma analysis, where some quality-impact volatiles (e.g., methi-onal and (E,E)-2,4-decadienal) could occur at trace concentrations [100].

In addition to the choice of modulation strategy, modulation period (typically 3-8 s) and 2D column dimensions must therefore be jointly optimised so that the 2D separation completes within each modulation cycle without wraparound. Equally critical is the selection of the stationary phase, since the choice of column pair governs the selectivity of each dimension and how effectively the ²D separation space is occupied. In a fully orthogonal system, the two columns separate analytes by independent retention mechanisms, distributing peaks across the ²D plane and minimising co-elution. In practice, however, true or-thogonality is rarely achieved because the retention of an analyte on any given phase re-flects a combination of interactions, and the contributions of volatility and polarity to re-tention are partially correlated across real analyte sets. Moreover, the term ‘polarity’ itself is an umbrella for several distinct intermolecular interactions, including cavity formation and dispersion interactions (l), lone-pair and π-electron interactions (e), dipolari-ty/polarisability (s), and hydrogen-bond acidity (a) and basicity (b), each of which contrib-utes differently to retention depending on the functional groups of the analyte and the chemistry of the stationary phase. The linear solvation energy relationship (LSER) frame-work, developed by Abraham and co-workers [101, 102], decomposes these contributions into corresponding solute (referring to analytes; L, E, S, A, and B) and solvent (referring to the stationary phase as the partitioning medium; l, e, s, a, and b) descriptors and has been widely applied to characterise GC stationary phases and to guide the selection of orthog-onal column pairs for GC×GC [103].”

The following statement was added to page 25, lines 1112-1126: “This arrangement is beneficial for resolving the chemical-class clusters relevant to milk flavour, such as separating alcohols, aldehydes, and ketones from one another, and resolving aromatic from aliphatic hydrocarbons. The reversed-polarity configuration (polar ¹D × non-polar ²D) is less commonly applied in dairy analysis, but it can be advantageous for separating closely related polar oxygenates that are difficult to resolve on non-polar ¹D columns, such as short-chain alcohols and aldehydes. Selectivity tuning through alternative ²D chemistries (e.g., ionic liquid phases for free fatty acids [105], or trifluoropropyl-substituted phases for compounds containing electron-rich functional groups [106]) represents a further avenue for targeted enhancement of milk aroma profiling, although such configurations remain underexplored in the dairy literature.

Volatile identification in this study combined NIST library spectral matching with linear retention indices (LRI) generated from an n-alkane mixture, providing an additional metric for confirming compound identities [20]. Peak matching was carried out using LECO ChromaTOF software, and compounds with a library match score ≥80% were considered tentatively identified.”

Comment 6: Several responses in the rebuttal rely on the absence of sufficient literature as justification for not expanding discussion points. While this limitation is understandable, the manuscript would benefit from stronger expert interpretation regarding the implications of these gaps and the methodological challenges contributing to them.

Overall, while the manuscript has improved, I do not believe the revision sufficiently addresses the major conceptual concerns raised during the first review round.

Response 6: Thank you for your comment, highlighting that we needed to include more interpretation of the existing literature. We expanded the literature discussions across multiple sections of the manuscript, enhancing the technical and comparative components of the review. Our discussions provide additional context whilst still maintaining the journal's scope by highlighting the application of analytical techniques to milk analysis. We have made significant additions to the following sections of the manuscript: Section 2, Section 3.2, Section 3.3, Section 3.4, Section 5, and Section 6.

Reviewer 3 Report

Comments and Suggestions for Authors

I accept all the corrections done by the authors.

Author Response

Comment 1: I accept all the corrections done by the authors.

Response 1: Thank you for approving the revised version of the manuscript. We appreciate the time taken to conduct the review process on this manuscript.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

I have very carefully reviewed the latest revised version of the manuscript and I would like to thank the authors for finally taking the reviewer's constructive suggestions seriously and significantly improving the work. The paper now provides a much stronger critical synthesis and expert interpretation of the literature. I am fully satisfied with the current version and recommend the manuscript for publication.