INTRODUCTION

Sex dimorphism is the main concern as a first step of identifying victims in mass disasters. Ethnic group and other variables could influence the assignation of the correct sex. In principle, the patella seems not to be influenced by ancestry or postmortem processes; so is a perfect bone to take into account when investigating of unknown ancestry (^{Introna et al., 1998}).

Some studies have identified the patella as a potentially useful bone in relation to sexual dimorphism, although they show differences in mean values between males and females. Therefore, a systematic review is essential in order to collate all the empirical evidence that meets these inclusion criteria to answer some specific questions, minimize bias risk and support an objective approach to finding realistic conclusions (^{Antman et al., 1992}; ^{Oxman & Guyatt, 1993}).

To perform a meta-analysis in a very easy way, the Cochrane Consortium developed a software tool. Cochrane Consortium was defined as a “a global independent network of health practitioners, researchers, patient advocates and others, responding to the challenge of making the vast amounts of evidence generated through research useful for informing decisions about health” (^{Higgins et al., 2017}).

Thus, the aim of this study was to use meta-analytic techniques to test sexual dimorphism in patella depending on several techniques of measurement supported by the most important studies found by reviewing them systematically.

MATERIAL AND METHOD

This study is part of a PhD dissertation which is being performed at the University of Granada (Spain), by the first author. A systematic review was performed following the PRISMA statement (^{Hutton et al., 2016}). The next PICO questions were formulated to elucidate the problem of sexual dimorphism by means of patella measurements, following the previously published recommendations (^{Schardt et al., 2007}):

1. Are the patella measurements, carried out using several techniques (linear-classical, 3D or CBCT models), able to statistically discriminate between genders, in a healthy adult contemporary population of any given ethnic ancestry?

The variables considered in this review were previously defined by Introna et al., as shown in Table I:

Table I Definitions of the variables. 

Catálogo CISNE, Web of Science and Scopus were used as searching engines to find out the most relevant bibliography in an objective way. Catálogo CISNE is a library resource provided to their researchers by the Complutense University of Madrid. The key words “patella” AND “sex” were introduced, in all of them. Figure 1 shows a summary of the systematic searches following the PICO methodology recommendations. The meta-analysis assessment and discussion were performed using the software RevMan 5.3 (Review Manager (RevMan). Two kinds of meta-analysis were performed: one based on the all results (global), and the second one, grouping them according to measurement techniques (classical or using caliper, radiography, tomography and magnetic resonance).

Fig. 1 Summary of the findings following a systematic review. 

For any kind of explanation about meta-analysis terms, it is recommendable to read the materials and methods section published by ^{McFadden et al., (McFadden & Oxenham, 2018}). The type of proposed data was continuous with an inverse variance random model statistical method. The effect measurement was carried out using the mean difference between males and females according to the recommendations postulated by ^{Borenstein et al. (2010}). Tables II and III show the results by total and subtotals, with a 95 % C.I. for each study, and for the total C.I., indicating the number of male and female patellae analyzed on each study and on the total analysis and showing the mean difference between males and females, the standard deviation (SD), the weight of each study and their year of publication. Heterogeneity is represented by means of Tau2 and Chi2 values, freedom degree (with the associated p value) and I2 value, while a z test with its associated p value was performed to assess the global effect heterogeneity.

Table II Global meta-analysis table summary for the 3 measurements of the patella. 

Table III Meta-analysis table summary of the patella measurements divided by techniques. 

Risk of bias: The authors assessed the risk bias of the studies analyzing the next items: A) JCR journal?: whether the data had been extracted from JCR indexed journals or not. B) Age between 18 to 65?: When the sample includes individuals ranging from 18 to 65 years approx. C) Modern remains/bones?: Whether the bones/remains are modern or not. D) Dried bones?: When the data concerns dried bone. E) Inclusion and exclusion criteria?: When inclusion and exclusion criteria have been stated. F) Laterality?: if the studies refer to both laterality measurements analysis or otherwise, which side left or right these measurements are performed on. G) Intra-observer analysis?: if the repeatability of the measurements has been assessed. H) Inter-observer analysis?: in case other researchers have participated in data recording. I) Same n_male than N_female?: when approximately the same number of males and females have been included in the studies (with a maximum difference of 20 allowed).

The answers to the risk bias questions were classified in the figures, according to the following color codes:

+ AFFIRMATIVE answer.

There is not enough data to answer the question.

- NEGATIVE answer.

RESULTS AND DISCUSSION

The systematic review detected 17 studies to be included in the meta-analysis: ^{Akhlaghi et al. (2009}, 2010), ^{Bidmos et al. (2005}), ^{Dayal & Bidmos (2005}), Introna et al., ^{Kemkes-Grottenthaler (2005}), ^{Mahfouz et al. (2007}), ^{Michiue et al. (2018}), ^{Miller et al. (1996}), ^{Abdel Moneim et al. (2008}), ^{Olateju et al. (2013}), ^{Peckmann & Fisher (2018}), Peckmann et al. (2016), ^{Phoophalee & Riengrojpitak (2012}), ^{Sakaue (2008}), ^{Shang et al. (2014}), ^{Yasar Teke et al. (2018}) and ^{Yoo et al. (2007}). The main results are showed as a summary in Tables II and III as global results and classified by techniques. The number of male and female patellae, the mean difference (range) into the mean male and mean female, tau2, Chi2, I2 and z are showed as well. Figure 2 shows the forest plot for the global MH measurement. The I2 value of 93 % shows that the heterogeneity between studies is extremely high and it supposes that other variables could be relevant in the patella metrics. As it can be seen in all the figures, there are no complete standard conditions when it comes to comparing the studies. Risk bias appeared to be significant, so it is necessary to devote careful research to this. There are no studies in anthropology which take into consideration the research journals and dissertations; therefore, a new method was developed to try to solve the problem. The two most important risks observed were two: the type of population and the type of remains. Many studies were performed on fresh remains, not in dried bone, and included a mixed population with young or senescent people outside the age range of 18 to 65. On the other hand, very few articles consider the intra or inter-observer issues, in the case of patella measurements given by different authors, must be taken with caution, especially those that have not been published in JCR journals.

Fig. 2 Forest plot for the global MH measurement. A statistically mean difference between both sexes was found (p value < 0.00001) for a C.I. of 95 %. 

Figure 3 classifies the MH measurement into the four most relevant techniques. There are not enough results that can be considered definitive and further searches will be needed in order to improve the meta-analysis conclusions. First, better heterogeneity can be seen in the subgroups, with a higher value approx. 59 - 61 % for the radiography and caliper cases indicating a moderate heterogeneity. Only 2 cases were found using CT scan and 1 measured the height using MRI. The evaluation of homogeneity in both of these is not statistically relevant but, when it is studied among the 4 techniques subgroups, a statistical difference appeared (p value < 0.0001) that would indicate that the method of measurement can influence the height of the patella, although more studies must be performed to test that conclusion. It was observed that all the measurements carried out by radiography showed a minor mean difference between males and females (Table III).

Fig. 3 Forest plot for the MH measurement divided by techniques. Statistically mean differences between subgroups can be observed (p-value < 0.00001) for a C.I. of 95 %. 

Global MW measurement is represented in Figure 4 and separated into techniques in Figure 5. This shows a heterogeneity of 94 % in both cases which is high and indicates again that there were no standard conditions between studies. Only 1 study was found for the width measured with tomography. For this reason, heterogeneity value was not applicable in this case. Magnetic resonance width shows 2 studies, but Yasar Teke et al., divided the results into several age ranges concluding that age could influence the measurement. However, a very low sample size was considered into the groups.

Taphonomic processes are non-significant in modern patella remains or are very difficult to measure. Studies included in meta-analysis do not take into account the actual taphonomic processes in the patella. Therefore, a thorough discussion about it cannot be carried out.

MGH: Maximal Height Global, the maximal height of the patella taking into account all the studies; MWG: Maximal Width Global, the maximal widht of the patella taking into account all the results; Mean Difference [Range]: range in the mean difference between the mean males and the mean females. Results are in mm; df; (p): degree of freedom and p-value with a C.I. of 95 %.

Fig. 4 Forest plot for the global MW measurement. A statistically mean difference between both sexes was found (p value < 0.00001) for a C.I. of 95 %. 

Fig. 5 Forest plot for the MW measurement divided by techniques. Statistically mean differences between subgroups can be observed (p-value = 0.007) for a C.I. of 95 %. 

CONCLUSION

In this paper the patella height, width and thickness were successfully analyzed by means of a new systematic and meta-analysis approximation. Meta-analysis is useful in the anthropology field, allowing a new tool to manage related bibliography in an objective way and supply a better understanding of the most accurate measurement. In all the patella measurements cases, statistically significant differences between males and females were found globally. When the different studies were separated according to their techniques, more differences were found comparing each subgroup. Thus, the patella measurement techniques could influence the results and hence, more thorough investigations must be performed. In addition, the moderate-high heterogeneity showed a disparity in radiography values for height and width, indicating the importance of normalizing the patella measurement methods in order to perform better analysis and achieve more realistic conclusions.